Release: v4.23.0

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

.github/ISSUE_TEMPLATE/bug-report.yml

@@ -1,6 +1,5 @@
 name: "\U0001F41B Bug Report"
 description: Submit a bug report to help us improve transformers
-labels: [ "bug" ]
 body:
   - type: textarea
     id: system-info

.github/workflows/add-model-like.yml

@@ -41,10 +41,12 @@ jobs:
         run: |
           . ~/venv/bin/activate
           python setup.py develop
-          transformer_loc=$(pip show transformers | grep "Location: " | cut -c11-)
-          transformer_repo_loc=$(pwd .)
-          if [ "$transformer_loc" != "$transformer_repo_loc/src" ]; then
-              echo "transformers is from $transformer_loc but it shoud be from $transformer_repo_loc/src."
+          transformers_install=$(pip list -e | grep transformers)
+          transformers_install_array=($transformers_install)
+          transformers_loc=${transformers_install_array[-1]}
+          transformers_repo_loc=$(pwd .)
+          if [ "$transformers_loc" != "$transformers_repo_loc" ]; then
+              echo "transformers is from $transformers_loc but it shoud be from $transformers_repo_loc/src."
               echo "A fix is required. Stop testing."
               exit 1
           fi

.github/workflows/build-docker-images.yml

@@ -6,6 +6,10 @@ on:
       - docker-image*
   repository_dispatch:
   workflow_call:
+    inputs:
+      image_postfix:
+        required: true
+        type: string
   schedule:
     - cron: "0 1 * * *"
 
@@ -23,7 +27,7 @@ jobs:
         uses: docker/setup-buildx-action@v1
       -
         name: Check out code
-        uses: actions/checkout@v2
+        uses: actions/checkout@v3
       -
         name: Login to DockerHub
         uses: docker/login-action@v1
@@ -38,10 +42,12 @@ jobs:
           build-args: |
             REF=main
           push: true
-          tags: huggingface/transformers-all-latest-gpu
+          tags: huggingface/transformers-all-latest-gpu${{ inputs.image_postfix }}
 
   latest-with-torch-nightly-docker:
     name: "Nightly PyTorch + Stable TensorFlow"
+    # Push CI doesn't need this image
+    if: inputs.image_postfix != '-push-ci'
     runs-on: ubuntu-latest
     steps:
       -
@@ -91,10 +97,12 @@ jobs:
           build-args: |
             REF=main
           push: true
-          tags: huggingface/transformers-pytorch-deepspeed-latest-gpu
+          tags: huggingface/transformers-pytorch-deepspeed-latest-gpu${{ inputs.image_postfix }}
 
   nightly-torch-deepspeed-docker:
     name: "Nightly PyTorch + DeepSpeed"
+    # Push CI doesn't need this image
+    if: inputs.image_postfix != '-push-ci'
     runs-on: ubuntu-latest
     steps:
       -
@@ -121,6 +129,8 @@ jobs:
 
   doc-builder:
     name: "Doc builder"
+    # Push CI doesn't need this image
+    if: inputs.image_postfix != '-push-ci'
     runs-on: ubuntu-latest
     steps:
       -
@@ -145,6 +155,8 @@ jobs:
 
   latest-pytorch:
     name: "Latest PyTorch [dev]"
+    # Push CI doesn't need this image
+    if: inputs.image_postfix != '-push-ci'
     runs-on: ubuntu-latest
     steps:
       -
@@ -171,6 +183,8 @@ jobs:
 
   latest-tensorflow:
     name: "Latest TensorFlow [dev]"
+    # Push CI doesn't need this image
+    if: inputs.image_postfix != '-push-ci'
     runs-on: ubuntu-latest
     steps:
       -

.github/workflows/build_documentation.yml

@@ -15,6 +15,6 @@ jobs:
       commit_sha: ${{ github.sha }}
       package: transformers
       notebook_folder: transformers_doc
-      languages: en es it pt
+      languages: de en es it pt
     secrets:
       token: ${{ secrets.HUGGINGFACE_PUSH }}

.github/workflows/build_pr_documentation.yml

@@ -14,4 +14,4 @@ jobs:
       commit_sha: ${{ github.event.pull_request.head.sha }}
       pr_number: ${{ github.event.number }}
       package: transformers
-      languages: en es it pt
+      languages: de en es it pt

.github/workflows/check_runner_status.yml

@@ -0,0 +1,67 @@
+name: Self-hosted runner (check runner status)
+
+# Note that each job's dependencies go into a corresponding docker file.
+#
+# For example for `run_all_tests_torch_cuda_extensions_gpu` the docker image is
+# `huggingface/transformers-pytorch-deepspeed-latest-gpu`, which can be found at
+# `docker/transformers-pytorch-deepspeed-latest-gpu/Dockerfile`
+
+on:
+  repository_dispatch:
+  schedule:
+    # run per hour
+    - cron: "0 */1 * * *"
+
+env:
+  TRANSFORMERS_IS_CI: yes
+
+jobs:
+  check_runner_status:
+    name: Check Runner Status
+    runs-on: ubuntu-latest
+    outputs:
+      offline_runners: ${{ steps.set-offline_runners.outputs.offline_runners }}
+    steps:
+      - name: Checkout transformers
+        uses: actions/checkout@v2
+        with:
+          fetch-depth: 2
+
+      - name: Check Runner Status
+        run: python utils/check_self_hosted_runner.py --target_runners single-gpu-ci-runner-docker,multi-gpu-ci-runner-docker,single-gpu-scheduled-ci-runner-docker,multi-scheduled-scheduled-ci-runner-docker,single-gpu-doctest-ci-runner-docker --token ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
+
+      - id: set-offline_runners
+        name: Set output for offline runners
+        if: ${{ always() }}
+        run: |
+          offline_runners=$(python3 -c 'fp = open("offline_runners.txt"); failed = fp.read(); fp.close(); print(failed)')
+          echo "::set-output name=offline_runners::$offline_runners"
+
+  send_results:
+    name: Send results to webhook
+    runs-on: ubuntu-latest
+    needs: check_runner_status
+    if: ${{ failure() }}
+    steps:
+      - name: Preliminary job status
+        shell: bash
+        run: |
+          echo "Runner availability: ${{ needs.check_runner_status.result }}"
+
+      - uses: actions/checkout@v2
+      - uses: actions/download-artifact@v2
+      - name: Send message to Slack
+        env:
+          CI_SLACK_BOT_TOKEN: ${{ secrets.CI_SLACK_BOT_TOKEN }}
+          CI_SLACK_CHANNEL_ID: ${{ secrets.CI_SLACK_CHANNEL_ID }}
+          CI_SLACK_CHANNEL_ID_DAILY: ${{ secrets.CI_SLACK_CHANNEL_ID_DAILY }}
+          CI_SLACK_CHANNEL_DUMMY_TESTS: ${{ secrets.CI_SLACK_CHANNEL_DUMMY_TESTS }}
+          CI_SLACK_REPORT_CHANNEL_ID: ${{ secrets.CI_SLACK_CHANNEL_ID_DAILY }}
+          CI_EVENT: runner status check
+          RUNNER_STATUS: ${{ needs.check_runner_status.result }}
+          OFFLINE_RUNNERS: ${{ needs.check_runner_status.outputs.offline_runners }}
+        # We pass `needs.setup.outputs.matrix` as the argument. A processing in `notification_service.py` to change
+        # `models/bert` to `models_bert` is required, as the artifact names use `_` instead of `/`.
+        run: |
+          pip install slack_sdk
+          python utils/notification_service.py

.github/workflows/self-nightly-scheduled.yml

@@ -8,8 +8,9 @@ name: Self-hosted runner (nightly)
 
 on:
   repository_dispatch:
-  schedule:
-    - cron: "0 16 * * *"
+# Disable temporarily until the test suite can be run under 12 hours.
+#  schedule:
+#    - cron: "0 16 * * *"
 
 env:
   HF_HOME: /mnt/cache
@@ -22,8 +23,36 @@ env:
   RUN_PT_TF_CROSS_TESTS: 1
 
 jobs:
+  check_runner_status:
+    name: Check Runner Status
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout transformers
+        uses: actions/checkout@v2
+        with:
+          fetch-depth: 2
+
+      - name: Check Runner Status
+        run: python utils/check_self_hosted_runner.py --target_runners single-gpu-scheduled-ci-runner-docker,multi-gpu-scheduled-ci-runner-docker --token ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
+
+  check_runners:
+    name: Check Runners
+    needs: check_runner_status
+    strategy:
+      matrix:
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: ${{ format('{0}-{1}', matrix.machine_type, 'docker') }}
+    container:
+      image: huggingface/transformers-all-latest-torch-nightly-gpu
+      options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    steps:
+      - name: NVIDIA-SMI
+        run: |
+          nvidia-smi
+
   setup:
     name: Setup
+    needs: check_runners
     strategy:
       matrix:
         machine_type: [single-gpu, multi-gpu]
@@ -178,6 +207,9 @@ jobs:
         working-directory: /workspace/transformers
         run: git fetch && git checkout ${{ github.sha }}
 
+      - name: Remove cached torch extensions
+        run: rm -rf /github/home/.cache/torch_extensions/
+
       # To avoid unknown test failures
       - name: Pre build DeepSpeed *again*
         working-directory: /workspace
@@ -185,7 +217,7 @@ jobs:
           python3 -m pip uninstall -y deepspeed
           rm -rf DeepSpeed
           git clone https://github.com/microsoft/DeepSpeed && cd DeepSpeed && rm -rf build
-          DS_BUILD_CPU_ADAM=1 DS_BUILD_AIO=1 DS_BUILD_UTILS=1 python3 -m pip install . --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check
+          DS_BUILD_CPU_ADAM=1 DS_BUILD_FUSED_ADAM=1 DS_BUILD_AIO=1 DS_BUILD_UTILS=1 python3 -m pip install . --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check
 
       - name: NVIDIA-SMI
         run: |
@@ -217,8 +249,23 @@ jobs:
     name: Send results to webhook
     runs-on: ubuntu-latest
     if: always()
-    needs: [setup, run_tests_single_gpu, run_tests_multi_gpu, run_all_tests_torch_cuda_extensions_gpu]
+    needs: [
+      check_runner_status,
+      check_runners,
+      setup,
+      run_tests_single_gpu,
+      run_tests_multi_gpu,
+      run_all_tests_torch_cuda_extensions_gpu
+    ]
     steps:
+      - name: Preliminary job status
+        shell: bash
+        # For the meaning of these environment variables, see the job `Setup`
+        run: |
+          echo "Runner availability: ${{ needs.check_runner_status.result }}"
+          echo "Runner status: ${{ needs.check_runners.result }}"
+          echo "Setup status: ${{ needs.setup.result }}"
+
       - uses: actions/checkout@v2
       - uses: actions/download-artifact@v2
       - name: Send message to Slack
@@ -229,6 +276,9 @@ jobs:
           CI_SLACK_CHANNEL_DUMMY_TESTS: ${{ secrets.CI_SLACK_CHANNEL_DUMMY_TESTS }}
           CI_SLACK_REPORT_CHANNEL_ID: ${{ secrets.CI_SLACK_CHANNEL_ID_PAST_FUTURE }}
           CI_EVENT: nightly-build
+          RUNNER_STATUS: ${{ needs.check_runner_status.result }}
+          RUNNER_ENV_STATUS: ${{ needs.check_runners.result }}
+          SETUP_STATUS: ${{ needs.setup.result }}
         # We pass `needs.setup.outputs.matrix` as the argument. A processing in `notification_service.py` to change
         # `models/bert` to `models_bert` is required, as the artifact names use `_` instead of `/`.
         run: |

.github/workflows/self-past.yml

@@ -15,6 +15,11 @@ on:
       version:
         required: true
         type: string
+      # Use this to control the commit to test against
+      sha:
+        default: 'main'
+        required: false
+        type: string
 
 env:
   HF_HOME: /mnt/cache
@@ -27,24 +32,59 @@ env:
   RUN_PT_TF_CROSS_TESTS: 1
 
 jobs:
-  setup:
-    name: Setup
+  check_runner_status:
+    name: Check Runner Status
     runs-on: ubuntu-latest
-    outputs:
-      matrix: ${{ steps.set-matrix.outputs.matrix }}
     steps:
       - name: Checkout transformers
         uses: actions/checkout@v2
         with:
           fetch-depth: 2
 
+      - name: Check Runner Status
+        run: python utils/check_self_hosted_runner.py --target_runners single-gpu-past-ci-runner-docker,multi-gpu-past-ci-runner-docker --token ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
+
+  check_runners:
+    name: Check Runners
+    needs: check_runner_status
+    strategy:
+      matrix:
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: ${{ format('{0}-{1}', matrix.machine_type, 'docker-past-ci') }}
+    container:
+      image: huggingface/transformers-${{ inputs.framework }}-past-${{ inputs.version }}-gpu
+      options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    steps:
+      - name: NVIDIA-SMI
+        run: |
+          nvidia-smi
+
+  setup:
+    name: Setup
+    needs: check_runners
+    strategy:
+      matrix:
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: ${{ format('{0}-{1}', matrix.machine_type, 'docker-past-ci') }}
+    container:
+      image: huggingface/transformers-${{ inputs.framework }}-past-${{ inputs.version }}-gpu
+      options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    outputs:
+      matrix: ${{ steps.set-matrix.outputs.matrix }}
+    steps:
+      - name: Update clone
+        working-directory: /transformers
+        run: git fetch && git checkout ${{ inputs.sha }}
+
       - name: Cleanup
+        working-directory: /transformers
         run: |
           rm -rf tests/__pycache__
           rm -rf tests/models/__pycache__
           rm -rf reports
 
       - id: set-matrix
+        working-directory: /transformers
         name: Identify models to test
         run: |
           cd tests
@@ -65,7 +105,7 @@ jobs:
     steps:
       - name: Update clone
         working-directory: /transformers
-        run: git fetch && git checkout ${{ github.sha }}
+        run: git fetch && git checkout ${{ inputs.sha }}
 
       - name: Echo folder ${{ matrix.folders }}
         shell: bash
@@ -96,6 +136,15 @@ jobs:
         continue-on-error: true
         run: cat /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}/failures_short.txt
 
+      - name: Save job name
+        if: ${{ always() }}
+        shell: bash
+        run: |
+          matrix_folders=${matrix_folders/'models_'/'models/'}
+          job_name="Model tests ($matrix_folders, ${{ matrix.machine_type }})"
+          echo "$job_name"
+          echo "$job_name" > /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}/job_name.txt
+
       - name: Test suite reports artifacts
         if: ${{ always() }}
         uses: actions/upload-artifact@v2
@@ -118,7 +167,7 @@ jobs:
     steps:
       - name: Update clone
         working-directory: /transformers
-        run: git fetch && git checkout ${{ github.sha }}
+        run: git fetch && git checkout ${{ inputs.sha }}
 
       - name: Echo folder ${{ matrix.folders }}
         shell: bash
@@ -149,6 +198,15 @@ jobs:
         continue-on-error: true
         run: cat /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}/failures_short.txt
 
+      - name: Save job name
+        if: ${{ always() }}
+        shell: bash
+        run: |
+          matrix_folders=${matrix_folders/'models_'/'models/'}
+          job_name="Model tests ($matrix_folders, ${{ matrix.machine_type }})"
+          echo "$job_name"
+          echo "$job_name" > /transformers/reports/${{ matrix.machine_type }}_tests_gpu_${{ matrix.folders }}/job_name.txt
+
       - name: Test suite reports artifacts
         if: ${{ always() }}
         uses: actions/upload-artifact@v2
@@ -160,8 +218,16 @@ jobs:
     name: Send results to webhook
     runs-on: ubuntu-latest
     if: always()
-    needs: [setup, run_tests_single_gpu, run_tests_multi_gpu]
+    needs: [check_runner_status, check_runners, setup, run_tests_single_gpu, run_tests_multi_gpu]
     steps:
+      - name: Preliminary job status
+        shell: bash
+        # For the meaning of these environment variables, see the job `Setup`
+        run: |
+          echo "Runner availability: ${{ needs.check_runner_status.result }}"
+          echo "Runner status: ${{ needs.check_runners.result }}"
+          echo "Setup status: ${{ needs.setup.result }}"
+
       - uses: actions/checkout@v2
       - uses: actions/download-artifact@v2
 
@@ -177,6 +243,9 @@ jobs:
           CI_SLACK_CHANNEL_DUMMY_TESTS: ${{ secrets.CI_SLACK_CHANNEL_DUMMY_TESTS }}
           CI_SLACK_REPORT_CHANNEL_ID: ${{ secrets.CI_SLACK_CHANNEL_ID_PAST_FUTURE }}
           CI_EVENT: Past CI - ${{ inputs.framework }}-${{ inputs.version }}
+          RUNNER_STATUS: ${{ needs.check_runner_status.result }}
+          RUNNER_ENV_STATUS: ${{ needs.check_runners.result }}
+          SETUP_STATUS: ${{ needs.setup.result }}
         # We pass `needs.setup.outputs.matrix` as the argument. A processing in `notification_service.py` to change
         # `models/bert` to `models_bert` is required, as the artifact names use `_` instead of `/`.
         run: |

.github/workflows/self-push-caller.yml

@@ -40,6 +40,8 @@ jobs:
     needs: check-for-setup
     if: (github.event_name == 'push') && (needs.check-for-setup.outputs.changed == '1')
     uses: ./.github/workflows/build-docker-images.yml
+    with:
+      image_postfix: "-push-ci"
     secrets: inherit
 
   run_push_ci:

.github/workflows/self-push.yml

@@ -27,9 +27,43 @@ env:
   RUN_PT_TF_CROSS_TESTS: 1
 
 jobs:
+  check_runner_status:
+    name: Check Runner Status
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout transformers
+        uses: actions/checkout@v2
+        with:
+          fetch-depth: 2
+
+      - name: Check Runner Status
+        run: python utils/check_self_hosted_runner.py --target_runners single-gpu-ci-runner-docker,multi-gpu-ci-runner-docker --token ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
+
+  check_runners:
+    name: Check Runners
+    needs: check_runner_status
+    strategy:
+      matrix:
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: [self-hosted, docker-gpu, '${{ matrix.machine_type }}']
+    container:
+      image: huggingface/transformers-all-latest-gpu-push-ci
+      options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    steps:
+      - name: NVIDIA-SMI
+        run: |
+          nvidia-smi
+
   setup:
     name: Setup
-    runs-on: ubuntu-latest
+    needs: check_runners
+    strategy:
+      matrix:
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: [self-hosted, docker-gpu, '${{ matrix.machine_type }}']
+    container:
+      image: huggingface/transformers-all-latest-gpu-push-ci
+      options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
     outputs:
       matrix: ${{ steps.set-matrix.outputs.matrix }}
       test_map: ${{ steps.set-matrix.outputs.test_map }}
@@ -62,12 +96,8 @@ jobs:
           echo "env.CI_BRANCH = ${{ env.CI_BRANCH }}"
           echo "env.CI_SHA = ${{ env.CI_SHA }}"
 
-      - name: Checkout transformers
-        uses: actions/checkout@v2
-        with:
-          fetch-depth: 2
-
       - name: Update clone using environment variables
+        working-directory: /transformers
         run: |
           echo "original branch = $(git branch --show-current)"
           git fetch && git checkout ${{ env.CI_BRANCH }}
@@ -76,12 +106,14 @@ jobs:
           echo "log = $(git log -n 1)"
 
       - name: Cleanup
+        working-directory: /transformers
         run: |
           rm -rf tests/__pycache__
           rm -rf tests/models/__pycache__
           rm -rf reports
 
       - name: Fetch the tests to run
+        working-directory: /transformers
         # TODO: add `git-python` in the docker images
         run: |
           pip install --upgrade git-python
@@ -91,10 +123,11 @@ jobs:
         uses: actions/upload-artifact@v2
         with:
           name: test_fetched
-          path: test_preparation.txt
+          path: /transformers/test_preparation.txt
 
       - id: set-matrix
         name: Organize tests into models
+        working-directory: /transformers
         # The `keys` is used as GitHub actions matrix for jobs, i.e. `models/bert`, `tokenization`, `pipeline`, etc.
         # The `test_map` is used to get the actual identified test files under each key.
         # If no test to run (so no `test_map.json` file), create a dummy map (empty matrix will fail)
@@ -123,7 +156,7 @@ jobs:
         machine_type: [single-gpu]
     runs-on: [self-hosted, docker-gpu, '${{ matrix.machine_type }}']
     container:
-      image: huggingface/transformers-all-latest-gpu
+      image: huggingface/transformers-all-latest-gpu-push-ci
       options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
     steps:
       # Necessary to get the correct branch name and commit SHA for `workflow_run` event
@@ -208,7 +241,7 @@ jobs:
         machine_type: [multi-gpu]
     runs-on: [self-hosted, docker-gpu, '${{ matrix.machine_type }}']
     container:
-      image: huggingface/transformers-all-latest-gpu
+      image: huggingface/transformers-all-latest-gpu-push-ci
       options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
     steps:
       # Necessary to get the correct branch name and commit SHA for `workflow_run` event
@@ -293,7 +326,7 @@ jobs:
         machine_type: [single-gpu]
     runs-on: [self-hosted, docker-gpu, '${{ matrix.machine_type }}']
     container:
-      image: huggingface/transformers-pytorch-deepspeed-latest-gpu
+      image: huggingface/transformers-pytorch-deepspeed-latest-gpu-push-ci
       options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
     steps:
       # Necessary to get the correct branch name and commit SHA for `workflow_run` event
@@ -328,12 +361,15 @@ jobs:
           git checkout ${{ env.CI_SHA }}
           echo "log = $(git log -n 1)"
 
+      - name: Remove cached torch extensions
+        run: rm -rf /github/home/.cache/torch_extensions/
+
       # To avoid unknown test failures
       - name: Pre build DeepSpeed *again*
         working-directory: /workspace
         run: |
           python3 -m pip uninstall -y deepspeed
-          DS_BUILD_CPU_ADAM=1 DS_BUILD_AIO=1 DS_BUILD_UTILS=1 python3 -m pip install deepspeed --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check
+          DS_BUILD_CPU_ADAM=1 DS_BUILD_FUSED_ADAM=1 DS_BUILD_AIO=1 DS_BUILD_UTILS=1 python3 -m pip install deepspeed --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check
 
       - name: NVIDIA-SMI
         run: |
@@ -372,7 +408,7 @@ jobs:
         machine_type: [multi-gpu]
     runs-on: [self-hosted, docker-gpu, '${{ matrix.machine_type }}']
     container:
-      image: huggingface/transformers-pytorch-deepspeed-latest-gpu
+      image: huggingface/transformers-pytorch-deepspeed-latest-gpu-push-ci
       options: --gpus all --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
     steps:
       # Necessary to get the correct branch name and commit SHA for `workflow_run` event
@@ -407,12 +443,15 @@ jobs:
           git checkout ${{ env.CI_SHA }}
           echo "log = $(git log -n 1)"
 
+      - name: Remove cached torch extensions
+        run: rm -rf /github/home/.cache/torch_extensions/
+
       # To avoid unknown test failures
       - name: Pre build DeepSpeed *again*
         working-directory: /workspace
         run: |
           python3 -m pip uninstall -y deepspeed
-          DS_BUILD_CPU_ADAM=1 DS_BUILD_AIO=1 DS_BUILD_UTILS=1 python3 -m pip install deepspeed --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check
+          DS_BUILD_CPU_ADAM=1 DS_BUILD_FUSED_ADAM=1 DS_BUILD_AIO=1 DS_BUILD_UTILS=1 python3 -m pip install deepspeed --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check
 
       - name: NVIDIA-SMI
         run: |
@@ -446,6 +485,8 @@ jobs:
     runs-on: ubuntu-latest
     if: always()
     needs: [
+        check_runner_status,
+        check_runners,
         setup,
         run_tests_single_gpu,
         run_tests_multi_gpu,
@@ -453,6 +494,14 @@ jobs:
         run_tests_torch_cuda_extensions_multi_gpu
     ]
     steps:
+      - name: Preliminary job status
+        shell: bash
+        # For the meaning of these environment variables, see the job `Setup`
+        run: |
+          echo "Runner availability: ${{ needs.check_runner_status.result }}"
+          echo "Setup status: ${{ needs.setup.result }}"
+          echo "Runner status: ${{ needs.check_runners.result }}"
+
       # Necessary to get the correct branch name and commit SHA for `workflow_run` event
       # We also take into account the `push` event (we might want to test some changes in a branch)
       - name: Prepare custom environment variables
@@ -498,6 +547,10 @@ jobs:
           CI_TITLE_PUSH: ${{ github.event.head_commit.message }}
           CI_TITLE_WORKFLOW_RUN: ${{ github.event.workflow_run.head_commit.message }}
           CI_SHA: ${{ env.CI_SHA }}
+          RUNNER_STATUS: ${{ needs.check_runner_status.result }}
+          RUNNER_ENV_STATUS: ${{ needs.check_runners.result }}
+          SETUP_STATUS: ${{ needs.setup.result }}
+
         # We pass `needs.setup.outputs.matrix` as the argument. A processing in `notification_service.py` to change
         # `models/bert` to `models_bert` is required, as the artifact names use `_` instead of `/`.
         run: |

.github/workflows/self-scheduled.yml

@@ -22,8 +22,36 @@ env:
   RUN_PT_TF_CROSS_TESTS: 1
 
 jobs:
+  check_runner_status:
+    name: Check Runner Status
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout transformers
+        uses: actions/checkout@v2
+        with:
+          fetch-depth: 2
+
+      - name: Check Runner Status
+        run: python utils/check_self_hosted_runner.py --target_runners single-gpu-scheduled-ci-runner-docker,multi-gpu-scheduled-ci-runner-docker --token ${{ secrets.ACCESS_REPO_INFO_TOKEN }}
+
+  check_runners:
+    name: Check Runners
+    needs: check_runner_status
+    strategy:
+      matrix:
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: ${{ format('{0}-{1}', matrix.machine_type, 'docker') }}
+    container:
+      image: huggingface/transformers-all-latest-gpu
+      options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    steps:
+      - name: NVIDIA-SMI
+        run: |
+          nvidia-smi
+
   setup:
     name: Setup
+    needs: check_runners
     strategy:
       matrix:
         machine_type: [single-gpu, multi-gpu]
@@ -187,19 +215,19 @@ jobs:
         working-directory: /transformers
         run: |
           pip install -r examples/pytorch/_tests_requirements.txt
-          python3 -m pytest -v --make-reports=examples_gpu examples/pytorch
+          python3 -m pytest -v --make-reports=single-gpu_examples_gpu examples/pytorch
 
       - name: Failure short reports
         if: ${{ failure() }}
         continue-on-error: true
-        run: cat /transformers/reports/examples_gpu/failures_short.txt
+        run: cat /transformers/reports/single-gpu_examples_gpu/failures_short.txt
 
       - name: Test suite reports artifacts
         if: ${{ always() }}
         uses: actions/upload-artifact@v2
         with:
-          name: run_examples_gpu
-          path: /transformers/reports/examples_gpu
+          name: single-gpu_run_examples_gpu
+          path: /transformers/reports/single-gpu_examples_gpu
 
   run_pipelines_torch_gpu:
     name: PyTorch pipelines
@@ -228,10 +256,8 @@ jobs:
 
       - name: Run all pipeline tests on GPU
         working-directory: /transformers
-        env:
-          RUN_PIPELINE_TESTS: yes
         run: |
-          python3 -m pytest -n 1 -v --dist=loadfile -m is_pipeline_test --make-reports=${{ matrix.machine_type }}_tests_torch_pipeline_gpu tests
+          python3 -m pytest -n 1 -v --dist=loadfile --make-reports=${{ matrix.machine_type }}_tests_torch_pipeline_gpu tests/pipelines
 
       - name: Failure short reports
         if: ${{ failure() }}
@@ -273,10 +299,8 @@ jobs:
 
       - name: Run all pipeline tests on GPU
         working-directory: /transformers
-        env:
-          RUN_PIPELINE_TESTS: yes
         run: |
-          python3 -m pytest -n 1 -v --dist=loadfile -m is_pipeline_test --make-reports=${{ matrix.machine_type }}_tests_tf_pipeline_gpu tests
+          python3 -m pytest -n 1 -v --dist=loadfile --make-reports=${{ matrix.machine_type }}_tests_tf_pipeline_gpu tests/pipelines
 
       - name: Failure short reports
         if: ${{ always() }}
@@ -306,12 +330,15 @@ jobs:
         working-directory: /workspace/transformers
         run: git fetch && git checkout ${{ github.sha }}
 
+      - name: Remove cached torch extensions
+        run: rm -rf /github/home/.cache/torch_extensions/
+
       # To avoid unknown test failures
       - name: Pre build DeepSpeed *again*
         working-directory: /workspace
         run: |
           python3 -m pip uninstall -y deepspeed
-          DS_BUILD_CPU_ADAM=1 DS_BUILD_AIO=1 DS_BUILD_UTILS=1 python3 -m pip install deepspeed --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check
+          DS_BUILD_CPU_ADAM=1 DS_BUILD_FUSED_ADAM=1 DS_BUILD_AIO=1 DS_BUILD_UTILS=1 python3 -m pip install deepspeed --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check
 
       - name: NVIDIA-SMI
         run: |
@@ -343,8 +370,26 @@ jobs:
     name: Send results to webhook
     runs-on: ubuntu-latest
     if: always()
-    needs: [setup, run_tests_single_gpu, run_tests_multi_gpu, run_examples_gpu, run_pipelines_tf_gpu, run_pipelines_torch_gpu, run_all_tests_torch_cuda_extensions_gpu]
+    needs: [
+      check_runner_status,
+      check_runners,
+      setup,
+      run_tests_single_gpu,
+      run_tests_multi_gpu,
+      run_examples_gpu,
+      run_pipelines_tf_gpu,
+      run_pipelines_torch_gpu,
+      run_all_tests_torch_cuda_extensions_gpu
+    ]
     steps:
+      - name: Preliminary job status
+        shell: bash
+        # For the meaning of these environment variables, see the job `Setup`
+        run: |
+          echo "Runner availability: ${{ needs.check_runner_status.result }}"
+          echo "Runner status: ${{ needs.check_runners.result }}"
+          echo "Setup status: ${{ needs.setup.result }}"
+
       - uses: actions/checkout@v2
       - uses: actions/download-artifact@v2
       - name: Send message to Slack
@@ -355,6 +400,9 @@ jobs:
           CI_SLACK_CHANNEL_DUMMY_TESTS: ${{ secrets.CI_SLACK_CHANNEL_DUMMY_TESTS }}
           CI_SLACK_REPORT_CHANNEL_ID: ${{ secrets.CI_SLACK_CHANNEL_ID_DAILY }}
           CI_EVENT: scheduled
+          RUNNER_STATUS: ${{ needs.check_runner_status.result }}
+          RUNNER_ENV_STATUS: ${{ needs.check_runners.result }}
+          SETUP_STATUS: ${{ needs.setup.result }}
         # We pass `needs.setup.outputs.matrix` as the argument. A processing in `notification_service.py` to change
         # `models/bert` to `models_bert` is required, as the artifact names use `_` instead of `/`.
         run: |

.github/workflows/stale.yml

@@ -12,10 +12,10 @@ jobs:
     env:
       GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
     steps:
-    - uses: actions/checkout@v2
+    - uses: actions/checkout@v3
 
     - name: Setup Python
-      uses: actions/setup-python@v1
+      uses: actions/setup-python@v4
       with:
         python-version: 3.7
 
@@ -24,4 +24,4 @@ jobs:
         pip install PyGithub
     - name: Close stale issues
       run: |
-        python scripts/stale.py
+        python scripts/stale.py

CONTRIBUTING.md

@@ -171,6 +171,14 @@ Follow these steps to start contributing ([supported Python versions](https://gi
 
    If you have already cloned that repo, you might need to `git pull` to get the most recent changes in the `datasets`
    library.
+   
+   Depending on your OS, you might need to install some external libraries, as well, if the `pip` installation fails.
+   
+   For macOS, you will likely need [MeCab](https://taku910.github.io/mecab/), which can be installed from Homebrew:
+   
+   ```bash
+   brew install mecab
+   ```
 
 5. Develop the features on your branch.
 

Makefile

@@ -41,6 +41,7 @@ repo-consistency:
 	python utils/check_inits.py
 	python utils/check_config_docstrings.py
 	python utils/tests_fetcher.py --sanity_check
+	python utils/update_metadata.py --check-only
 
 # this target runs checks on all files
 

README.md

@@ -55,13 +55,13 @@ limitations under the License.
     <a href="https://hf.co/course"><img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/course_banner.png"></a>
 </h3>
 
-🤗 Transformers provides thousands of pretrained models to perform tasks on different modalities such as text, vision, and audio. 
+🤗 Transformers provides thousands of pretrained models to perform tasks on different modalities such as text, vision, and audio.
 
 These models can be applied on:
 
-* 📝 Text, for tasks like text classification, information extraction, question answering, summarization, translation, text generation, in over 100 languages. 
-* 🖼️ Images, for tasks like image classification, object detection, and segmentation. 
-* 🗣️ Audio, for tasks like speech recognition and audio classification. 
+* 📝 Text, for tasks like text classification, information extraction, question answering, summarization, translation, text generation, in over 100 languages.
+* 🖼️ Images, for tasks like image classification, object detection, and segmentation.
+* 🗣️ Audio, for tasks like speech recognition and audio classification.
 
 Transformer models can also perform tasks on **several modalities combined**, such as table question answering, optical character recognition, information extraction from scanned documents, video classification, and visual question answering.
 
@@ -87,12 +87,16 @@ Here are a few examples:
 In Computer Vision:
 - [Image classification with ViT](https://huggingface.co/google/vit-base-patch16-224)
 - [Object Detection with DETR](https://huggingface.co/facebook/detr-resnet-50)
-- [Image Segmentation with DETR](https://huggingface.co/facebook/detr-resnet-50-panoptic)
+- [Semantic Segmentation with SegFormer](https://huggingface.co/nvidia/segformer-b0-finetuned-ade-512-512)
+- [Panoptic Segmentation with DETR](https://huggingface.co/facebook/detr-resnet-50-panoptic)
 
 In Audio:
 - [Automatic Speech Recognition with Wav2Vec2](https://huggingface.co/facebook/wav2vec2-base-960h)
 - [Keyword Spotting with Wav2Vec2](https://huggingface.co/superb/wav2vec2-base-superb-ks)
 
+In Multimodal tasks:
+- [Visual Question Answering with ViLT](https://huggingface.co/dandelin/vilt-b32-finetuned-vqa)
+
 **[Write With Transformer](https://transformer.huggingface.co)**, built by the Hugging Face team, is the official demo of this repo’s text generation capabilities.
 
 ## If you are looking for custom support from the Hugging Face team
@@ -157,7 +161,7 @@ Here we get a list of objects detected in the image, with a box surrounding the
 
 You can learn more about the tasks supported by the `pipeline` API in [this tutorial](https://huggingface.co/docs/transformers/task_summary).
 
-To download and use any of the pretrained models on your given task, all it takes is three lines of code. Here is the PyTorch version:
+In addition to `pipeline`, to download and use any of the pretrained models on your given task, all it takes is three lines of code. Here is the PyTorch version:
 ```python
 >>> from transformers import AutoTokenizer, AutoModel
 
@@ -181,7 +185,7 @@ And here is the equivalent code for TensorFlow:
 
 The tokenizer is responsible for all the preprocessing the pretrained model expects, and can be called directly on a single string (as in the above examples) or a list. It will output a dictionary that you can use in downstream code or simply directly pass to your model using the ** argument unpacking operator.
 
-The model itself is a regular [Pytorch `nn.Module`](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) or a [TensorFlow `tf.keras.Model`](https://www.tensorflow.org/api_docs/python/tf/keras/Model) (depending on your backend) which you can use normally. [This tutorial](https://huggingface.co/docs/transformers/training) explains how to integrate such a model into a classic PyTorch or TensorFlow training loop, or how to use our `Trainer` API to quickly fine-tune on a new dataset.
+The model itself is a regular [Pytorch `nn.Module`](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) or a [TensorFlow `tf.keras.Model`](https://www.tensorflow.org/api_docs/python/tf/keras/Model) (depending on your backend) which you can use as usual. [This tutorial](https://huggingface.co/docs/transformers/training) explains how to integrate such a model into a classic PyTorch or TensorFlow training loop, or how to use our `Trainer` API to quickly fine-tune on a new dataset.
 
 ## Why should I use transformers?
 
@@ -194,7 +198,7 @@ The model itself is a regular [Pytorch `nn.Module`](https://pytorch.org/docs/sta
 1. Lower compute costs, smaller carbon footprint:
     - Researchers can share trained models instead of always retraining.
     - Practitioners can reduce compute time and production costs.
-    - Dozens of architectures with over 20,000 pretrained models, some in more than 100 languages.
+    - Dozens of architectures with over 60,000 pretrained models across all modalities.
 
 1. Choose the right framework for every part of a model's lifetime:
     - Train state-of-the-art models in 3 lines of code.
@@ -209,7 +213,7 @@ The model itself is a regular [Pytorch `nn.Module`](https://pytorch.org/docs/sta
 ## Why shouldn't I use transformers?
 
 - This library is not a modular toolbox of building blocks for neural nets. The code in the model files is not refactored with additional abstractions on purpose, so that researchers can quickly iterate on each of the models without diving into additional abstractions/files.
-- The training API is not intended to work on any model but is optimized to work with the models provided by the library. For generic machine learning loops, you should use another library.
+- The training API is not intended to work on any model but is optimized to work with the models provided by the library. For generic machine learning loops, you should use another library (possibly, [Accelerate](https://huggingface.co/docs/accelerate)).
 - While we strive to present as many use cases as possible, the scripts in our [examples folder](https://github.com/huggingface/transformers/tree/main/examples) are just that: examples. It is expected that they won't work out-of-the box on your specific problem and that you will be required to change a few lines of code to adapt them to your needs.
 
 ## Installation
@@ -245,6 +249,8 @@ conda install -c huggingface transformers
 
 Follow the installation pages of Flax, PyTorch or TensorFlow to see how to install them with conda.
 
+> **_NOTE:_**  On Windows, you may be prompted to activate Developer Mode in order to benefit from caching. If this is not an option for you, please let us know in [this issue](https://github.com/huggingface/huggingface_hub/issues/1062).
+
 ## Model architectures
 
 **[All the model checkpoints](https://huggingface.co/models)** provided by 🤗 Transformers are seamlessly integrated from the huggingface.co [model hub](https://huggingface.co) where they are uploaded directly by [users](https://huggingface.co/users) and [organizations](https://huggingface.co/organizations).
@@ -272,6 +278,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 1. **[CANINE](https://huggingface.co/docs/transformers/model_doc/canine)** (from Google Research) released with the paper [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) by Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting.
 1. **[CLIP](https://huggingface.co/docs/transformers/model_doc/clip)** (from OpenAI) released with the paper [Learning Transferable Visual Models From Natural Language Supervision](https://arxiv.org/abs/2103.00020) by Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, Ilya Sutskever.
 1. **[CodeGen](https://huggingface.co/docs/transformers/model_doc/codegen)** (from Salesforce) released with the paper [A Conversational Paradigm for Program Synthesis](https://arxiv.org/abs/2203.13474) by Erik Nijkamp, Bo Pang, Hiroaki Hayashi, Lifu Tu, Huan Wang, Yingbo Zhou, Silvio Savarese, Caiming Xiong.
+1. **[Conditional DETR](https://huggingface.co/docs/transformers/model_doc/conditional_detr)** (from Microsoft Research Asia) released with the paper [Conditional DETR for Fast Training Convergence](https://arxiv.org/abs/2108.06152) by Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, Jingdong Wang.
 1. **[ConvBERT](https://huggingface.co/docs/transformers/model_doc/convbert)** (from YituTech) released with the paper [ConvBERT: Improving BERT with Span-based Dynamic Convolution](https://arxiv.org/abs/2008.02496) by Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan.
 1. **[ConvNeXT](https://huggingface.co/docs/transformers/model_doc/convnext)** (from Facebook AI) released with the paper [A ConvNet for the 2020s](https://arxiv.org/abs/2201.03545) by Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie.
 1. **[CPM](https://huggingface.co/docs/transformers/model_doc/cpm)** (from Tsinghua University) released with the paper [CPM: A Large-scale Generative Chinese Pre-trained Language Model](https://arxiv.org/abs/2012.00413) by Zhengyan Zhang, Xu Han, Hao Zhou, Pei Ke, Yuxian Gu, Deming Ye, Yujia Qin, Yusheng Su, Haozhe Ji, Jian Guan, Fanchao Qi, Xiaozhi Wang, Yanan Zheng, Guoyang Zeng, Huanqi Cao, Shengqi Chen, Daixuan Li, Zhenbo Sun, Zhiyuan Liu, Minlie Huang, Wentao Han, Jie Tang, Juanzi Li, Xiaoyan Zhu, Maosong Sun.
@@ -281,15 +288,19 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 1. **[DeBERTa](https://huggingface.co/docs/transformers/model_doc/deberta)** (from Microsoft) released with the paper [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://arxiv.org/abs/2006.03654) by Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen.
 1. **[DeBERTa-v2](https://huggingface.co/docs/transformers/model_doc/deberta-v2)** (from Microsoft) released with the paper [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://arxiv.org/abs/2006.03654) by Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen.
 1. **[Decision Transformer](https://huggingface.co/docs/transformers/model_doc/decision_transformer)** (from Berkeley/Facebook/Google) released with the paper [Decision Transformer: Reinforcement Learning via Sequence Modeling](https://arxiv.org/abs/2106.01345) by Lili Chen, Kevin Lu, Aravind Rajeswaran, Kimin Lee, Aditya Grover, Michael Laskin, Pieter Abbeel, Aravind Srinivas, Igor Mordatch.
+1. **[Deformable DETR](https://huggingface.co/docs/transformers/model_doc/deformable_detr)** (from SenseTime Research) released with the paper [Deformable DETR: Deformable Transformers for End-to-End Object Detection](https://arxiv.org/abs/2010.04159) by Xizhou Zhu, Weijie Su, Lewei Lu, Bin Li, Xiaogang Wang, Jifeng Dai.
 1. **[DeiT](https://huggingface.co/docs/transformers/model_doc/deit)** (from Facebook) released with the paper [Training data-efficient image transformers & distillation through attention](https://arxiv.org/abs/2012.12877) by Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, Hervé Jégou.
 1. **[DETR](https://huggingface.co/docs/transformers/model_doc/detr)** (from Facebook) released with the paper [End-to-End Object Detection with Transformers](https://arxiv.org/abs/2005.12872) by Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, Sergey Zagoruyko.
 1. **[DialoGPT](https://huggingface.co/docs/transformers/model_doc/dialogpt)** (from Microsoft Research) released with the paper [DialoGPT: Large-Scale Generative Pre-training for Conversational Response Generation](https://arxiv.org/abs/1911.00536) by Yizhe Zhang, Siqi Sun, Michel Galley, Yen-Chun Chen, Chris Brockett, Xiang Gao, Jianfeng Gao, Jingjing Liu, Bill Dolan.
 1. **[DistilBERT](https://huggingface.co/docs/transformers/model_doc/distilbert)** (from HuggingFace), released together with the paper [DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter](https://arxiv.org/abs/1910.01108) by Victor Sanh, Lysandre Debut and Thomas Wolf. The same method has been applied to compress GPT2 into [DistilGPT2](https://github.com/huggingface/transformers/tree/main/examples/research_projects/distillation), RoBERTa into [DistilRoBERTa](https://github.com/huggingface/transformers/tree/main/examples/research_projects/distillation), Multilingual BERT into [DistilmBERT](https://github.com/huggingface/transformers/tree/main/examples/research_projects/distillation) and a German version of DistilBERT.
 1. **[DiT](https://huggingface.co/docs/transformers/model_doc/dit)** (from Microsoft Research) released with the paper [DiT: Self-supervised Pre-training for Document Image Transformer](https://arxiv.org/abs/2203.02378) by Junlong Li, Yiheng Xu, Tengchao Lv, Lei Cui, Cha Zhang, Furu Wei.
+1. **[Donut](https://huggingface.co/docs/transformers/model_doc/donut)** (from NAVER), released together with the paper [OCR-free Document Understanding Transformer](https://arxiv.org/abs/2111.15664) by Geewook Kim, Teakgyu Hong, Moonbin Yim, Jeongyeon Nam, Jinyoung Park, Jinyeong Yim, Wonseok Hwang, Sangdoo Yun, Dongyoon Han, Seunghyun Park.
 1. **[DPR](https://huggingface.co/docs/transformers/model_doc/dpr)** (from Facebook) released with the paper [Dense Passage Retrieval for Open-Domain Question Answering](https://arxiv.org/abs/2004.04906) by Vladimir Karpukhin, Barlas Oğuz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih.
 1. **[DPT](https://huggingface.co/docs/transformers/master/model_doc/dpt)** (from Intel Labs) released with the paper [Vision Transformers for Dense Prediction](https://arxiv.org/abs/2103.13413) by René Ranftl, Alexey Bochkovskiy, Vladlen Koltun.
 1. **[ELECTRA](https://huggingface.co/docs/transformers/model_doc/electra)** (from Google Research/Stanford University) released with the paper [ELECTRA: Pre-training text encoders as discriminators rather than generators](https://arxiv.org/abs/2003.10555) by Kevin Clark, Minh-Thang Luong, Quoc V. Le, Christopher D. Manning.
 1. **[EncoderDecoder](https://huggingface.co/docs/transformers/model_doc/encoder-decoder)** (from Google Research) released with the paper [Leveraging Pre-trained Checkpoints for Sequence Generation Tasks](https://arxiv.org/abs/1907.12461) by Sascha Rothe, Shashi Narayan, Aliaksei Severyn.
+1. **[ERNIE](https://huggingface.co/docs/transformers/model_doc/ernie)** (from Baidu) released with the paper [ERNIE: Enhanced Representation through Knowledge Integration](https://arxiv.org/abs/1904.09223) by Yu Sun, Shuohuan Wang, Yukun Li, Shikun Feng, Xuyi Chen, Han Zhang, Xin Tian, Danxiang Zhu, Hao Tian, Hua Wu.
+1. **[ESM](https://huggingface.co/docs/transformers/model_doc/esm)** (from Meta AI) are transformer protein language models.  **ESM-1b** was released with the paper [Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences](https://www.pnas.org/content/118/15/e2016239118) by Alexander Rives, Joshua Meier, Tom Sercu, Siddharth Goyal, Zeming Lin, Jason Liu, Demi Guo, Myle Ott, C. Lawrence Zitnick, Jerry Ma, and Rob Fergus. **ESM-1v** was released with the paper [Language models enable zero-shot prediction of the effects of mutations on protein function](https://doi.org/10.1101/2021.07.09.450648) by Joshua Meier, Roshan Rao, Robert Verkuil, Jason Liu, Tom Sercu and Alexander Rives. **ESM-2** was released with the paper [Language models of protein sequences at the scale of evolution enable accurate structure prediction](https://doi.org/10.1101/2022.07.20.500902) by Zeming Lin, Halil Akin, Roshan Rao, Brian Hie, Zhongkai Zhu, Wenting Lu, Allan dos Santos Costa, Maryam Fazel-Zarandi, Tom Sercu, Sal Candido, Alexander Rives.
 1. **[FlauBERT](https://huggingface.co/docs/transformers/model_doc/flaubert)** (from CNRS) released with the paper [FlauBERT: Unsupervised Language Model Pre-training for French](https://arxiv.org/abs/1912.05372) by Hang Le, Loïc Vial, Jibril Frej, Vincent Segonne, Maximin Coavoux, Benjamin Lecouteux, Alexandre Allauzen, Benoît Crabbé, Laurent Besacier, Didier Schwab.
 1. **[FLAVA](https://huggingface.co/docs/transformers/model_doc/flava)** (from Facebook AI) released with the paper [FLAVA: A Foundational Language And Vision Alignment Model](https://arxiv.org/abs/2112.04482) by Amanpreet Singh, Ronghang Hu, Vedanuj Goswami, Guillaume Couairon, Wojciech Galuba, Marcus Rohrbach, and Douwe Kiela.
 1. **[FNet](https://huggingface.co/docs/transformers/model_doc/fnet)** (from Google Research) released with the paper [FNet: Mixing Tokens with Fourier Transforms](https://arxiv.org/abs/2105.03824) by James Lee-Thorp, Joshua Ainslie, Ilya Eckstein, Santiago Ontanon.
@@ -298,6 +309,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 1. **[GPT](https://huggingface.co/docs/transformers/model_doc/openai-gpt)** (from OpenAI) released with the paper [Improving Language Understanding by Generative Pre-Training](https://blog.openai.com/language-unsupervised/) by Alec Radford, Karthik Narasimhan, Tim Salimans and Ilya Sutskever.
 1. **[GPT Neo](https://huggingface.co/docs/transformers/model_doc/gpt_neo)** (from EleutherAI) released in the repository [EleutherAI/gpt-neo](https://github.com/EleutherAI/gpt-neo) by Sid Black, Stella Biderman, Leo Gao, Phil Wang and Connor Leahy.
 1. **[GPT NeoX](https://huggingface.co/docs/transformers/model_doc/gpt_neox)** (from EleutherAI) released with the paper [GPT-NeoX-20B: An Open-Source Autoregressive Language Model](https://arxiv.org/abs/2204.06745) by Sid Black, Stella Biderman, Eric Hallahan, Quentin Anthony, Leo Gao, Laurence Golding, Horace He, Connor Leahy, Kyle McDonell, Jason Phang, Michael Pieler, USVSN Sai Prashanth, Shivanshu Purohit, Laria Reynolds, Jonathan Tow, Ben Wang, Samuel Weinbach
+1. **[GPT NeoX Japanese](https://huggingface.co/docs/transformers/model_doc/gpt_neox_japanese)** (from ABEJA) released by Shinya Otani, Takayoshi Makabe, Anuj Arora, and Kyo Hattori.
 1. **[GPT-2](https://huggingface.co/docs/transformers/model_doc/gpt2)** (from OpenAI) released with the paper [Language Models are Unsupervised Multitask Learners](https://blog.openai.com/better-language-models/) by Alec Radford*, Jeffrey Wu*, Rewon Child, David Luan, Dario Amodei** and Ilya Sutskever**.
 1. **[GPT-J](https://huggingface.co/docs/transformers/model_doc/gptj)** (from EleutherAI) released in the repository [kingoflolz/mesh-transformer-jax](https://github.com/kingoflolz/mesh-transformer-jax/) by Ben Wang and Aran Komatsuzaki.
 1. **[GroupViT](https://huggingface.co/docs/transformers/model_doc/groupvit)** (from UCSD, NVIDIA) released with the paper [GroupViT: Semantic Segmentation Emerges from Text Supervision](https://arxiv.org/abs/2202.11094) by Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, Xiaolong Wang.
@@ -307,7 +319,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 1. **[LayoutLM](https://huggingface.co/docs/transformers/model_doc/layoutlm)** (from Microsoft Research Asia) released with the paper [LayoutLM: Pre-training of Text and Layout for Document Image Understanding](https://arxiv.org/abs/1912.13318) by Yiheng Xu, Minghao Li, Lei Cui, Shaohan Huang, Furu Wei, Ming Zhou.
 1. **[LayoutLMv2](https://huggingface.co/docs/transformers/model_doc/layoutlmv2)** (from Microsoft Research Asia) released with the paper [LayoutLMv2: Multi-modal Pre-training for Visually-Rich Document Understanding](https://arxiv.org/abs/2012.14740) by Yang Xu, Yiheng Xu, Tengchao Lv, Lei Cui, Furu Wei, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Wanxiang Che, Min Zhang, Lidong Zhou.
 1. **[LayoutLMv3](https://huggingface.co/docs/transformers/model_doc/layoutlmv3)** (from Microsoft Research Asia) released with the paper [LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking](https://arxiv.org/abs/2204.08387) by Yupan Huang, Tengchao Lv, Lei Cui, Yutong Lu, Furu Wei.
-1. **[LayoutXLM](https://huggingface.co/docs/transformers/model_doc/layoutlmv2)** (from Microsoft Research Asia) released with the paper [LayoutXLM: Multimodal Pre-training for Multilingual Visually-rich Document Understanding](https://arxiv.org/abs/2104.08836) by Yiheng Xu, Tengchao Lv, Lei Cui, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Furu Wei.
+1. **[LayoutXLM](https://huggingface.co/docs/transformers/model_doc/layoutxlm)** (from Microsoft Research Asia) released with the paper [LayoutXLM: Multimodal Pre-training for Multilingual Visually-rich Document Understanding](https://arxiv.org/abs/2104.08836) by Yiheng Xu, Tengchao Lv, Lei Cui, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Furu Wei.
 1. **[LED](https://huggingface.co/docs/transformers/model_doc/led)** (from AllenAI) released with the paper [Longformer: The Long-Document Transformer](https://arxiv.org/abs/2004.05150) by Iz Beltagy, Matthew E. Peters, Arman Cohan.
 1. **[LeViT](https://huggingface.co/docs/transformers/model_doc/levit)** (from Meta AI) released with the paper [LeViT: A Vision Transformer in ConvNet's Clothing for Faster Inference](https://arxiv.org/abs/2104.01136) by Ben Graham, Alaaeldin El-Nouby, Hugo Touvron, Pierre Stock, Armand Joulin, Hervé Jégou, Matthijs Douze.
 1. **[Longformer](https://huggingface.co/docs/transformers/model_doc/longformer)** (from AllenAI) released with the paper [Longformer: The Long-Document Transformer](https://arxiv.org/abs/2004.05150) by Iz Beltagy, Matthew E. Peters, Arman Cohan.
@@ -317,6 +329,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 1. **[M-CTC-T](https://huggingface.co/docs/transformers/model_doc/mctct)** (from Facebook) released with the paper [Pseudo-Labeling For Massively Multilingual Speech Recognition](https://arxiv.org/abs/2111.00161) by Loren Lugosch, Tatiana Likhomanenko, Gabriel Synnaeve, and Ronan Collobert.
 1. **[M2M100](https://huggingface.co/docs/transformers/model_doc/m2m_100)** (from Facebook) released with the paper [Beyond English-Centric Multilingual Machine Translation](https://arxiv.org/abs/2010.11125) by Angela Fan, Shruti Bhosale, Holger Schwenk, Zhiyi Ma, Ahmed El-Kishky, Siddharth Goyal, Mandeep Baines, Onur Celebi, Guillaume Wenzek, Vishrav Chaudhary, Naman Goyal, Tom Birch, Vitaliy Liptchinsky, Sergey Edunov, Edouard Grave, Michael Auli, Armand Joulin.
 1. **[MarianMT](https://huggingface.co/docs/transformers/model_doc/marian)** Machine translation models trained using [OPUS](http://opus.nlpl.eu/) data by Jörg Tiedemann. The [Marian Framework](https://marian-nmt.github.io/) is being developed by the Microsoft Translator Team.
+1. **[MarkupLM](https://huggingface.co/docs/transformers/model_doc/markuplm)** (from Microsoft Research Asia) released with the paper [MarkupLM: Pre-training of Text and Markup Language for Visually-rich Document Understanding](https://arxiv.org/abs/2110.08518) by Junlong Li, Yiheng Xu, Lei Cui, Furu Wei.
 1. **[MaskFormer](https://huggingface.co/docs/transformers/model_doc/maskformer)** (from Meta and UIUC) released with the paper [Per-Pixel Classification is Not All You Need for Semantic Segmentation](https://arxiv.org/abs/2107.06278) by Bowen Cheng, Alexander G. Schwing, Alexander Kirillov.
 1. **[mBART](https://huggingface.co/docs/transformers/model_doc/mbart)** (from Facebook) released with the paper [Multilingual Denoising Pre-training for Neural Machine Translation](https://arxiv.org/abs/2001.08210) by Yinhan Liu, Jiatao Gu, Naman Goyal, Xian Li, Sergey Edunov, Marjan Ghazvininejad, Mike Lewis, Luke Zettlemoyer.
 1. **[mBART-50](https://huggingface.co/docs/transformers/model_doc/mbart)** (from Facebook) released with the paper [Multilingual Translation with Extensible Multilingual Pretraining and Finetuning](https://arxiv.org/abs/2008.00401) by Yuqing Tang, Chau Tran, Xian Li, Peng-Jen Chen, Naman Goyal, Vishrav Chaudhary, Jiatao Gu, Angela Fan.
@@ -334,6 +347,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 1. **[OPT](https://huggingface.co/docs/transformers/master/model_doc/opt)** (from Meta AI) released with the paper [OPT: Open Pre-trained Transformer Language Models](https://arxiv.org/abs/2205.01068) by Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen et al.
 1. **[OWL-ViT](https://huggingface.co/docs/transformers/model_doc/owlvit)** (from Google AI) released with the paper [Simple Open-Vocabulary Object Detection with Vision Transformers](https://arxiv.org/abs/2205.06230) by Matthias Minderer, Alexey Gritsenko, Austin Stone, Maxim Neumann, Dirk Weissenborn, Alexey Dosovitskiy, Aravindh Mahendran, Anurag Arnab, Mostafa Dehghani, Zhuoran Shen, Xiao Wang, Xiaohua Zhai, Thomas Kipf, and Neil Houlsby.
 1. **[Pegasus](https://huggingface.co/docs/transformers/model_doc/pegasus)** (from Google) released with the paper [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu.
+1. **[PEGASUS-X](https://huggingface.co/docs/transformers/model_doc/pegasus_x)** (from Google) released with the paper [Investigating Efficiently Extending Transformers for Long Input Summarization](https://arxiv.org/abs/2208.04347) by Jason Phang, Yao Zhao, and Peter J. Liu.
 1. **[Perceiver IO](https://huggingface.co/docs/transformers/model_doc/perceiver)** (from Deepmind) released with the paper [Perceiver IO: A General Architecture for Structured Inputs & Outputs](https://arxiv.org/abs/2107.14795) by Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier Hénaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, João Carreira.
 1. **[PhoBERT](https://huggingface.co/docs/transformers/model_doc/phobert)** (from VinAI Research) released with the paper [PhoBERT: Pre-trained language models for Vietnamese](https://www.aclweb.org/anthology/2020.findings-emnlp.92/) by Dat Quoc Nguyen and Anh Tuan Nguyen.
 1. **[PLBart](https://huggingface.co/docs/transformers/model_doc/plbart)** (from UCLA NLP) released with the paper [Unified Pre-training for Program Understanding and Generation](https://arxiv.org/abs/2103.06333) by Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, Kai-Wei Chang.
@@ -356,11 +370,12 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 1. **[Splinter](https://huggingface.co/docs/transformers/model_doc/splinter)** (from Tel Aviv University), released together with the paper [Few-Shot Question Answering by Pretraining Span Selection](https://arxiv.org/abs/2101.00438) by Ori Ram, Yuval Kirstain, Jonathan Berant, Amir Globerson, Omer Levy.
 1. **[SqueezeBERT](https://huggingface.co/docs/transformers/model_doc/squeezebert)** (from Berkeley) released with the paper [SqueezeBERT: What can computer vision teach NLP about efficient neural networks?](https://arxiv.org/abs/2006.11316) by Forrest N. Iandola, Albert E. Shaw, Ravi Krishna, and Kurt W. Keutzer.
 1. **[Swin Transformer](https://huggingface.co/docs/transformers/model_doc/swin)** (from Microsoft) released with the paper [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://arxiv.org/abs/2103.14030) by Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, Baining Guo.
-1. **[Swin Transformer V2](https://huggingface.co/docs/transformers/main/model_doc/swinv2)** (from Microsoft) released with the paper [Swin Transformer V2: Scaling Up Capacity and Resolution](https://arxiv.org/abs/2111.09883) by Ze Liu, Han Hu, Yutong Lin, Zhuliang Yao, Zhenda Xie, Yixuan Wei, Jia Ning, Yue Cao, Zheng Zhang, Li Dong, Furu Wei, Baining Guo.
+1. **[Swin Transformer V2](https://huggingface.co/docs/transformers/model_doc/swinv2)** (from Microsoft) released with the paper [Swin Transformer V2: Scaling Up Capacity and Resolution](https://arxiv.org/abs/2111.09883) by Ze Liu, Han Hu, Yutong Lin, Zhuliang Yao, Zhenda Xie, Yixuan Wei, Jia Ning, Yue Cao, Zheng Zhang, Li Dong, Furu Wei, Baining Guo.
 1. **[T5](https://huggingface.co/docs/transformers/model_doc/t5)** (from Google AI) released with the paper [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://arxiv.org/abs/1910.10683) by Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu.
 1. **[T5v1.1](https://huggingface.co/docs/transformers/model_doc/t5v1.1)** (from Google AI) released in the repository [google-research/text-to-text-transfer-transformer](https://github.com/google-research/text-to-text-transfer-transformer/blob/main/released_checkpoints.md#t511) by Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu.
 1. **[TAPAS](https://huggingface.co/docs/transformers/model_doc/tapas)** (from Google AI) released with the paper [TAPAS: Weakly Supervised Table Parsing via Pre-training](https://arxiv.org/abs/2004.02349) by Jonathan Herzig, Paweł Krzysztof Nowak, Thomas Müller, Francesco Piccinno and Julian Martin Eisenschlos.
 1. **[TAPEX](https://huggingface.co/docs/transformers/model_doc/tapex)** (from Microsoft Research) released with the paper [TAPEX: Table Pre-training via Learning a Neural SQL Executor](https://arxiv.org/abs/2107.07653) by Qian Liu, Bei Chen, Jiaqi Guo, Morteza Ziyadi, Zeqi Lin, Weizhu Chen, Jian-Guang Lou.
+1. **[Time Series Transformer](https://huggingface.co/docs/transformers/model_doc/time_series_transformer)**  (from HuggingFace).
 1. **[Trajectory Transformer](https://huggingface.co/docs/transformers/model_doc/trajectory_transformers)** (from the University of California at Berkeley) released with the paper [Offline Reinforcement Learning as One Big Sequence Modeling Problem](https://arxiv.org/abs/2106.02039) by Michael Janner, Qiyang Li, Sergey Levine
 1. **[Transformer-XL](https://huggingface.co/docs/transformers/model_doc/transfo-xl)** (from Google/CMU) released with the paper [Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context](https://arxiv.org/abs/1901.02860) by Zihang Dai*, Zhilin Yang*, Yiming Yang, Jaime Carbonell, Quoc V. Le, Ruslan Salakhutdinov.
 1. **[TrOCR](https://huggingface.co/docs/transformers/model_doc/trocr)** (from Microsoft), released together with the paper [TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models](https://arxiv.org/abs/2109.10282) by Minghao Li, Tengchao Lv, Lei Cui, Yijuan Lu, Dinei Florencio, Cha Zhang, Zhoujun Li, Furu Wei.
@@ -368,14 +383,18 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 1. **[UniSpeech](https://huggingface.co/docs/transformers/model_doc/unispeech)** (from Microsoft Research) released with the paper [UniSpeech: Unified Speech Representation Learning with Labeled and Unlabeled Data](https://arxiv.org/abs/2101.07597) by Chengyi Wang, Yu Wu, Yao Qian, Kenichi Kumatani, Shujie Liu, Furu Wei, Michael Zeng, Xuedong Huang.
 1. **[UniSpeechSat](https://huggingface.co/docs/transformers/model_doc/unispeech-sat)** (from Microsoft Research) released with the paper [UNISPEECH-SAT: UNIVERSAL SPEECH REPRESENTATION LEARNING WITH SPEAKER AWARE PRE-TRAINING](https://arxiv.org/abs/2110.05752) by Sanyuan Chen, Yu Wu, Chengyi Wang, Zhengyang Chen, Zhuo Chen, Shujie Liu, Jian Wu, Yao Qian, Furu Wei, Jinyu Li, Xiangzhan Yu.
 1. **[VAN](https://huggingface.co/docs/transformers/model_doc/van)** (from Tsinghua University and Nankai University) released with the paper [Visual Attention Network](https://arxiv.org/abs/2202.09741) by Meng-Hao Guo, Cheng-Ze Lu, Zheng-Ning Liu, Ming-Ming Cheng, Shi-Min Hu.
+1. **[VideoMAE](https://huggingface.co/docs/transformers/model_doc/videomae)** (from Multimedia Computing Group, Nanjing University) released with the paper [VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training](https://arxiv.org/abs/2203.12602) by Zhan Tong, Yibing Song, Jue Wang, Limin Wang.
 1. **[ViLT](https://huggingface.co/docs/transformers/model_doc/vilt)** (from NAVER AI Lab/Kakao Enterprise/Kakao Brain) released with the paper [ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision](https://arxiv.org/abs/2102.03334) by Wonjae Kim, Bokyung Son, Ildoo Kim.
 1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.
 1. **[VisualBERT](https://huggingface.co/docs/transformers/model_doc/visual_bert)** (from UCLA NLP) released with the paper [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) by Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang.
 1. **[ViTMAE](https://huggingface.co/docs/transformers/model_doc/vit_mae)** (from Meta AI) released with the paper [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) by Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick.
+1. **[ViTMSN](https://huggingface.co/docs/transformers/model_doc/vit_msn)** (from Meta AI) released with the paper [Masked Siamese Networks for Label-Efficient Learning](https://arxiv.org/abs/2204.07141) by Mahmoud Assran, Mathilde Caron, Ishan Misra, Piotr Bojanowski, Florian Bordes, Pascal Vincent, Armand Joulin, Michael Rabbat, Nicolas Ballas.
 1. **[Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/wav2vec2)** (from Facebook AI) released with the paper [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) by Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli.
 1. **[Wav2Vec2-Conformer](https://huggingface.co/docs/transformers/model_doc/wav2vec2-conformer)** (from Facebook AI) released with the paper [FAIRSEQ S2T: Fast Speech-to-Text Modeling with FAIRSEQ](https://arxiv.org/abs/2010.05171) by Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Sravya Popuri, Dmytro Okhonko, Juan Pino.
 1. **[Wav2Vec2Phoneme](https://huggingface.co/docs/transformers/model_doc/wav2vec2_phoneme)** (from Facebook AI) released with the paper [Simple and Effective Zero-shot Cross-lingual Phoneme Recognition](https://arxiv.org/abs/2109.11680) by Qiantong Xu, Alexei Baevski, Michael Auli.
 1. **[WavLM](https://huggingface.co/docs/transformers/model_doc/wavlm)** (from Microsoft Research) released with the paper [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://arxiv.org/abs/2110.13900) by Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng, Furu Wei.
+1. **[Whisper](https://huggingface.co/docs/transformers/model_doc/whisper)** (from OpenAI) released with the paper [Robust Speech Recognition via Large-Scale Weak Supervision](https://cdn.openai.com/papers/whisper.pdf) by Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, Ilya Sutskever.
+1. **[X-CLIP](https://huggingface.co/docs/transformers/model_doc/xclip)** (from Microsoft Research) released with the paper [Expanding Language-Image Pretrained Models for General Video Recognition](https://arxiv.org/abs/2208.02816) by Bolin Ni, Houwen Peng, Minghao Chen, Songyang Zhang, Gaofeng Meng, Jianlong Fu, Shiming Xiang, Haibin Ling.
 1. **[XGLM](https://huggingface.co/docs/transformers/model_doc/xglm)** (From Facebook AI) released with the paper [Few-shot Learning with Multilingual Language Models](https://arxiv.org/abs/2112.10668) by Xi Victoria Lin, Todor Mihaylov, Mikel Artetxe, Tianlu Wang, Shuohui Chen, Daniel Simig, Myle Ott, Naman Goyal, Shruti Bhosale, Jingfei Du, Ramakanth Pasunuru, Sam Shleifer, Punit Singh Koura, Vishrav Chaudhary, Brian O'Horo, Jeff Wang, Luke Zettlemoyer, Zornitsa Kozareva, Mona Diab, Veselin Stoyanov, Xian Li.
 1. **[XLM](https://huggingface.co/docs/transformers/model_doc/xlm)** (from Facebook) released together with the paper [Cross-lingual Language Model Pretraining](https://arxiv.org/abs/1901.07291) by Guillaume Lample and Alexis Conneau.
 1. **[XLM-ProphetNet](https://huggingface.co/docs/transformers/model_doc/xlm-prophetnet)** (from Microsoft Research) released with the paper [ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training](https://arxiv.org/abs/2001.04063) by Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou.
@@ -390,7 +409,7 @@ Current number of checkpoints: ![](https://img.shields.io/endpoint?url=https://h
 
 To check if each model has an implementation in Flax, PyTorch or TensorFlow, or has an associated tokenizer backed by the 🤗 Tokenizers library, refer to [this table](https://huggingface.co/docs/transformers/index#supported-frameworks).
 
-These implementations have been tested on several datasets (see the example scripts) and should match the performance of the original implementations. You can find more details on performance in the Examples section of the [documentation](https://huggingface.co/docs/transformers/examples).
+These implementations have been tested on several datasets (see the example scripts) and should match the performance of the original implementations. You can find more details on performance in the Examples section of the [documentation](https://github.com/huggingface/transformers/tree/main/examples).
 
 
 ## Learn more

README_ko.md

@@ -59,7 +59,7 @@ limitations under the License.
 
 🤗 Transformers는 이러한 사전학습 모델을 빠르게 다운로드해 특정 텍스트에 사용하고, 원하는 데이터로 fine-tuning해 커뮤니티나 우리의 [모델 허브](https://huggingface.co/models)에 공유할 수 있도록 API를 제공합니다. 또한, 모델 구조를 정의하는 각 파이썬 모듈은 완전히 독립적이여서 연구 실험을 위해 손쉽게 수정할 수 있습니다.
 
-🤗 Transformers는 가장 유명한 3개의 딥러닝 라이브러리를 지원합니다. 이들은 서로 완벽히 연동됩니다 — [Jax](https://jax.readthedocs.io/en/latest/), [PyTorch](https://pytorch.org/), [TensorFlow](https://www.tensorflow.org/). 간단하게 이 라이브러리 중 하나로 모델을 학습하고, 또 다른 라이브러리로 추론을 위해 모델을 불러올 수 있습니다. 
+🤗 Transformers는 가장 유명한 3개의 딥러닝 라이브러리를 지원합니다. 이들은 서로 완벽히 연동됩니다 — [Jax](https://jax.readthedocs.io/en/latest/), [PyTorch](https://pytorch.org/), [TensorFlow](https://www.tensorflow.org/). 간단하게 이 라이브러리 중 하나로 모델을 학습하고, 또 다른 라이브러리로 추론을 위해 모델을 불러올 수 있습니다.
 
 ## 온라인 데모
 
@@ -74,7 +74,7 @@ limitations under the License.
 - [DistilBERT를 이용한 질문 답변](https://huggingface.co/distilbert-base-uncased-distilled-squad?text=Which+name+is+also+used+to+describe+the+Amazon+rainforest+in+English%3F&context=The+Amazon+rainforest+%28Portuguese%3A+Floresta+Amaz%C3%B4nica+or+Amaz%C3%B4nia%3B+Spanish%3A+Selva+Amaz%C3%B3nica%2C+Amazon%C3%ADa+or+usually+Amazonia%3B+French%3A+For%C3%AAt+amazonienne%3B+Dutch%3A+Amazoneregenwoud%29%2C+also+known+in+English+as+Amazonia+or+the+Amazon+Jungle%2C+is+a+moist+broadleaf+forest+that+covers+most+of+the+Amazon+basin+of+South+America.+This+basin+encompasses+7%2C000%2C000+square+kilometres+%282%2C700%2C000+sq+mi%29%2C+of+which+5%2C500%2C000+square+kilometres+%282%2C100%2C000+sq+mi%29+are+covered+by+the+rainforest.+This+region+includes+territory+belonging+to+nine+nations.+The+majority+of+the+forest+is+contained+within+Brazil%2C+with+60%25+of+the+rainforest%2C+followed+by+Peru+with+13%25%2C+Colombia+with+10%25%2C+and+with+minor+amounts+in+Venezuela%2C+Ecuador%2C+Bolivia%2C+Guyana%2C+Suriname+and+French+Guiana.+States+or+departments+in+four+nations+contain+%22Amazonas%22+in+their+names.+The+Amazon+represents+over+half+of+the+planet%27s+remaining+rainforests%2C+and+comprises+the+largest+and+most+biodiverse+tract+of+tropical+rainforest+in+the+world%2C+with+an+estimated+390+billion+individual+trees+divided+into+16%2C000+species)
 - [T5로 번역하기](https://huggingface.co/t5-base?text=My+name+is+Wolfgang+and+I+live+in+Berlin)
 
-**[Transformer와 글쓰기](https://transformer.huggingface.co)** 는 이 저장소의 텍스트 생성 능력에 관한 Hugging Face 팀의 공식 데모입니다. 
+**[Transformer와 글쓰기](https://transformer.huggingface.co)** 는 이 저장소의 텍스트 생성 능력에 관한 Hugging Face 팀의 공식 데모입니다.
 
 ## Hugging Face 팀의 커스텀 지원을 원한다면
 
@@ -228,6 +228,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[CANINE](https://huggingface.co/docs/transformers/model_doc/canine)** (from Google Research) released with the paper [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) by Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting.
 1. **[CLIP](https://huggingface.co/docs/transformers/model_doc/clip)** (from OpenAI) released with the paper [Learning Transferable Visual Models From Natural Language Supervision](https://arxiv.org/abs/2103.00020) by Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, Ilya Sutskever.
 1. **[CodeGen](https://huggingface.co/docs/transformers/model_doc/codegen)** (from Salesforce) released with the paper [A Conversational Paradigm for Program Synthesis](https://arxiv.org/abs/2203.13474) by Erik Nijkamp, Bo Pang, Hiroaki Hayashi, Lifu Tu, Huan Wang, Yingbo Zhou, Silvio Savarese, Caiming Xiong.
+1. **[Conditional DETR](https://huggingface.co/docs/transformers/model_doc/conditional_detr)** (from Microsoft Research Asia) released with the paper [Conditional DETR for Fast Training Convergence](https://arxiv.org/abs/2108.06152) by Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, Jingdong Wang.
 1. **[ConvBERT](https://huggingface.co/docs/transformers/model_doc/convbert)** (from YituTech) released with the paper [ConvBERT: Improving BERT with Span-based Dynamic Convolution](https://arxiv.org/abs/2008.02496) by Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan.
 1. **[ConvNeXT](https://huggingface.co/docs/transformers/model_doc/convnext)** (from Facebook AI) released with the paper [A ConvNet for the 2020s](https://arxiv.org/abs/2201.03545) by Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie.
 1. **[CPM](https://huggingface.co/docs/transformers/model_doc/cpm)** (from Tsinghua University) released with the paper [CPM: A Large-scale Generative Chinese Pre-trained Language Model](https://arxiv.org/abs/2012.00413) by Zhengyan Zhang, Xu Han, Hao Zhou, Pei Ke, Yuxian Gu, Deming Ye, Yujia Qin, Yusheng Su, Haozhe Ji, Jian Guan, Fanchao Qi, Xiaozhi Wang, Yanan Zheng, Guoyang Zeng, Huanqi Cao, Shengqi Chen, Daixuan Li, Zhenbo Sun, Zhiyuan Liu, Minlie Huang, Wentao Han, Jie Tang, Juanzi Li, Xiaoyan Zhu, Maosong Sun.
@@ -237,15 +238,19 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[DeBERTa](https://huggingface.co/docs/transformers/model_doc/deberta)** (from Microsoft) released with the paper [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://arxiv.org/abs/2006.03654) by Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen.
 1. **[DeBERTa-v2](https://huggingface.co/docs/transformers/model_doc/deberta-v2)** (from Microsoft) released with the paper [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://arxiv.org/abs/2006.03654) by Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen.
 1. **[Decision Transformer](https://huggingface.co/docs/transformers/model_doc/decision_transformer)** (from Berkeley/Facebook/Google) released with the paper [Decision Transformer: Reinforcement Learning via Sequence Modeling](https://arxiv.org/abs/2106.01345) by Lili Chen, Kevin Lu, Aravind Rajeswaran, Kimin Lee, Aditya Grover, Michael Laskin, Pieter Abbeel, Aravind Srinivas, Igor Mordatch.
+1. **[Deformable DETR](https://huggingface.co/docs/transformers/model_doc/deformable_detr)** (from SenseTime Research) released with the paper [Deformable DETR: Deformable Transformers for End-to-End Object Detection](https://arxiv.org/abs/2010.04159) by Xizhou Zhu, Weijie Su, Lewei Lu, Bin Li, Xiaogang Wang, Jifeng Dai.
 1. **[DeiT](https://huggingface.co/docs/transformers/model_doc/deit)** (from Facebook) released with the paper [Training data-efficient image transformers & distillation through attention](https://arxiv.org/abs/2012.12877) by Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, Hervé Jégou.
 1. **[DETR](https://huggingface.co/docs/transformers/model_doc/detr)** (from Facebook) released with the paper [End-to-End Object Detection with Transformers](https://arxiv.org/abs/2005.12872) by Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, Sergey Zagoruyko.
 1. **[DialoGPT](https://huggingface.co/docs/transformers/model_doc/dialogpt)** (from Microsoft Research) released with the paper [DialoGPT: Large-Scale Generative Pre-training for Conversational Response Generation](https://arxiv.org/abs/1911.00536) by Yizhe Zhang, Siqi Sun, Michel Galley, Yen-Chun Chen, Chris Brockett, Xiang Gao, Jianfeng Gao, Jingjing Liu, Bill Dolan.
 1. **[DistilBERT](https://huggingface.co/docs/transformers/model_doc/distilbert)** (from HuggingFace), released together with the paper [DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter](https://arxiv.org/abs/1910.01108) by Victor Sanh, Lysandre Debut and Thomas Wolf. The same method has been applied to compress GPT2 into [DistilGPT2](https://github.com/huggingface/transformers/tree/main/examples/distillation), RoBERTa into [DistilRoBERTa](https://github.com/huggingface/transformers/tree/main/examples/distillation), Multilingual BERT into [DistilmBERT](https://github.com/huggingface/transformers/tree/main/examples/distillation) and a German version of DistilBERT.
 1. **[DiT](https://huggingface.co/docs/transformers/model_doc/dit)** (from Microsoft Research) released with the paper [DiT: Self-supervised Pre-training for Document Image Transformer](https://arxiv.org/abs/2203.02378) by Junlong Li, Yiheng Xu, Tengchao Lv, Lei Cui, Cha Zhang, Furu Wei.
+1. **[Donut](https://huggingface.co/docs/transformers/model_doc/donut)** (from NAVER) released with the paper [OCR-free Document Understanding Transformer](https://arxiv.org/abs/2111.15664) by Geewook Kim, Teakgyu Hong, Moonbin Yim, Jeongyeon Nam, Jinyoung Park, Jinyeong Yim, Wonseok Hwang, Sangdoo Yun, Dongyoon Han, Seunghyun Park.
 1. **[DPR](https://huggingface.co/docs/transformers/model_doc/dpr)** (from Facebook) released with the paper [Dense Passage Retrieval for Open-Domain Question Answering](https://arxiv.org/abs/2004.04906) by Vladimir Karpukhin, Barlas Oğuz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih.
 1. **[DPT](https://huggingface.co/docs/transformers/master/model_doc/dpt)** (from Intel Labs) released with the paper [Vision Transformers for Dense Prediction](https://arxiv.org/abs/2103.13413) by René Ranftl, Alexey Bochkovskiy, Vladlen Koltun.
 1. **[ELECTRA](https://huggingface.co/docs/transformers/model_doc/electra)** (from Google Research/Stanford University) released with the paper [ELECTRA: Pre-training text encoders as discriminators rather than generators](https://arxiv.org/abs/2003.10555) by Kevin Clark, Minh-Thang Luong, Quoc V. Le, Christopher D. Manning.
 1. **[EncoderDecoder](https://huggingface.co/docs/transformers/model_doc/encoder-decoder)** (from Google Research) released with the paper [Leveraging Pre-trained Checkpoints for Sequence Generation Tasks](https://arxiv.org/abs/1907.12461) by Sascha Rothe, Shashi Narayan, Aliaksei Severyn.
+1. **[ERNIE](https://huggingface.co/docs/transformers/model_doc/ernie)** (from Baidu) released with the paper [ERNIE: Enhanced Representation through Knowledge Integration](https://arxiv.org/abs/1904.09223) by Yu Sun, Shuohuan Wang, Yukun Li, Shikun Feng, Xuyi Chen, Han Zhang, Xin Tian, Danxiang Zhu, Hao Tian, Hua Wu.
+1. **[ESM](https://huggingface.co/docs/transformers/model_doc/esm)** (from Meta AI) are transformer protein language models.  **ESM-1b** was released with the paper [Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences](https://www.pnas.org/content/118/15/e2016239118) by Alexander Rives, Joshua Meier, Tom Sercu, Siddharth Goyal, Zeming Lin, Jason Liu, Demi Guo, Myle Ott, C. Lawrence Zitnick, Jerry Ma, and Rob Fergus. **ESM-1v** was released with the paper [Language models enable zero-shot prediction of the effects of mutations on protein function](https://doi.org/10.1101/2021.07.09.450648) by Joshua Meier, Roshan Rao, Robert Verkuil, Jason Liu, Tom Sercu and Alexander Rives. **ESM-2** was released with the paper [Language models of protein sequences at the scale of evolution enable accurate structure prediction](https://doi.org/10.1101/2022.07.20.500902) by Zeming Lin, Halil Akin, Roshan Rao, Brian Hie, Zhongkai Zhu, Wenting Lu, Allan dos Santos Costa, Maryam Fazel-Zarandi, Tom Sercu, Sal Candido, Alexander Rives.
 1. **[FlauBERT](https://huggingface.co/docs/transformers/model_doc/flaubert)** (from CNRS) released with the paper [FlauBERT: Unsupervised Language Model Pre-training for French](https://arxiv.org/abs/1912.05372) by Hang Le, Loïc Vial, Jibril Frej, Vincent Segonne, Maximin Coavoux, Benjamin Lecouteux, Alexandre Allauzen, Benoît Crabbé, Laurent Besacier, Didier Schwab.
 1. **[FLAVA](https://huggingface.co/docs/transformers/model_doc/flava)** (from Facebook AI) released with the paper [FLAVA: A Foundational Language And Vision Alignment Model](https://arxiv.org/abs/2112.04482) by Amanpreet Singh, Ronghang Hu, Vedanuj Goswami, Guillaume Couairon, Wojciech Galuba, Marcus Rohrbach, and Douwe Kiela.
 1. **[FNet](https://huggingface.co/docs/transformers/model_doc/fnet)** (from Google Research) released with the paper [FNet: Mixing Tokens with Fourier Transforms](https://arxiv.org/abs/2105.03824) by James Lee-Thorp, Joshua Ainslie, Ilya Eckstein, Santiago Ontanon.
@@ -253,7 +258,8 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[GLPN](https://huggingface.co/docs/transformers/model_doc/glpn)** (from KAIST) released with the paper [Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth](https://arxiv.org/abs/2201.07436) by Doyeon Kim, Woonghyun Ga, Pyungwhan Ahn, Donggyu Joo, Sehwan Chun, Junmo Kim.
 1. **[GPT](https://huggingface.co/docs/transformers/model_doc/openai-gpt)** (from OpenAI) released with the paper [Improving Language Understanding by Generative Pre-Training](https://blog.openai.com/language-unsupervised/) by Alec Radford, Karthik Narasimhan, Tim Salimans and Ilya Sutskever.
 1. **[GPT Neo](https://huggingface.co/docs/transformers/model_doc/gpt_neo)** (from EleutherAI) released in the repository [EleutherAI/gpt-neo](https://github.com/EleutherAI/gpt-neo) by Sid Black, Stella Biderman, Leo Gao, Phil Wang and Connor Leahy.
-1. **[GPT NeoX](https://huggingface.co/docs/transformers/model_doc/gpt_neox)** (from EleutherAI) released with the paper [GPT-NeoX-20B: An Open-Source Autoregressive Language Model](https://arxiv.org/abs/2204.06745) by Sid Black, Stella Biderman, Eric Hallahan, Quentin Anthony, Leo Gao, Laurence Golding, Horace He, Connor Leahy, Kyle McDonell, Jason Phang, Michael Pieler, USVSN Sai Prashanth, Shivanshu Purohit, Laria Reynolds, Jonathan Tow, Ben Wang, Samuel Weinbach 
+1. **[GPT NeoX](https://huggingface.co/docs/transformers/model_doc/gpt_neox)** (from EleutherAI) released with the paper [GPT-NeoX-20B: An Open-Source Autoregressive Language Model](https://arxiv.org/abs/2204.06745) by Sid Black, Stella Biderman, Eric Hallahan, Quentin Anthony, Leo Gao, Laurence Golding, Horace He, Connor Leahy, Kyle McDonell, Jason Phang, Michael Pieler, USVSN Sai Prashanth, Shivanshu Purohit, Laria Reynolds, Jonathan Tow, Ben Wang, Samuel Weinbach
+1. **[GPT NeoX Japanese](https://huggingface.co/docs/transformers/model_doc/gpt_neox_japanese)** (from ABEJA) released by Shinya Otani, Takayoshi Makabe, Anuj Arora, and Kyo Hattori.
 1. **[GPT-2](https://huggingface.co/docs/transformers/model_doc/gpt2)** (from OpenAI) released with the paper [Language Models are Unsupervised Multitask Learners](https://blog.openai.com/better-language-models/) by Alec Radford*, Jeffrey Wu*, Rewon Child, David Luan, Dario Amodei** and Ilya Sutskever**.
 1. **[GPT-J](https://huggingface.co/docs/transformers/model_doc/gptj)** (from EleutherAI) released in the repository [kingoflolz/mesh-transformer-jax](https://github.com/kingoflolz/mesh-transformer-jax/) by Ben Wang and Aran Komatsuzaki.
 1. **[GroupViT](https://huggingface.co/docs/transformers/model_doc/groupvit)** (from UCSD, NVIDIA) released with the paper [GroupViT: Semantic Segmentation Emerges from Text Supervision](https://arxiv.org/abs/2202.11094) by Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, Xiaolong Wang.
@@ -263,7 +269,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[LayoutLM](https://huggingface.co/docs/transformers/model_doc/layoutlm)** (from Microsoft Research Asia) released with the paper [LayoutLM: Pre-training of Text and Layout for Document Image Understanding](https://arxiv.org/abs/1912.13318) by Yiheng Xu, Minghao Li, Lei Cui, Shaohan Huang, Furu Wei, Ming Zhou.
 1. **[LayoutLMv2](https://huggingface.co/docs/transformers/model_doc/layoutlmv2)** (from Microsoft Research Asia) released with the paper [LayoutLMv2: Multi-modal Pre-training for Visually-Rich Document Understanding](https://arxiv.org/abs/2012.14740) by Yang Xu, Yiheng Xu, Tengchao Lv, Lei Cui, Furu Wei, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Wanxiang Che, Min Zhang, Lidong Zhou.
 1. **[LayoutLMv3](https://huggingface.co/docs/transformers/model_doc/layoutlmv3)** (from Microsoft Research Asia) released with the paper [LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking](https://arxiv.org/abs/2204.08387) by Yupan Huang, Tengchao Lv, Lei Cui, Yutong Lu, Furu Wei.
-1. **[LayoutXLM](https://huggingface.co/docs/transformers/model_doc/layoutlmv2)** (from Microsoft Research Asia) released with the paper [LayoutXLM: Multimodal Pre-training for Multilingual Visually-rich Document Understanding](https://arxiv.org/abs/2104.08836) by Yiheng Xu, Tengchao Lv, Lei Cui, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Furu Wei.
+1. **[LayoutXLM](https://huggingface.co/docs/transformers/model_doc/layoutxlm)** (from Microsoft Research Asia) released with the paper [LayoutXLM: Multimodal Pre-training for Multilingual Visually-rich Document Understanding](https://arxiv.org/abs/2104.08836) by Yiheng Xu, Tengchao Lv, Lei Cui, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Furu Wei.
 1. **[LED](https://huggingface.co/docs/transformers/model_doc/led)** (from AllenAI) released with the paper [Longformer: The Long-Document Transformer](https://arxiv.org/abs/2004.05150) by Iz Beltagy, Matthew E. Peters, Arman Cohan.
 1. **[LeViT](https://huggingface.co/docs/transformers/model_doc/levit)** (from Meta AI) released with the paper [LeViT: A Vision Transformer in ConvNet's Clothing for Faster Inference](https://arxiv.org/abs/2104.01136) by Ben Graham, Alaaeldin El-Nouby, Hugo Touvron, Pierre Stock, Armand Joulin, Hervé Jégou, Matthijs Douze.
 1. **[Longformer](https://huggingface.co/docs/transformers/model_doc/longformer)** (from AllenAI) released with the paper [Longformer: The Long-Document Transformer](https://arxiv.org/abs/2004.05150) by Iz Beltagy, Matthew E. Peters, Arman Cohan.
@@ -273,6 +279,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[M-CTC-T](https://huggingface.co/docs/transformers/model_doc/mctct)** (from Facebook) released with the paper [Pseudo-Labeling For Massively Multilingual Speech Recognition](https://arxiv.org/abs/2111.00161) by Loren Lugosch, Tatiana Likhomanenko, Gabriel Synnaeve, and Ronan Collobert.
 1. **[M2M100](https://huggingface.co/docs/transformers/model_doc/m2m_100)** (from Facebook) released with the paper [Beyond English-Centric Multilingual Machine Translation](https://arxiv.org/abs/2010.11125) by Angela Fan, Shruti Bhosale, Holger Schwenk, Zhiyi Ma, Ahmed El-Kishky, Siddharth Goyal, Mandeep Baines, Onur Celebi, Guillaume Wenzek, Vishrav Chaudhary, Naman Goyal, Tom Birch, Vitaliy Liptchinsky, Sergey Edunov, Edouard Grave, Michael Auli, Armand Joulin.
 1. **[MarianMT](https://huggingface.co/docs/transformers/model_doc/marian)** Machine translation models trained using [OPUS](http://opus.nlpl.eu/) data by Jörg Tiedemann. The [Marian Framework](https://marian-nmt.github.io/) is being developed by the Microsoft Translator Team.
+1. **[MarkupLM](https://huggingface.co/docs/transformers/model_doc/markuplm)** (from Microsoft Research Asia) released with the paper [MarkupLM: Pre-training of Text and Markup Language for Visually-rich Document Understanding](https://arxiv.org/abs/2110.08518) by Junlong Li, Yiheng Xu, Lei Cui, Furu Wei.
 1. **[MaskFormer](https://huggingface.co/docs/transformers/model_doc/maskformer)** (from Meta and UIUC) released with the paper [Per-Pixel Classification is Not All You Need for Semantic Segmentation](https://arxiv.org/abs/2107.06278) by Bowen Cheng, Alexander G. Schwing, Alexander Kirillov.
 1. **[mBART](https://huggingface.co/docs/transformers/model_doc/mbart)** (from Facebook) released with the paper [Multilingual Denoising Pre-training for Neural Machine Translation](https://arxiv.org/abs/2001.08210) by Yinhan Liu, Jiatao Gu, Naman Goyal, Xian Li, Sergey Edunov, Marjan Ghazvininejad, Mike Lewis, Luke Zettlemoyer.
 1. **[mBART-50](https://huggingface.co/docs/transformers/model_doc/mbart)** (from Facebook) released with the paper [Multilingual Translation with Extensible Multilingual Pretraining and Finetuning](https://arxiv.org/abs/2008.00401) by Yuqing Tang, Chau Tran, Xian Li, Peng-Jen Chen, Naman Goyal, Vishrav Chaudhary, Jiatao Gu, Angela Fan.
@@ -290,6 +297,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[OPT](https://huggingface.co/docs/transformers/master/model_doc/opt)** (from Meta AI) released with the paper [OPT: Open Pre-trained Transformer Language Models](https://arxiv.org/abs/2205.01068) by Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen et al.
 1. **[OWL-ViT](https://huggingface.co/docs/transformers/model_doc/owlvit)** (from Google AI) released with the paper [Simple Open-Vocabulary Object Detection with Vision Transformers](https://arxiv.org/abs/2205.06230) by Matthias Minderer, Alexey Gritsenko, Austin Stone, Maxim Neumann, Dirk Weissenborn, Alexey Dosovitskiy, Aravindh Mahendran, Anurag Arnab, Mostafa Dehghani, Zhuoran Shen, Xiao Wang, Xiaohua Zhai, Thomas Kipf, and Neil Houlsby.
 1. **[Pegasus](https://huggingface.co/docs/transformers/model_doc/pegasus)** (from Google) released with the paper [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu.
+1. **[PEGASUS-X](https://huggingface.co/docs/transformers/model_doc/pegasus_x)** (from Google) released with the paper [Investigating Efficiently Extending Transformers for Long Input Summarization](https://arxiv.org/abs/2208.04347) by Jason Phang, Yao Zhao, Peter J. Liu.
 1. **[Perceiver IO](https://huggingface.co/docs/transformers/model_doc/perceiver)** (from Deepmind) released with the paper [Perceiver IO: A General Architecture for Structured Inputs & Outputs](https://arxiv.org/abs/2107.14795) by Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier Hénaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, João Carreira.
 1. **[PhoBERT](https://huggingface.co/docs/transformers/model_doc/phobert)** (from VinAI Research) released with the paper [PhoBERT: Pre-trained language models for Vietnamese](https://www.aclweb.org/anthology/2020.findings-emnlp.92/) by Dat Quoc Nguyen and Anh Tuan Nguyen.
 1. **[PLBart](https://huggingface.co/docs/transformers/model_doc/plbart)** (from UCLA NLP) released with the paper [Unified Pre-training for Program Understanding and Generation](https://arxiv.org/abs/2103.06333) by Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, Kai-Wei Chang.
@@ -299,9 +307,9 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[RAG](https://huggingface.co/docs/transformers/model_doc/rag)** (from Facebook) released with the paper [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://arxiv.org/abs/2005.11401) by Patrick Lewis, Ethan Perez, Aleksandara Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, Douwe Kiela.
 1. **[REALM](https://huggingface.co/docs/transformers/model_doc/realm.html)** (from Google Research) released with the paper [REALM: Retrieval-Augmented Language Model Pre-Training](https://arxiv.org/abs/2002.08909) by Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat and Ming-Wei Chang.
 1. **[Reformer](https://huggingface.co/docs/transformers/model_doc/reformer)** (from Google Research) released with the paper [Reformer: The Efficient Transformer](https://arxiv.org/abs/2001.04451) by Nikita Kitaev, Łukasz Kaiser, Anselm Levskaya.
-1. **[RegNet](https://huggingface.co/docs/transformers/model_doc/regnet)** (from META Research) released with the paper [Designing Network Design Space](https://arxiv.org/abs/2003.13678) by Ilija Radosavovic, Raj Prateek Kosaraju, Ross Girshick, Kaiming He, Piotr Dollár. 
+1. **[RegNet](https://huggingface.co/docs/transformers/model_doc/regnet)** (from META Research) released with the paper [Designing Network Design Space](https://arxiv.org/abs/2003.13678) by Ilija Radosavovic, Raj Prateek Kosaraju, Ross Girshick, Kaiming He, Piotr Dollár.
 1. **[RemBERT](https://huggingface.co/docs/transformers/model_doc/rembert)** (from Google Research) released with the paper [Rethinking embedding coupling in pre-trained language models](https://arxiv.org/pdf/2010.12821.pdf) by Hyung Won Chung, Thibault Févry, Henry Tsai, M. Johnson, Sebastian Ruder.
-1. **[ResNet](https://huggingface.co/docs/transformers/model_doc/resnet)** (from Microsoft Research) released with the paper [Deep Residual Learning for Image Recognition](https://arxiv.org/abs/1512.03385) by Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun. 
+1. **[ResNet](https://huggingface.co/docs/transformers/model_doc/resnet)** (from Microsoft Research) released with the paper [Deep Residual Learning for Image Recognition](https://arxiv.org/abs/1512.03385) by Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun.
 1. **[RoBERTa](https://huggingface.co/docs/transformers/model_doc/roberta)** (from Facebook), released together with the paper a [Robustly Optimized BERT Pretraining Approach](https://arxiv.org/abs/1907.11692) by Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, Veselin Stoyanov.
 1. **[RoFormer](https://huggingface.co/docs/transformers/model_doc/roformer)** (from ZhuiyiTechnology), released together with the paper a [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://arxiv.org/pdf/2104.09864v1.pdf) by Jianlin Su and Yu Lu and Shengfeng Pan and Bo Wen and Yunfeng Liu.
 1. **[SegFormer](https://huggingface.co/docs/transformers/model_doc/segformer)** (from NVIDIA) released with the paper [SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers](https://arxiv.org/abs/2105.15203) by Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M. Alvarez, Ping Luo.
@@ -312,27 +320,32 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[Splinter](https://huggingface.co/docs/transformers/model_doc/splinter)** (from Tel Aviv University), released together with the paper [Few-Shot Question Answering by Pretraining Span Selection](https://arxiv.org/abs/2101.00438) by Ori Ram, Yuval Kirstain, Jonathan Berant, Amir Globerson, Omer Levy.
 1. **[SqueezeBERT](https://huggingface.co/docs/transformers/model_doc/squeezebert)** (from Berkeley) released with the paper [SqueezeBERT: What can computer vision teach NLP about efficient neural networks?](https://arxiv.org/abs/2006.11316) by Forrest N. Iandola, Albert E. Shaw, Ravi Krishna, and Kurt W. Keutzer.
 1. **[Swin Transformer](https://huggingface.co/docs/transformers/model_doc/swin)** (from Microsoft) released with the paper [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://arxiv.org/abs/2103.14030) by Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, Baining Guo.
-1. **[Swin Transformer V2](https://huggingface.co/docs/transformers/main/model_doc/swinv2)** (from Microsoft) released with the paper [Swin Transformer V2: Scaling Up Capacity and Resolution](https://arxiv.org/abs/2111.09883) by Ze Liu, Han Hu, Yutong Lin, Zhuliang Yao, Zhenda Xie, Yixuan Wei, Jia Ning, Yue Cao, Zheng Zhang, Li Dong, Furu Wei, Baining Guo.
+1. **[Swin Transformer V2](https://huggingface.co/docs/transformers/model_doc/swinv2)** (from Microsoft) released with the paper [Swin Transformer V2: Scaling Up Capacity and Resolution](https://arxiv.org/abs/2111.09883) by Ze Liu, Han Hu, Yutong Lin, Zhuliang Yao, Zhenda Xie, Yixuan Wei, Jia Ning, Yue Cao, Zheng Zhang, Li Dong, Furu Wei, Baining Guo.
 1. **[T5](https://huggingface.co/docs/transformers/model_doc/t5)** (from Google AI) released with the paper [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://arxiv.org/abs/1910.10683) by Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu.
 1. **[T5v1.1](https://huggingface.co/docs/transformers/model_doc/t5v1.1)** (from Google AI) released in the repository [google-research/text-to-text-transfer-transformer](https://github.com/google-research/text-to-text-transfer-transformer/blob/main/released_checkpoints.md#t511) by Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu.
 1. **[TAPAS](https://huggingface.co/docs/transformers/model_doc/tapas)** (from Google AI) released with the paper [TAPAS: Weakly Supervised Table Parsing via Pre-training](https://arxiv.org/abs/2004.02349) by Jonathan Herzig, Paweł Krzysztof Nowak, Thomas Müller, Francesco Piccinno and Julian Martin Eisenschlos.
 1. **[TAPEX](https://huggingface.co/docs/transformers/model_doc/tapex)** (from Microsoft Research) released with the paper [TAPEX: Table Pre-training via Learning a Neural SQL Executor](https://arxiv.org/abs/2107.07653) by Qian Liu, Bei Chen, Jiaqi Guo, Morteza Ziyadi, Zeqi Lin, Weizhu Chen, Jian-Guang Lou.
-1. **[Trajectory Transformer](https://huggingface.co/docs/transformers/model_doc/trajectory_transformers)** (from the University of California at Berkeley) released with the paper [Offline Reinforcement Learning as One Big Sequence Modeling Problem](https://arxiv.org/abs/2106.02039) by Michael Janner, Qiyang Li, Sergey Levine 
+1. **[Time Series Transformer](https://huggingface.co/docs/transformers/model_doc/time_series_transformer)** (from HuggingFace).
+1. **[Trajectory Transformer](https://huggingface.co/docs/transformers/model_doc/trajectory_transformers)** (from the University of California at Berkeley) released with the paper [Offline Reinforcement Learning as One Big Sequence Modeling Problem](https://arxiv.org/abs/2106.02039) by Michael Janner, Qiyang Li, Sergey Levine
 1. **[Transformer-XL](https://huggingface.co/docs/transformers/model_doc/transfo-xl)** (from Google/CMU) released with the paper [Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context](https://arxiv.org/abs/1901.02860) by Zihang Dai*, Zhilin Yang*, Yiming Yang, Jaime Carbonell, Quoc V. Le, Ruslan Salakhutdinov.
 1. **[TrOCR](https://huggingface.co/docs/transformers/model_doc/trocr)** (from Microsoft), released together with the paper [TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models](https://arxiv.org/abs/2109.10282) by Minghao Li, Tengchao Lv, Lei Cui, Yijuan Lu, Dinei Florencio, Cha Zhang, Zhoujun Li, Furu Wei.
-1. **[UL2](https://huggingface.co/docs/transformers/model_doc/ul2)** (from Google Research) released with the paper [Unifying Language Learning Paradigms](https://arxiv.org/abs/2205.05131v1) by Yi Tay, Mostafa Dehghani, Vinh Q. Tran, Xavier Garcia, Dara Bahri, Tal Schuster, Huaixiu Steven Zheng, Neil Houlsby, Donald Metzler 
+1. **[UL2](https://huggingface.co/docs/transformers/model_doc/ul2)** (from Google Research) released with the paper [Unifying Language Learning Paradigms](https://arxiv.org/abs/2205.05131v1) by Yi Tay, Mostafa Dehghani, Vinh Q. Tran, Xavier Garcia, Dara Bahri, Tal Schuster, Huaixiu Steven Zheng, Neil Houlsby, Donald Metzler
 1. **[UniSpeech](https://huggingface.co/docs/transformers/model_doc/unispeech)** (from Microsoft Research) released with the paper [UniSpeech: Unified Speech Representation Learning with Labeled and Unlabeled Data](https://arxiv.org/abs/2101.07597) by Chengyi Wang, Yu Wu, Yao Qian, Kenichi Kumatani, Shujie Liu, Furu Wei, Michael Zeng, Xuedong Huang.
 1. **[UniSpeechSat](https://huggingface.co/docs/transformers/model_doc/unispeech-sat)** (from Microsoft Research) released with the paper [UNISPEECH-SAT: UNIVERSAL SPEECH REPRESENTATION LEARNING WITH SPEAKER AWARE PRE-TRAINING](https://arxiv.org/abs/2110.05752) by Sanyuan Chen, Yu Wu, Chengyi Wang, Zhengyang Chen, Zhuo Chen, Shujie Liu, Jian Wu, Yao Qian, Furu Wei, Jinyu Li, Xiangzhan Yu.
 1. **[VAN](https://huggingface.co/docs/transformers/model_doc/van)** (from Tsinghua University and Nankai University) released with the paper [Visual Attention Network](https://arxiv.org/pdf/2202.09741.pdf) by Meng-Hao Guo, Cheng-Ze Lu, Zheng-Ning Liu, Ming-Ming Cheng, Shi-Min Hu.
+1. **[VideoMAE](https://huggingface.co/docs/transformers/model_doc/videomae)** (from Multimedia Computing Group, Nanjing University) released with the paper [VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training](https://arxiv.org/abs/2203.12602) by Zhan Tong, Yibing Song, Jue Wang, Limin Wang.
 1. **[ViLT](https://huggingface.co/docs/transformers/model_doc/vilt)** (from NAVER AI Lab/Kakao Enterprise/Kakao Brain) released with the paper [ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision](https://arxiv.org/abs/2102.03334) by Wonjae Kim, Bokyung Son, Ildoo Kim.
 1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.
 1. **[VisualBERT](https://huggingface.co/docs/transformers/model_doc/visual_bert)** (from UCLA NLP) released with the paper [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) by Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang.
 1. **[ViTMAE](https://huggingface.co/docs/transformers/model_doc/vit_mae)** (from Meta AI) released with the paper [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) by Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick.
+1. **[ViTMSN](https://huggingface.co/docs/transformers/model_doc/vit_msn)** (from Meta AI) released with the paper [Masked Siamese Networks for Label-Efficient Learning](https://arxiv.org/abs/2204.07141) by Mahmoud Assran, Mathilde Caron, Ishan Misra, Piotr Bojanowski, Florian Bordes, Pascal Vincent, Armand Joulin, Michael Rabbat, Nicolas Ballas.
 1. **[Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/wav2vec2)** (from Facebook AI) released with the paper [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) by Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli.
 1. **[Wav2Vec2-Conformer](https://huggingface.co/docs/transformers/model_doc/wav2vec2-conformer)** (from Facebook AI) released with the paper [FAIRSEQ S2T: Fast Speech-to-Text Modeling with FAIRSEQ](https://arxiv.org/abs/2010.05171) by Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Sravya Popuri, Dmytro Okhonko, Juan Pino.
 1. **[Wav2Vec2Phoneme](https://huggingface.co/docs/transformers/model_doc/wav2vec2_phoneme)** (from Facebook AI) released with the paper [Simple and Effective Zero-shot Cross-lingual Phoneme Recognition](https://arxiv.org/abs/2109.11680) by Qiantong Xu, Alexei Baevski, Michael Auli.
 1. **[WavLM](https://huggingface.co/docs/transformers/model_doc/wavlm)** (from Microsoft Research) released with the paper [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://arxiv.org/abs/2110.13900) by Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng, Furu Wei.
-1. **[XGLM](https://huggingface.co/docs/transformers/model_doc/xglm)** (From Facebook AI) released with the paper [Few-shot Learning with Multilingual Language Models](https://arxiv.org/abs/2112.10668) by Xi Victoria Lin, Todor Mihaylov, Mikel Artetxe, Tianlu Wang, Shuohui Chen, Daniel Simig, Myle Ott, Naman Goyal, Shruti Bhosale, Jingfei Du, Ramakanth Pasunuru, Sam Shleifer, Punit Singh Koura, Vishrav Chaudhary, Brian O'Horo, Jeff Wang, Luke Zettlemoyer, Zornitsa Kozareva, Mona Diab, Veselin Stoyanov, Xian Li. 
+1. **[Whisper](https://huggingface.co/docs/transformers/model_doc/whisper)** (from OpenAI) released with the paper [Robust Speech Recognition via Large-Scale Weak Supervision](https://cdn.openai.com/papers/whisper.pdf) by Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, Ilya Sutskever.
+1. **[X-CLIP](https://huggingface.co/docs/transformers/model_doc/xclip)** (from Microsoft Research) released with the paper [Expanding Language-Image Pretrained Models for General Video Recognition](https://arxiv.org/abs/2208.02816) by Bolin Ni, Houwen Peng, Minghao Chen, Songyang Zhang, Gaofeng Meng, Jianlong Fu, Shiming Xiang, Haibin Ling.
+1. **[XGLM](https://huggingface.co/docs/transformers/model_doc/xglm)** (From Facebook AI) released with the paper [Few-shot Learning with Multilingual Language Models](https://arxiv.org/abs/2112.10668) by Xi Victoria Lin, Todor Mihaylov, Mikel Artetxe, Tianlu Wang, Shuohui Chen, Daniel Simig, Myle Ott, Naman Goyal, Shruti Bhosale, Jingfei Du, Ramakanth Pasunuru, Sam Shleifer, Punit Singh Koura, Vishrav Chaudhary, Brian O'Horo, Jeff Wang, Luke Zettlemoyer, Zornitsa Kozareva, Mona Diab, Veselin Stoyanov, Xian Li.
 1. **[XLM](https://huggingface.co/docs/transformers/model_doc/xlm)** (from Facebook) released together with the paper [Cross-lingual Language Model Pretraining](https://arxiv.org/abs/1901.07291) by Guillaume Lample and Alexis Conneau.
 1. **[XLM-ProphetNet](https://huggingface.co/docs/transformers/model_doc/xlm-prophetnet)** (from Microsoft Research) released with the paper [ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training](https://arxiv.org/abs/2001.04063) by Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou.
 1. **[XLM-RoBERTa](https://huggingface.co/docs/transformers/model_doc/xlm-roberta)** (from Facebook AI), released together with the paper [Unsupervised Cross-lingual Representation Learning at Scale](https://arxiv.org/abs/1911.02116) by Alexis Conneau*, Kartikay Khandelwal*, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer and Veselin Stoyanov.
@@ -342,7 +355,7 @@ Flax, PyTorch, TensorFlow 설치 페이지에서 이들을 conda로 설치하는
 1. **[XLSR-Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/xlsr_wav2vec2)** (from Facebook AI) released with the paper [Unsupervised Cross-Lingual Representation Learning For Speech Recognition](https://arxiv.org/abs/2006.13979) by Alexis Conneau, Alexei Baevski, Ronan Collobert, Abdelrahman Mohamed, Michael Auli.
 1. **[YOLOS](https://huggingface.co/docs/transformers/model_doc/yolos)** (from Huazhong University of Science & Technology) released with the paper [You Only Look at One Sequence: Rethinking Transformer in Vision through Object Detection](https://arxiv.org/abs/2106.00666) by Yuxin Fang, Bencheng Liao, Xinggang Wang, Jiemin Fang, Jiyang Qi, Rui Wu, Jianwei Niu, Wenyu Liu.
 1. **[YOSO](https://huggingface.co/docs/transformers/model_doc/yoso)** (from the University of Wisconsin - Madison) released with the paper [You Only Sample (Almost) by Zhanpeng Zeng, Yunyang Xiong, Sathya N. Ravi, Shailesh Acharya, Glenn Fung, Vikas Singh.
-1. 새로운 모델을 올리고 싶나요? 우리가 **상세한 가이드와 템플릿** 으로 새로운 모델을 올리도록 도와드릴게요. 가이드와 템플릿은 이 저장소의 [`templates`](./templates) 폴더에서 확인하실 수 있습니다. [컨트리뷰션 가이드라인](./CONTRIBUTING.md)을 꼭 확인해주시고, PR을 올리기 전에 메인테이너에게 연락하거나 이슈를 오픈해 피드백을 받으시길 바랍니다. 
+1. 새로운 모델을 올리고 싶나요? 우리가 **상세한 가이드와 템플릿** 으로 새로운 모델을 올리도록 도와드릴게요. 가이드와 템플릿은 이 저장소의 [`templates`](./templates) 폴더에서 확인하실 수 있습니다. [컨트리뷰션 가이드라인](./CONTRIBUTING.md)을 꼭 확인해주시고, PR을 올리기 전에 메인테이너에게 연락하거나 이슈를 오픈해 피드백을 받으시길 바랍니다.
 
 각 모델이 Flax, PyTorch, TensorFlow으로 구현되었는지 또는 🤗 Tokenizers 라이브러리가 지원하는 토크나이저를 사용하는지 확인하려면, [이 표](https://huggingface.co/docs/transformers/index#supported-frameworks)를 확인하세요.
 

README_zh-hans.md

@@ -173,7 +173,7 @@ checkpoint: 检查点
     - 对所有模型统一的API
 
 1. 更低计算开销，更少的碳排放：
-    - 研究人员可以分享亿训练的模型而非次次从头开始训练
+    - 研究人员可以分享已训练的模型而非每次从头开始训练
     - 工程师可以减少计算用时和生产环境开销
     - 数十种模型架构、两千多个预训练模型、100多种语言支持
 
@@ -245,13 +245,14 @@ conda install -c huggingface transformers
 1. **[BigBird-RoBERTa](https://huggingface.co/docs/transformers/model_doc/big_bird)** (来自 Google Research) 伴随论文 [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) 由 Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed 发布。
 1. **[Blenderbot](https://huggingface.co/docs/transformers/model_doc/blenderbot)** (来自 Facebook) 伴随论文 [Recipes for building an open-domain chatbot](https://arxiv.org/abs/2004.13637) 由 Stephen Roller, Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Kurt Shuster, Eric M. Smith, Y-Lan Boureau, Jason Weston 发布。
 1. **[BlenderbotSmall](https://huggingface.co/docs/transformers/model_doc/blenderbot-small)** (来自 Facebook) 伴随论文 [Recipes for building an open-domain chatbot](https://arxiv.org/abs/2004.13637) 由 Stephen Roller, Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Kurt Shuster, Eric M. Smith, Y-Lan Boureau, Jason Weston 发布。
-1. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigSicence Workshop](https://bigscience.huggingface.co/). 
+1. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigSicence Workshop](https://bigscience.huggingface.co/).
 1. **[BORT](https://huggingface.co/docs/transformers/model_doc/bort)** (来自 Alexa) 伴随论文 [Optimal Subarchitecture Extraction For BERT](https://arxiv.org/abs/2010.10499) 由 Adrian de Wynter and Daniel J. Perry 发布。
 1. **[ByT5](https://huggingface.co/docs/transformers/model_doc/byt5)** (来自 Google Research) 伴随论文 [ByT5: Towards a token-free future with pre-trained byte-to-byte models](https://arxiv.org/abs/2105.13626) 由 Linting Xue, Aditya Barua, Noah Constant, Rami Al-Rfou, Sharan Narang, Mihir Kale, Adam Roberts, Colin Raffel 发布。
 1. **[CamemBERT](https://huggingface.co/docs/transformers/model_doc/camembert)** (来自 Inria/Facebook/Sorbonne) 伴随论文 [CamemBERT: a Tasty French Language Model](https://arxiv.org/abs/1911.03894) 由 Louis Martin*, Benjamin Muller*, Pedro Javier Ortiz Suárez*, Yoann Dupont, Laurent Romary, Éric Villemonte de la Clergerie, Djamé Seddah and Benoît Sagot 发布。
 1. **[CANINE](https://huggingface.co/docs/transformers/model_doc/canine)** (来自 Google Research) 伴随论文 [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) 由 Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting 发布。
 1. **[CLIP](https://huggingface.co/docs/transformers/model_doc/clip)** (来自 OpenAI) 伴随论文 [Learning Transferable Visual Models From Natural Language Supervision](https://arxiv.org/abs/2103.00020) 由 Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, Ilya Sutskever 发布。
 1. **[CodeGen](https://huggingface.co/docs/transformers/model_doc/codegen)** (来自 Salesforce) 伴随论文 [A Conversational Paradigm for Program Synthesis](https://arxiv.org/abs/2203.13474) 由 Erik Nijkamp, Bo Pang, Hiroaki Hayashi, Lifu Tu, Huan Wang, Yingbo Zhou, Silvio Savarese, Caiming Xiong 发布。
+1. **[Conditional DETR](https://huggingface.co/docs/transformers/model_doc/conditional_detr)** (来自 Microsoft Research Asia) 伴随论文 [Conditional DETR for Fast Training Convergence](https://arxiv.org/abs/2108.06152) 由 Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, Jingdong Wang 发布。
 1. **[ConvBERT](https://huggingface.co/docs/transformers/model_doc/convbert)** (来自 YituTech) 伴随论文 [ConvBERT: Improving BERT with Span-based Dynamic Convolution](https://arxiv.org/abs/2008.02496) 由 Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan 发布。
 1. **[ConvNeXT](https://huggingface.co/docs/transformers/model_doc/convnext)** (来自 Facebook AI) 伴随论文 [A ConvNet for the 2020s](https://arxiv.org/abs/2201.03545) 由 Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie 发布。
 1. **[CPM](https://huggingface.co/docs/transformers/model_doc/cpm)** (来自 Tsinghua University) 伴随论文 [CPM: A Large-scale Generative Chinese Pre-trained Language Model](https://arxiv.org/abs/2012.00413) 由 Zhengyan Zhang, Xu Han, Hao Zhou, Pei Ke, Yuxian Gu, Deming Ye, Yujia Qin, Yusheng Su, Haozhe Ji, Jian Guan, Fanchao Qi, Xiaozhi Wang, Yanan Zheng, Guoyang Zeng, Huanqi Cao, Shengqi Chen, Daixuan Li, Zhenbo Sun, Zhiyuan Liu, Minlie Huang, Wentao Han, Jie Tang, Juanzi Li, Xiaoyan Zhu, Maosong Sun 发布。
@@ -261,15 +262,19 @@ conda install -c huggingface transformers
 1. **[DeBERTa](https://huggingface.co/docs/transformers/model_doc/deberta)** (来自 Microsoft) 伴随论文 [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://arxiv.org/abs/2006.03654) 由 Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen 发布。
 1. **[DeBERTa-v2](https://huggingface.co/docs/transformers/model_doc/deberta-v2)** (来自 Microsoft) 伴随论文 [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://arxiv.org/abs/2006.03654) 由 Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen 发布。
 1. **[Decision Transformer](https://huggingface.co/docs/transformers/model_doc/decision_transformer)** (来自 Berkeley/Facebook/Google) 伴随论文 [Decision Transformer: Reinforcement Learning via Sequence Modeling](https://arxiv.org/abs/2106.01345) 由 Lili Chen, Kevin Lu, Aravind Rajeswaran, Kimin Lee, Aditya Grover, Michael Laskin, Pieter Abbeel, Aravind Srinivas, Igor Mordatch 发布。
+1. **[Deformable DETR](https://huggingface.co/docs/transformers/model_doc/deformable_detr)** (来自 SenseTime Research) 伴随论文 [Deformable DETR: Deformable Transformers for End-to-End Object Detection](https://arxiv.org/abs/2010.04159) 由 Xizhou Zhu, Weijie Su, Lewei Lu, Bin Li, Xiaogang Wang, Jifeng Dai 发布。
 1. **[DeiT](https://huggingface.co/docs/transformers/model_doc/deit)** (来自 Facebook) 伴随论文 [Training data-efficient image transformers & distillation through attention](https://arxiv.org/abs/2012.12877) 由 Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, Hervé Jégou 发布。
 1. **[DETR](https://huggingface.co/docs/transformers/model_doc/detr)** (来自 Facebook) 伴随论文 [End-to-End Object Detection with Transformers](https://arxiv.org/abs/2005.12872) 由 Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, Sergey Zagoruyko 发布。
 1. **[DialoGPT](https://huggingface.co/docs/transformers/model_doc/dialogpt)** (来自 Microsoft Research) 伴随论文 [DialoGPT: Large-Scale Generative Pre-training for Conversational Response Generation](https://arxiv.org/abs/1911.00536) 由 Yizhe Zhang, Siqi Sun, Michel Galley, Yen-Chun Chen, Chris Brockett, Xiang Gao, Jianfeng Gao, Jingjing Liu, Bill Dolan 发布。
 1. **[DistilBERT](https://huggingface.co/docs/transformers/model_doc/distilbert)** (来自 HuggingFace), 伴随论文 [DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter](https://arxiv.org/abs/1910.01108) 由 Victor Sanh, Lysandre Debut and Thomas Wolf 发布。 同样的方法也应用于压缩 GPT-2 到 [DistilGPT2](https://github.com/huggingface/transformers/tree/main/examples/distillation), RoBERTa 到 [DistilRoBERTa](https://github.com/huggingface/transformers/tree/main/examples/distillation), Multilingual BERT 到 [DistilmBERT](https://github.com/huggingface/transformers/tree/main/examples/distillation) 和德语版 DistilBERT。
 1. **[DiT](https://huggingface.co/docs/transformers/model_doc/dit)** (来自 Microsoft Research) 伴随论文 [DiT: Self-supervised Pre-training for Document Image Transformer](https://arxiv.org/abs/2203.02378) 由 Junlong Li, Yiheng Xu, Tengchao Lv, Lei Cui, Cha Zhang, Furu Wei 发布。
+1. **[Donut](https://huggingface.co/docs/transformers/model_doc/donut)** (来自 NAVER) 伴随论文 [OCR-free Document Understanding Transformer](https://arxiv.org/abs/2111.15664) 由 Geewook Kim, Teakgyu Hong, Moonbin Yim, Jeongyeon Nam, Jinyoung Park, Jinyeong Yim, Wonseok Hwang, Sangdoo Yun, Dongyoon Han, Seunghyun Park 发布。
 1. **[DPR](https://huggingface.co/docs/transformers/model_doc/dpr)** (来自 Facebook) 伴随论文 [Dense Passage Retrieval for Open-Domain Question Answering](https://arxiv.org/abs/2004.04906) 由 Vladimir Karpukhin, Barlas Oğuz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih 发布。
 1. **[DPT](https://huggingface.co/docs/transformers/master/model_doc/dpt)** (来自 Intel Labs) 伴随论文 [Vision Transformers for Dense Prediction](https://arxiv.org/abs/2103.13413) 由 René Ranftl, Alexey Bochkovskiy, Vladlen Koltun 发布。
 1. **[ELECTRA](https://huggingface.co/docs/transformers/model_doc/electra)** (来自 Google Research/Stanford University) 伴随论文 [ELECTRA: Pre-training text encoders as discriminators rather than generators](https://arxiv.org/abs/2003.10555) 由 Kevin Clark, Minh-Thang Luong, Quoc V. Le, Christopher D. Manning 发布。
 1. **[EncoderDecoder](https://huggingface.co/docs/transformers/model_doc/encoder-decoder)** (来自 Google Research) 伴随论文 [Leveraging Pre-trained Checkpoints for Sequence Generation Tasks](https://arxiv.org/abs/1907.12461) 由 Sascha Rothe, Shashi Narayan, Aliaksei Severyn 发布。
+1. **[ERNIE](https://huggingface.co/docs/transformers/model_doc/ernie)** (来自 Baidu) 伴随论文 [ERNIE: Enhanced Representation through Knowledge Integration](https://arxiv.org/abs/1904.09223) by Yu Sun, Shuohuan Wang, Yukun Li, Shikun Feng, Xuyi Chen, Han Zhang, Xin Tian, Danxiang Zhu, Hao Tian, Hua Wu 发布。
+1. **[ESM](https://huggingface.co/docs/transformers/model_doc/esm)** (from Meta AI) are transformer protein language models.  **ESM-1b** was released with the paper [Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences](https://www.pnas.org/content/118/15/e2016239118) by Alexander Rives, Joshua Meier, Tom Sercu, Siddharth Goyal, Zeming Lin, Jason Liu, Demi Guo, Myle Ott, C. Lawrence Zitnick, Jerry Ma, and Rob Fergus. **ESM-1v** was released with the paper [Language models enable zero-shot prediction of the effects of mutations on protein function](https://doi.org/10.1101/2021.07.09.450648) by Joshua Meier, Roshan Rao, Robert Verkuil, Jason Liu, Tom Sercu and Alexander Rives. **ESM-2** was released with the paper [Language models of protein sequences at the scale of evolution enable accurate structure prediction](https://doi.org/10.1101/2022.07.20.500902) by Zeming Lin, Halil Akin, Roshan Rao, Brian Hie, Zhongkai Zhu, Wenting Lu, Allan dos Santos Costa, Maryam Fazel-Zarandi, Tom Sercu, Sal Candido, Alexander Rives.
 1. **[FlauBERT](https://huggingface.co/docs/transformers/model_doc/flaubert)** (来自 CNRS) 伴随论文 [FlauBERT: Unsupervised Language Model Pre-training for French](https://arxiv.org/abs/1912.05372) 由 Hang Le, Loïc Vial, Jibril Frej, Vincent Segonne, Maximin Coavoux, Benjamin Lecouteux, Alexandre Allauzen, Benoît Crabbé, Laurent Besacier, Didier Schwab 发布。
 1. **[FLAVA](https://huggingface.co/docs/transformers/model_doc/flava)** (来自 Facebook AI) 伴随论文 [FLAVA: A Foundational Language And Vision Alignment Model](https://arxiv.org/abs/2112.04482) 由 Amanpreet Singh, Ronghang Hu, Vedanuj Goswami, Guillaume Couairon, Wojciech Galuba, Marcus Rohrbach, and Douwe Kiela 发布。
 1. **[FNet](https://huggingface.co/docs/transformers/model_doc/fnet)** (来自 Google Research) 伴随论文 [FNet: Mixing Tokens with Fourier Transforms](https://arxiv.org/abs/2105.03824) 由 James Lee-Thorp, Joshua Ainslie, Ilya Eckstein, Santiago Ontanon 发布。
@@ -277,7 +282,8 @@ conda install -c huggingface transformers
 1. **[GLPN](https://huggingface.co/docs/transformers/model_doc/glpn)** (来自 KAIST) 伴随论文 [Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth](https://arxiv.org/abs/2201.07436) 由 Doyeon Kim, Woonghyun Ga, Pyungwhan Ahn, Donggyu Joo, Sehwan Chun, Junmo Kim 发布。
 1. **[GPT](https://huggingface.co/docs/transformers/model_doc/openai-gpt)** (来自 OpenAI) 伴随论文 [Improving Language Understanding by Generative Pre-Training](https://blog.openai.com/language-unsupervised/) 由 Alec Radford, Karthik Narasimhan, Tim Salimans and Ilya Sutskever 发布。
 1. **[GPT Neo](https://huggingface.co/docs/transformers/model_doc/gpt_neo)** (来自 EleutherAI) 随仓库 [EleutherAI/gpt-neo](https://github.com/EleutherAI/gpt-neo) 发布。作者为 Sid Black, Stella Biderman, Leo Gao, Phil Wang and Connor Leahy 发布。
-1. **[GPT NeoX](https://huggingface.co/docs/transformers/model_doc/gpt_neox)** (from EleutherAI) released with the paper [GPT-NeoX-20B: An Open-Source Autoregressive Language Model](https://arxiv.org/abs/2204.06745) by Sid Black, Stella Biderman, Eric Hallahan, Quentin Anthony, Leo Gao, Laurence Golding, Horace He, Connor Leahy, Kyle McDonell, Jason Phang, Michael Pieler, USVSN Sai Prashanth, Shivanshu Purohit, Laria Reynolds, Jonathan Tow, Ben Wang, Samuel Weinbach 
+1. **[GPT NeoX](https://huggingface.co/docs/transformers/model_doc/gpt_neox)** (from EleutherAI) released with the paper [GPT-NeoX-20B: An Open-Source Autoregressive Language Model](https://arxiv.org/abs/2204.06745) by Sid Black, Stella Biderman, Eric Hallahan, Quentin Anthony, Leo Gao, Laurence Golding, Horace He, Connor Leahy, Kyle McDonell, Jason Phang, Michael Pieler, USVSN Sai Prashanth, Shivanshu Purohit, Laria Reynolds, Jonathan Tow, Ben Wang, Samuel Weinbach
+1. **[GPT NeoX Japanese](https://huggingface.co/docs/transformers/model_doc/gpt_neox_japanese)** (来自 ABEJA) 由 Shinya Otani, Takayoshi Makabe, Anuj Arora, Kyo Hattori。
 1. **[GPT-2](https://huggingface.co/docs/transformers/model_doc/gpt2)** (来自 OpenAI) 伴随论文 [Language Models are Unsupervised Multitask Learners](https://blog.openai.com/better-language-models/) 由 Alec Radford*, Jeffrey Wu*, Rewon Child, David Luan, Dario Amodei** and Ilya Sutskever** 发布。
 1. **[GPT-J](https://huggingface.co/docs/transformers/model_doc/gptj)** (来自 EleutherAI) 伴随论文 [kingoflolz/mesh-transformer-jax](https://github.com/kingoflolz/mesh-transformer-jax/) 由 Ben Wang and Aran Komatsuzaki 发布。
 1. **[GroupViT](https://huggingface.co/docs/transformers/model_doc/groupvit)** (来自 UCSD, NVIDIA) 伴随论文 [GroupViT: Semantic Segmentation Emerges from Text Supervision](https://arxiv.org/abs/2202.11094) 由 Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, Xiaolong Wang 发布。
@@ -287,7 +293,7 @@ conda install -c huggingface transformers
 1. **[LayoutLM](https://huggingface.co/docs/transformers/model_doc/layoutlm)** (来自 Microsoft Research Asia) 伴随论文 [LayoutLM: Pre-training of Text and Layout for Document Image Understanding](https://arxiv.org/abs/1912.13318) 由 Yiheng Xu, Minghao Li, Lei Cui, Shaohan Huang, Furu Wei, Ming Zhou 发布。
 1. **[LayoutLMv2](https://huggingface.co/docs/transformers/model_doc/layoutlmv2)** (来自 Microsoft Research Asia) 伴随论文 [LayoutLMv2: Multi-modal Pre-training for Visually-Rich Document Understanding](https://arxiv.org/abs/2012.14740) 由 Yang Xu, Yiheng Xu, Tengchao Lv, Lei Cui, Furu Wei, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Wanxiang Che, Min Zhang, Lidong Zhou 发布。
 1. **[LayoutLMv3](https://huggingface.co/docs/transformers/model_doc/layoutlmv3)** (来自 Microsoft Research Asia) 伴随论文 [LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking](https://arxiv.org/abs/2204.08387) 由 Yupan Huang, Tengchao Lv, Lei Cui, Yutong Lu, Furu Wei 发布。
-1. **[LayoutXLM](https://huggingface.co/docs/transformers/model_doc/layoutlmv2)** (来自 Microsoft Research Asia) 伴随论文 [LayoutXLM: Multimodal Pre-training for Multilingual Visually-rich Document Understanding](https://arxiv.org/abs/2104.08836) 由 Yiheng Xu, Tengchao Lv, Lei Cui, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Furu Wei 发布。
+1. **[LayoutXLM](https://huggingface.co/docs/transformers/model_doc/layoutxlm)** (来自 Microsoft Research Asia) 伴随论文 [LayoutXLM: Multimodal Pre-training for Multilingual Visually-rich Document Understanding](https://arxiv.org/abs/2104.08836) 由 Yiheng Xu, Tengchao Lv, Lei Cui, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Furu Wei 发布。
 1. **[LED](https://huggingface.co/docs/transformers/model_doc/led)** (来自 AllenAI) 伴随论文 [Longformer: The Long-Document Transformer](https://arxiv.org/abs/2004.05150) 由 Iz Beltagy, Matthew E. Peters, Arman Cohan 发布。
 1. **[LeViT](https://huggingface.co/docs/transformers/model_doc/levit)** (来自 Meta AI) 伴随论文 [LeViT: A Vision Transformer in ConvNet's Clothing for Faster Inference](https://arxiv.org/abs/2104.01136) 由 Ben Graham, Alaaeldin El-Nouby, Hugo Touvron, Pierre Stock, Armand Joulin, Hervé Jégou, Matthijs Douze 发布。
 1. **[Longformer](https://huggingface.co/docs/transformers/model_doc/longformer)** (来自 AllenAI) 伴随论文 [Longformer: The Long-Document Transformer](https://arxiv.org/abs/2004.05150) 由 Iz Beltagy, Matthew E. Peters, Arman Cohan 发布。
@@ -297,7 +303,8 @@ conda install -c huggingface transformers
 1. **[M-CTC-T](https://huggingface.co/docs/transformers/model_doc/mctct)** (来自 Facebook) 伴随论文 [Pseudo-Labeling For Massively Multilingual Speech Recognition](https://arxiv.org/abs/2111.00161) 由 Loren Lugosch, Tatiana Likhomanenko, Gabriel Synnaeve, and Ronan Collobert 发布。
 1. **[M2M100](https://huggingface.co/docs/transformers/model_doc/m2m_100)** (来自 Facebook) 伴随论文 [Beyond English-Centric Multilingual Machine Translation](https://arxiv.org/abs/2010.11125) 由 Angela Fan, Shruti Bhosale, Holger Schwenk, Zhiyi Ma, Ahmed El-Kishky, Siddharth Goyal, Mandeep Baines, Onur Celebi, Guillaume Wenzek, Vishrav Chaudhary, Naman Goyal, Tom Birch, Vitaliy Liptchinsky, Sergey Edunov, Edouard Grave, Michael Auli, Armand Joulin 发布。
 1. **[MarianMT](https://huggingface.co/docs/transformers/model_doc/marian)** 用 [OPUS](http://opus.nlpl.eu/) 数据训练的机器翻译模型由 Jörg Tiedemann 发布。[Marian Framework](https://marian-nmt.github.io/) 由微软翻译团队开发。
-1. **[MaskFormer](https://huggingface.co/docs/transformers/model_doc/maskformer)** (from Meta and UIUC) released with the paper [Per-Pixel Classification is Not All You Need for Semantic Segmentation](https://arxiv.org/abs/2107.06278) by Bowen Cheng, Alexander G. Schwing, Alexander Kirillov 
+1. **[MarkupLM](https://huggingface.co/docs/transformers/model_doc/markuplm)** (来自 Microsoft Research Asia) 伴随论文 [MarkupLM: Pre-training of Text and Markup Language for Visually-rich Document Understanding](https://arxiv.org/abs/2110.08518) 由 Junlong Li, Yiheng Xu, Lei Cui, Furu Wei 发布。
+1. **[MaskFormer](https://huggingface.co/docs/transformers/model_doc/maskformer)** (from Meta and UIUC) released with the paper [Per-Pixel Classification is Not All You Need for Semantic Segmentation](https://arxiv.org/abs/2107.06278) by Bowen Cheng, Alexander G. Schwing, Alexander Kirillov  >>>>>>> Fix rebase
 1. **[mBART](https://huggingface.co/docs/transformers/model_doc/mbart)** (来自 Facebook) 伴随论文 [Multilingual Denoising Pre-training for Neural Machine Translation](https://arxiv.org/abs/2001.08210) 由 Yinhan Liu, Jiatao Gu, Naman Goyal, Xian Li, Sergey Edunov, Marjan Ghazvininejad, Mike Lewis, Luke Zettlemoyer 发布。
 1. **[mBART-50](https://huggingface.co/docs/transformers/model_doc/mbart)** (来自 Facebook) 伴随论文 [Multilingual Translation with Extensible Multilingual Pretraining and Finetuning](https://arxiv.org/abs/2008.00401) 由 Yuqing Tang, Chau Tran, Xian Li, Peng-Jen Chen, Naman Goyal, Vishrav Chaudhary, Jiatao Gu, Angela Fan 发布。
 1. **[Megatron-BERT](https://huggingface.co/docs/transformers/model_doc/megatron-bert)** (来自 NVIDIA) 伴随论文 [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) 由 Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro 发布。
@@ -307,13 +314,14 @@ conda install -c huggingface transformers
 1. **[MobileViT](https://huggingface.co/docs/transformers/model_doc/mobilevit)** (来自 Apple) 伴随论文 [MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer](https://arxiv.org/abs/2110.02178) 由 Sachin Mehta and Mohammad Rastegari 发布。
 1. **[MPNet](https://huggingface.co/docs/transformers/model_doc/mpnet)** (来自 Microsoft Research) 伴随论文 [MPNet: Masked and Permuted Pre-training for Language Understanding](https://arxiv.org/abs/2004.09297) 由 Kaitao Song, Xu Tan, Tao Qin, Jianfeng Lu, Tie-Yan Liu 发布。
 1. **[MT5](https://huggingface.co/docs/transformers/model_doc/mt5)** (来自 Google AI) 伴随论文 [mT5: A massively multilingual pre-trained text-to-text transformer](https://arxiv.org/abs/2010.11934) 由 Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, Colin Raffel 发布。
-1. **[MVP](https://huggingface.co/docs/transformers/model_doc/mvp)** (来自 中国人民大学 AI Box) 伴随论文 [MVP: Multi-task Supervised Pre-training for Natural Language Generation](https://arxiv.org/abs/2206.12131) 由 Tianyi Tang, Junyi Li, Wayne Xin Zhao and Ji-Rong Wen 发布。 
+1. **[MVP](https://huggingface.co/docs/transformers/model_doc/mvp)** (来自 中国人民大学 AI Box) 伴随论文 [MVP: Multi-task Supervised Pre-training for Natural Language Generation](https://arxiv.org/abs/2206.12131) 由 Tianyi Tang, Junyi Li, Wayne Xin Zhao and Ji-Rong Wen 发布。
 1. **[Nezha](https://huggingface.co/docs/transformers/model_doc/nezha)** (来自华为诺亚方舟实验室) 伴随论文 [NEZHA: Neural Contextualized Representation for Chinese Language Understanding](https://arxiv.org/abs/1909.00204) 由 Junqiu Wei, Xiaozhe Ren, Xiaoguang Li, Wenyong Huang, Yi Liao, Yasheng Wang, Jiashu Lin, Xin Jiang, Xiao Chen and Qun Liu 发布。
 1. **[NLLB](https://huggingface.co/docs/transformers/model_doc/nllb)** (来自 Meta) 伴随论文 [No Language Left Behind: Scaling Human-Centered Machine Translation](https://arxiv.org/abs/2207.04672) 由 the NLLB team 发布。
 1. **[Nyströmformer](https://huggingface.co/docs/transformers/model_doc/nystromformer)** (来自 the University of Wisconsin - Madison) 伴随论文 [Nyströmformer: A Nyström-Based Algorithm for Approximating Self-Attention](https://arxiv.org/abs/2102.03902) 由 Yunyang Xiong, Zhanpeng Zeng, Rudrasis Chakraborty, Mingxing Tan, Glenn Fung, Yin Li, Vikas Singh 发布。
 1. **[OPT](https://huggingface.co/docs/transformers/master/model_doc/opt)** (来自 Meta AI) 伴随论文 [OPT: Open Pre-trained Transformer Language Models](https://arxiv.org/abs/2205.01068) 由 Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen et al 发布。
 1. **[OWL-ViT](https://huggingface.co/docs/transformers/model_doc/owlvit)** (来自 Google AI) 伴随论文 [Simple Open-Vocabulary Object Detection with Vision Transformers](https://arxiv.org/abs/2205.06230) 由 Matthias Minderer, Alexey Gritsenko, Austin Stone, Maxim Neumann, Dirk Weissenborn, Alexey Dosovitskiy, Aravindh Mahendran, Anurag Arnab, Mostafa Dehghani, Zhuoran Shen, Xiao Wang, Xiaohua Zhai, Thomas Kipf, and Neil Houlsby 发布。
 1. **[Pegasus](https://huggingface.co/docs/transformers/model_doc/pegasus)** (来自 Google) 伴随论文 [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) 由 Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu 发布。
+1. **[PEGASUS-X](https://huggingface.co/docs/transformers/model_doc/pegasus_x)** (来自 Google) 伴随论文 [Investigating Efficiently Extending Transformers for Long Input Summarization](https://arxiv.org/abs/2208.04347) 由 Jason Phang, Yao Zhao, Peter J. Liu 发布。
 1. **[Perceiver IO](https://huggingface.co/docs/transformers/model_doc/perceiver)** (来自 Deepmind) 伴随论文 [Perceiver IO: A General Architecture for Structured Inputs & Outputs](https://arxiv.org/abs/2107.14795) 由 Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier Hénaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, João Carreira 发布。
 1. **[PhoBERT](https://huggingface.co/docs/transformers/model_doc/phobert)** (来自 VinAI Research) 伴随论文 [PhoBERT: Pre-trained language models for Vietnamese](https://www.aclweb.org/anthology/2020.findings-emnlp.92/) 由 Dat Quoc Nguyen and Anh Tuan Nguyen 发布。
 1. **[PLBart](https://huggingface.co/docs/transformers/model_doc/plbart)** (来自 UCLA NLP) 伴随论文 [Unified Pre-training for Program Understanding and Generation](https://arxiv.org/abs/2103.06333) 由 Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, Kai-Wei Chang 发布。
@@ -323,9 +331,9 @@ conda install -c huggingface transformers
 1. **[RAG](https://huggingface.co/docs/transformers/model_doc/rag)** (来自 Facebook) 伴随论文 [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://arxiv.org/abs/2005.11401) 由 Patrick Lewis, Ethan Perez, Aleksandara Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, Douwe Kiela 发布。
 1. **[REALM](https://huggingface.co/docs/transformers/model_doc/realm.html)** (来自 Google Research) 伴随论文 [REALM: Retrieval-Augmented Language Model Pre-Training](https://arxiv.org/abs/2002.08909) 由 Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat and Ming-Wei Chang 发布。
 1. **[Reformer](https://huggingface.co/docs/transformers/model_doc/reformer)** (来自 Google Research) 伴随论文 [Reformer: The Efficient Transformer](https://arxiv.org/abs/2001.04451) 由 Nikita Kitaev, Łukasz Kaiser, Anselm Levskaya 发布。
-1. **[RegNet](https://huggingface.co/docs/transformers/model_doc/regnet)** (from META Research) released with the paper [Designing Network Design Space](https://arxiv.org/abs/2003.13678) by Ilija Radosavovic, Raj Prateek Kosaraju, Ross Girshick, Kaiming He, Piotr Dollár. 
+1. **[RegNet](https://huggingface.co/docs/transformers/model_doc/regnet)** (from META Research) released with the paper [Designing Network Design Space](https://arxiv.org/abs/2003.13678) by Ilija Radosavovic, Raj Prateek Kosaraju, Ross Girshick, Kaiming He, Piotr Dollár.
 1. **[RemBERT](https://huggingface.co/docs/transformers/model_doc/rembert)** (来自 Google Research) 伴随论文 [Rethinking embedding coupling in pre-trained language models](https://arxiv.org/pdf/2010.12821.pdf) 由 Hyung Won Chung, Thibault Févry, Henry Tsai, M. Johnson, Sebastian Ruder 发布。
-1. **[ResNet](https://huggingface.co/docs/transformers/model_doc/resnet)** (from Microsoft Research) released with the paper [Deep Residual Learning for Image Recognition](https://arxiv.org/abs/1512.03385) by Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun. 
+1. **[ResNet](https://huggingface.co/docs/transformers/model_doc/resnet)** (from Microsoft Research) released with the paper [Deep Residual Learning for Image Recognition](https://arxiv.org/abs/1512.03385) by Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun.
 1. **[RoBERTa](https://huggingface.co/docs/transformers/model_doc/roberta)** (来自 Facebook), 伴随论文 [Robustly Optimized BERT Pretraining Approach](https://arxiv.org/abs/1907.11692) 由 Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, Veselin Stoyanov 发布。
 1. **[RoFormer](https://huggingface.co/docs/transformers/model_doc/roformer)** (来自 ZhuiyiTechnology), 伴随论文 [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://arxiv.org/pdf/2104.09864v1.pdf) 由 Jianlin Su and Yu Lu and Shengfeng Pan and Bo Wen and Yunfeng Liu 发布。
 1. **[SegFormer](https://huggingface.co/docs/transformers/model_doc/segformer)** (来自 NVIDIA) 伴随论文 [SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers](https://arxiv.org/abs/2105.15203) 由 Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M. Alvarez, Ping Luo 发布。
@@ -336,27 +344,32 @@ conda install -c huggingface transformers
 1. **[Splinter](https://huggingface.co/docs/transformers/model_doc/splinter)** (来自 Tel Aviv University) 伴随论文 [Few-Shot Question Answering by Pretraining Span Selection](https://arxiv.org/abs/2101.00438) 由 Ori Ram, Yuval Kirstain, Jonathan Berant, Amir Globerson, Omer Levy 发布。
 1. **[SqueezeBERT](https://huggingface.co/docs/transformers/model_doc/squeezebert)** (来自 Berkeley) 伴随论文 [SqueezeBERT: What can computer vision teach NLP about efficient neural networks?](https://arxiv.org/abs/2006.11316) 由 Forrest N. Iandola, Albert E. Shaw, Ravi Krishna, and Kurt W. Keutzer 发布。
 1. **[Swin Transformer](https://huggingface.co/docs/transformers/model_doc/swin)** (来自 Microsoft) 伴随论文 [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://arxiv.org/abs/2103.14030) 由 Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, Baining Guo 发布。
-1. **[Swin Transformer V2](https://huggingface.co/docs/transformers/main/model_doc/swinv2)** (来自 Microsoft) 伴随论文 [Swin Transformer V2: Scaling Up Capacity and Resolution](https://arxiv.org/abs/2111.09883) 由 Ze Liu, Han Hu, Yutong Lin, Zhuliang Yao, Zhenda Xie, Yixuan Wei, Jia Ning, Yue Cao, Zheng Zhang, Li Dong, Furu Wei, Baining Guo 发布。
+1. **[Swin Transformer V2](https://huggingface.co/docs/transformers/model_doc/swinv2)** (来自 Microsoft) 伴随论文 [Swin Transformer V2: Scaling Up Capacity and Resolution](https://arxiv.org/abs/2111.09883) 由 Ze Liu, Han Hu, Yutong Lin, Zhuliang Yao, Zhenda Xie, Yixuan Wei, Jia Ning, Yue Cao, Zheng Zhang, Li Dong, Furu Wei, Baining Guo 发布。
 1. **[T5](https://huggingface.co/docs/transformers/model_doc/t5)** (来自 Google AI) 伴随论文 [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://arxiv.org/abs/1910.10683) 由 Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu 发布。
 1. **[T5v1.1](https://huggingface.co/docs/transformers/model_doc/t5v1.1)** (来自 Google AI) 伴随论文 [google-research/text-to-text-transfer-transformer](https://github.com/google-research/text-to-text-transfer-transformer/blob/main/released_checkpoints.md#t511) 由 Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu 发布。
 1. **[TAPAS](https://huggingface.co/docs/transformers/model_doc/tapas)** (来自 Google AI) 伴随论文 [TAPAS: Weakly Supervised Table Parsing via Pre-training](https://arxiv.org/abs/2004.02349) 由 Jonathan Herzig, Paweł Krzysztof Nowak, Thomas Müller, Francesco Piccinno and Julian Martin Eisenschlos 发布。
 1. **[TAPEX](https://huggingface.co/docs/transformers/model_doc/tapex)** (来自 Microsoft Research) 伴随论文 [TAPEX: Table Pre-training via Learning a Neural SQL Executor](https://arxiv.org/abs/2107.07653) 由 Qian Liu, Bei Chen, Jiaqi Guo, Morteza Ziyadi, Zeqi Lin, Weizhu Chen, Jian-Guang Lou 发布。
-1. **[Trajectory Transformer](https://huggingface.co/docs/transformers/model_doc/trajectory_transformers)** (from the University of California at Berkeley) released with the paper [Offline Reinforcement Learning as One Big Sequence Modeling Problem](https://arxiv.org/abs/2106.02039) by Michael Janner, Qiyang Li, Sergey Levine 
+1. **[Time Series Transformer](https://huggingface.co/docs/transformers/model_doc/time_series_transformer)** (from HuggingFace).
+1. **[Trajectory Transformer](https://huggingface.co/docs/transformers/model_doc/trajectory_transformers)** (from the University of California at Berkeley) released with the paper [Offline Reinforcement Learning as One Big Sequence Modeling Problem](https://arxiv.org/abs/2106.02039) by Michael Janner, Qiyang Li, Sergey Levine
 1. **[Transformer-XL](https://huggingface.co/docs/transformers/model_doc/transfo-xl)** (来自 Google/CMU) 伴随论文 [Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context](https://arxiv.org/abs/1901.02860) 由 Zihang Dai*, Zhilin Yang*, Yiming Yang, Jaime Carbonell, Quoc V. Le, Ruslan Salakhutdinov 发布。
 1. **[TrOCR](https://huggingface.co/docs/transformers/model_doc/trocr)** (来自 Microsoft) 伴随论文 [TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models](https://arxiv.org/abs/2109.10282) 由 Minghao Li, Tengchao Lv, Lei Cui, Yijuan Lu, Dinei Florencio, Cha Zhang, Zhoujun Li, Furu Wei 发布。
-1. **[UL2](https://huggingface.co/docs/transformers/model_doc/ul2)** (from Google Research) released with the paper [Unifying Language Learning Paradigms](https://arxiv.org/abs/2205.05131v1) by Yi Tay, Mostafa Dehghani, Vinh Q. Tran, Xavier Garcia, Dara Bahri, Tal Schuster, Huaixiu Steven Zheng, Neil Houlsby, Donald Metzler 
+1. **[UL2](https://huggingface.co/docs/transformers/model_doc/ul2)** (from Google Research) released with the paper [Unifying Language Learning Paradigms](https://arxiv.org/abs/2205.05131v1) by Yi Tay, Mostafa Dehghani, Vinh Q. Tran, Xavier Garcia, Dara Bahri, Tal Schuster, Huaixiu Steven Zheng, Neil Houlsby, Donald Metzler
 1. **[UniSpeech](https://huggingface.co/docs/transformers/model_doc/unispeech)** (来自 Microsoft Research) 伴随论文 [UniSpeech: Unified Speech Representation Learning with Labeled and Unlabeled Data](https://arxiv.org/abs/2101.07597) 由 Chengyi Wang, Yu Wu, Yao Qian, Kenichi Kumatani, Shujie Liu, Furu Wei, Michael Zeng, Xuedong Huang 发布。
 1. **[UniSpeechSat](https://huggingface.co/docs/transformers/model_doc/unispeech-sat)** (来自 Microsoft Research) 伴随论文 [UNISPEECH-SAT: UNIVERSAL SPEECH REPRESENTATION LEARNING WITH SPEAKER AWARE PRE-TRAINING](https://arxiv.org/abs/2110.05752) 由 Sanyuan Chen, Yu Wu, Chengyi Wang, Zhengyang Chen, Zhuo Chen, Shujie Liu, Jian Wu, Yao Qian, Furu Wei, Jinyu Li, Xiangzhan Yu 发布。
 1. **[VAN](https://huggingface.co/docs/transformers/model_doc/van)** (来自 Tsinghua University and Nankai University) 伴随论文 [Visual Attention Network](https://arxiv.org/pdf/2202.09741.pdf) 由 Meng-Hao Guo, Cheng-Ze Lu, Zheng-Ning Liu, Ming-Ming Cheng, Shi-Min Hu 发布。
+1. **[VideoMAE](https://huggingface.co/docs/transformers/model_doc/videomae)** (来自 Multimedia Computing Group, Nanjing University) 伴随论文 [VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training](https://arxiv.org/abs/2203.12602) 由 Zhan Tong, Yibing Song, Jue Wang, Limin Wang 发布。
 1. **[ViLT](https://huggingface.co/docs/transformers/model_doc/vilt)** (来自 NAVER AI Lab/Kakao Enterprise/Kakao Brain) 伴随论文 [ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision](https://arxiv.org/abs/2102.03334) 由 Wonjae Kim, Bokyung Son, Ildoo Kim 发布。
 1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (来自 Google AI) 伴随论文 [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) 由 Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby 发布。
 1. **[VisualBERT](https://huggingface.co/docs/transformers/model_doc/visual_bert)** (来自 UCLA NLP) 伴随论文 [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) 由 Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang 发布。
 1. **[ViTMAE](https://huggingface.co/docs/transformers/model_doc/vit_mae)** (来自 Meta AI) 伴随论文 [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) 由 Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick 发布。
+1. **[ViTMSN](https://huggingface.co/docs/transformers/model_doc/vit_msn)** (来自 Meta AI) 伴随论文 [Masked Siamese Networks for Label-Efficient Learning](https://arxiv.org/abs/2204.07141) by Mahmoud Assran, Mathilde Caron, Ishan Misra, Piotr Bojanowski, Florian Bordes, Pascal Vincent, Armand Joulin, Michael Rabbat, Nicolas Ballas 发布.
 1. **[Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/wav2vec2)** (来自 Facebook AI) 伴随论文 [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) 由 Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli 发布。
 1. **[Wav2Vec2-Conformer](https://huggingface.co/docs/transformers/model_doc/wav2vec2-conformer)** (来自 Facebook AI) 伴随论文 [FAIRSEQ S2T: Fast Speech-to-Text Modeling with FAIRSEQ](https://arxiv.org/abs/2010.05171) 由 Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Sravya Popuri, Dmytro Okhonko, Juan Pino 发布。
 1. **[Wav2Vec2Phoneme](https://huggingface.co/docs/transformers/model_doc/wav2vec2_phoneme)** (来自 Facebook AI) 伴随论文 [Simple and Effective Zero-shot Cross-lingual Phoneme Recognition](https://arxiv.org/abs/2109.11680) 由 Qiantong Xu, Alexei Baevski, Michael Auli 发布。
 1. **[WavLM](https://huggingface.co/docs/transformers/model_doc/wavlm)** (from Microsoft Research) released with the paper [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://arxiv.org/abs/2110.13900) by Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng, Furu Wei.
-1. **[XGLM](https://huggingface.co/docs/transformers/model_doc/xglm)** (From Facebook AI) released with the paper [Few-shot Learning with Multilingual Language Models](https://arxiv.org/abs/2112.10668) by Xi Victoria Lin, Todor Mihaylov, Mikel Artetxe, Tianlu Wang, Shuohui Chen, Daniel Simig, Myle Ott, Naman Goyal, Shruti Bhosale, Jingfei Du, Ramakanth Pasunuru, Sam Shleifer, Punit Singh Koura, Vishrav Chaudhary, Brian O'Horo, Jeff Wang, Luke Zettlemoyer, Zornitsa Kozareva, Mona Diab, Veselin Stoyanov, Xian Li. 
+1. **[Whisper](https://huggingface.co/docs/transformers/model_doc/whisper)** (来自 OpenAI) 伴随论文 [Robust Speech Recognition via Large-Scale Weak Supervision](https://cdn.openai.com/papers/whisper.pdf) 由 Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, Ilya Sutskever 发布。
+1. **[X-CLIP](https://huggingface.co/docs/transformers/model_doc/xclip)** (来自 Microsoft Research) 伴随论文 [Expanding Language-Image Pretrained Models for General Video Recognition](https://arxiv.org/abs/2208.02816) 由 Bolin Ni, Houwen Peng, Minghao Chen, Songyang Zhang, Gaofeng Meng, Jianlong Fu, Shiming Xiang, Haibin Ling 发布。
+1. **[XGLM](https://huggingface.co/docs/transformers/model_doc/xglm)** (From Facebook AI) released with the paper [Few-shot Learning with Multilingual Language Models](https://arxiv.org/abs/2112.10668) by Xi Victoria Lin, Todor Mihaylov, Mikel Artetxe, Tianlu Wang, Shuohui Chen, Daniel Simig, Myle Ott, Naman Goyal, Shruti Bhosale, Jingfei Du, Ramakanth Pasunuru, Sam Shleifer, Punit Singh Koura, Vishrav Chaudhary, Brian O'Horo, Jeff Wang, Luke Zettlemoyer, Zornitsa Kozareva, Mona Diab, Veselin Stoyanov, Xian Li.
 1. **[XLM](https://huggingface.co/docs/transformers/model_doc/xlm)** (来自 Facebook) 伴随论文 [Cross-lingual Language Model Pretraining](https://arxiv.org/abs/1901.07291) 由 Guillaume Lample and Alexis Conneau 发布。
 1. **[XLM-ProphetNet](https://huggingface.co/docs/transformers/model_doc/xlm-prophetnet)** (来自 Microsoft Research) 伴随论文 [ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training](https://arxiv.org/abs/2001.04063) 由 Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou 发布。
 1. **[XLM-RoBERTa](https://huggingface.co/docs/transformers/model_doc/xlm-roberta)** (来自 Facebook AI), 伴随论文 [Unsupervised Cross-lingual Representation Learning at Scale](https://arxiv.org/abs/1911.02116) 由 Alexis Conneau*, Kartikay Khandelwal*, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer and Veselin Stoyanov 发布。

README_zh-hant.md

@@ -185,7 +185,7 @@ Tokenizer 為所有的預訓練模型提供了預處理，並可以直接轉換
     - 對所有模型使用的制式化API
 
 1. 更低的運算成本，更少的碳排放：
-    - 研究人員可以分享預訓練的模型而非從頭開始訓練
+    - 研究人員可以分享已訓練的模型而非每次從頭開始訓練
     - 工程師可以減少計算時間以及生產成本
     - 數十種模型架構、兩千多個預訓練模型、100多種語言支援
 
@@ -257,13 +257,14 @@ conda install -c huggingface transformers
 1. **[BigBird-RoBERTa](https://huggingface.co/docs/transformers/model_doc/big_bird)** (from Google Research) released with the paper [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed.
 1. **[Blenderbot](https://huggingface.co/docs/transformers/model_doc/blenderbot)** (from Facebook) released with the paper [Recipes for building an open-domain chatbot](https://arxiv.org/abs/2004.13637) by Stephen Roller, Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Kurt Shuster, Eric M. Smith, Y-Lan Boureau, Jason Weston.
 1. **[BlenderbotSmall](https://huggingface.co/docs/transformers/model_doc/blenderbot-small)** (from Facebook) released with the paper [Recipes for building an open-domain chatbot](https://arxiv.org/abs/2004.13637) by Stephen Roller, Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Kurt Shuster, Eric M. Smith, Y-Lan Boureau, Jason Weston.
-1. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigSicence Workshop](https://bigscience.huggingface.co/). 
+1. **[BLOOM](https://huggingface.co/docs/transformers/model_doc/bloom)** (from BigScience workshop) released by the [BigSicence Workshop](https://bigscience.huggingface.co/).
 1. **[BORT](https://huggingface.co/docs/transformers/model_doc/bort)** (from Alexa) released with the paper [Optimal Subarchitecture Extraction For BERT](https://arxiv.org/abs/2010.10499) by Adrian de Wynter and Daniel J. Perry.
 1. **[ByT5](https://huggingface.co/docs/transformers/model_doc/byt5)** (from Google Research) released with the paper [ByT5: Towards a token-free future with pre-trained byte-to-byte models](https://arxiv.org/abs/2105.13626) by Linting Xue, Aditya Barua, Noah Constant, Rami Al-Rfou, Sharan Narang, Mihir Kale, Adam Roberts, Colin Raffel.
 1. **[CamemBERT](https://huggingface.co/docs/transformers/model_doc/camembert)** (from Inria/Facebook/Sorbonne) released with the paper [CamemBERT: a Tasty French Language Model](https://arxiv.org/abs/1911.03894) by Louis Martin*, Benjamin Muller*, Pedro Javier Ortiz Suárez*, Yoann Dupont, Laurent Romary, Éric Villemonte de la Clergerie, Djamé Seddah and Benoît Sagot.
 1. **[CANINE](https://huggingface.co/docs/transformers/model_doc/canine)** (from Google Research) released with the paper [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) by Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting.
 1. **[CLIP](https://huggingface.co/docs/transformers/model_doc/clip)** (from OpenAI) released with the paper [Learning Transferable Visual Models From Natural Language Supervision](https://arxiv.org/abs/2103.00020) by Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, Ilya Sutskever.
 1. **[CodeGen](https://huggingface.co/docs/transformers/model_doc/codegen)** (from Salesforce) released with the paper [A Conversational Paradigm for Program Synthesis](https://arxiv.org/abs/2203.13474) by Erik Nijkamp, Bo Pang, Hiroaki Hayashi, Lifu Tu, Huan Wang, Yingbo Zhou, Silvio Savarese, Caiming Xiong.
+1. **[Conditional DETR](https://huggingface.co/docs/transformers/model_doc/conditional_detr)** (from Microsoft Research Asia) released with the paper [Conditional DETR for Fast Training Convergence](https://arxiv.org/abs/2108.06152) by Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, Jingdong Wang.
 1. **[ConvBERT](https://huggingface.co/docs/transformers/model_doc/convbert)** (from YituTech) released with the paper [ConvBERT: Improving BERT with Span-based Dynamic Convolution](https://arxiv.org/abs/2008.02496) by Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan.
 1. **[ConvNeXT](https://huggingface.co/docs/transformers/model_doc/convnext)** (from Facebook AI) released with the paper [A ConvNet for the 2020s](https://arxiv.org/abs/2201.03545) by Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie.
 1. **[CPM](https://huggingface.co/docs/transformers/model_doc/cpm)** (from Tsinghua University) released with the paper [CPM: A Large-scale Generative Chinese Pre-trained Language Model](https://arxiv.org/abs/2012.00413) by Zhengyan Zhang, Xu Han, Hao Zhou, Pei Ke, Yuxian Gu, Deming Ye, Yujia Qin, Yusheng Su, Haozhe Ji, Jian Guan, Fanchao Qi, Xiaozhi Wang, Yanan Zheng, Guoyang Zeng, Huanqi Cao, Shengqi Chen, Daixuan Li, Zhenbo Sun, Zhiyuan Liu, Minlie Huang, Wentao Han, Jie Tang, Juanzi Li, Xiaoyan Zhu, Maosong Sun.
@@ -273,15 +274,19 @@ conda install -c huggingface transformers
 1. **[DeBERTa](https://huggingface.co/docs/transformers/model_doc/deberta)** (from Microsoft) released with the paper [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://arxiv.org/abs/2006.03654) by Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen.
 1. **[DeBERTa-v2](https://huggingface.co/docs/transformers/model_doc/deberta-v2)** (from Microsoft) released with the paper [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://arxiv.org/abs/2006.03654) by Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen.
 1. **[Decision Transformer](https://huggingface.co/docs/transformers/model_doc/decision_transformer)** (from Berkeley/Facebook/Google) released with the paper [Decision Transformer: Reinforcement Learning via Sequence Modeling](https://arxiv.org/abs/2106.01345) by Lili Chen, Kevin Lu, Aravind Rajeswaran, Kimin Lee, Aditya Grover, Michael Laskin, Pieter Abbeel, Aravind Srinivas, Igor Mordatch.
+1. **[Deformable DETR](https://huggingface.co/docs/transformers/model_doc/deformable_detr)** (from SenseTime Research) released with the paper [Deformable DETR: Deformable Transformers for End-to-End Object Detection](https://arxiv.org/abs/2010.04159) by Xizhou Zhu, Weijie Su, Lewei Lu, Bin Li, Xiaogang Wang, Jifeng Dai.
 1. **[DeiT](https://huggingface.co/docs/transformers/model_doc/deit)** (from Facebook) released with the paper [Training data-efficient image transformers & distillation through attention](https://arxiv.org/abs/2012.12877) by Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, Hervé Jégou.
 1. **[DETR](https://huggingface.co/docs/transformers/model_doc/detr)** (from Facebook) released with the paper [End-to-End Object Detection with Transformers](https://arxiv.org/abs/2005.12872) by Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, Sergey Zagoruyko.
 1. **[DialoGPT](https://huggingface.co/docs/transformers/model_doc/dialogpt)** (from Microsoft Research) released with the paper [DialoGPT: Large-Scale Generative Pre-training for Conversational Response Generation](https://arxiv.org/abs/1911.00536) by Yizhe Zhang, Siqi Sun, Michel Galley, Yen-Chun Chen, Chris Brockett, Xiang Gao, Jianfeng Gao, Jingjing Liu, Bill Dolan.
 1. **[DistilBERT](https://huggingface.co/docs/transformers/model_doc/distilbert)** (from HuggingFace), released together with the paper [DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter](https://arxiv.org/abs/1910.01108) by Victor Sanh, Lysandre Debut and Thomas Wolf. The same method has been applied to compress GPT2 into [DistilGPT2](https://github.com/huggingface/transformers/tree/main/examples/distillation), RoBERTa into [DistilRoBERTa](https://github.com/huggingface/transformers/tree/main/examples/distillation), Multilingual BERT into [DistilmBERT](https://github.com/huggingface/transformers/tree/main/examples/distillation) and a German version of DistilBERT.
 1. **[DiT](https://huggingface.co/docs/transformers/model_doc/dit)** (from Microsoft Research) released with the paper [DiT: Self-supervised Pre-training for Document Image Transformer](https://arxiv.org/abs/2203.02378) by Junlong Li, Yiheng Xu, Tengchao Lv, Lei Cui, Cha Zhang, Furu Wei.
+1. **[Donut](https://huggingface.co/docs/transformers/model_doc/donut)** (from NAVER) released with the paper [OCR-free Document Understanding Transformer](https://arxiv.org/abs/2111.15664) by Geewook Kim, Teakgyu Hong, Moonbin Yim, Jeongyeon Nam, Jinyoung Park, Jinyeong Yim, Wonseok Hwang, Sangdoo Yun, Dongyoon Han, Seunghyun Park.
 1. **[DPR](https://huggingface.co/docs/transformers/model_doc/dpr)** (from Facebook) released with the paper [Dense Passage Retrieval for Open-Domain Question Answering](https://arxiv.org/abs/2004.04906) by Vladimir Karpukhin, Barlas Oğuz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih.
 1. **[DPT](https://huggingface.co/docs/transformers/master/model_doc/dpt)** (from Intel Labs) released with the paper [Vision Transformers for Dense Prediction](https://arxiv.org/abs/2103.13413) by René Ranftl, Alexey Bochkovskiy, Vladlen Koltun.
 1. **[ELECTRA](https://huggingface.co/docs/transformers/model_doc/electra)** (from Google Research/Stanford University) released with the paper [ELECTRA: Pre-training text encoders as discriminators rather than generators](https://arxiv.org/abs/2003.10555) by Kevin Clark, Minh-Thang Luong, Quoc V. Le, Christopher D. Manning.
 1. **[EncoderDecoder](https://huggingface.co/docs/transformers/model_doc/encoder-decoder)** (from Google Research) released with the paper [Leveraging Pre-trained Checkpoints for Sequence Generation Tasks](https://arxiv.org/abs/1907.12461) by Sascha Rothe, Shashi Narayan, Aliaksei Severyn.
+1. **[ERNIE](https://huggingface.co/docs/transformers/model_doc/ernie)** (from Baidu) released with the paper [ERNIE: Enhanced Representation through Knowledge Integration](https://arxiv.org/abs/1904.09223) by Yu Sun, Shuohuan Wang, Yukun Li, Shikun Feng, Xuyi Chen, Han Zhang, Xin Tian, Danxiang Zhu, Hao Tian, Hua Wu.
+1. **[ESM](https://huggingface.co/docs/transformers/model_doc/esm)** (from Meta AI) are transformer protein language models.  **ESM-1b** was released with the paper [Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences](https://www.pnas.org/content/118/15/e2016239118) by Alexander Rives, Joshua Meier, Tom Sercu, Siddharth Goyal, Zeming Lin, Jason Liu, Demi Guo, Myle Ott, C. Lawrence Zitnick, Jerry Ma, and Rob Fergus. **ESM-1v** was released with the paper [Language models enable zero-shot prediction of the effects of mutations on protein function](https://doi.org/10.1101/2021.07.09.450648) by Joshua Meier, Roshan Rao, Robert Verkuil, Jason Liu, Tom Sercu and Alexander Rives. **ESM-2** was released with the paper [Language models of protein sequences at the scale of evolution enable accurate structure prediction](https://doi.org/10.1101/2022.07.20.500902) by Zeming Lin, Halil Akin, Roshan Rao, Brian Hie, Zhongkai Zhu, Wenting Lu, Allan dos Santos Costa, Maryam Fazel-Zarandi, Tom Sercu, Sal Candido, Alexander Rives.
 1. **[FlauBERT](https://huggingface.co/docs/transformers/model_doc/flaubert)** (from CNRS) released with the paper [FlauBERT: Unsupervised Language Model Pre-training for French](https://arxiv.org/abs/1912.05372) by Hang Le, Loïc Vial, Jibril Frej, Vincent Segonne, Maximin Coavoux, Benjamin Lecouteux, Alexandre Allauzen, Benoît Crabbé, Laurent Besacier, Didier Schwab.
 1. **[FLAVA](https://huggingface.co/docs/transformers/model_doc/flava)** (from Facebook AI) released with the paper [FLAVA: A Foundational Language And Vision Alignment Model](https://arxiv.org/abs/2112.04482) by Amanpreet Singh, Ronghang Hu, Vedanuj Goswami, Guillaume Couairon, Wojciech Galuba, Marcus Rohrbach, and Douwe Kiela.
 1. **[FNet](https://huggingface.co/docs/transformers/model_doc/fnet)** (from Google Research) released with the paper [FNet: Mixing Tokens with Fourier Transforms](https://arxiv.org/abs/2105.03824) by James Lee-Thorp, Joshua Ainslie, Ilya Eckstein, Santiago Ontanon.
@@ -289,7 +294,8 @@ conda install -c huggingface transformers
 1. **[GLPN](https://huggingface.co/docs/transformers/model_doc/glpn)** (from KAIST) released with the paper [Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth](https://arxiv.org/abs/2201.07436) by Doyeon Kim, Woonghyun Ga, Pyungwhan Ahn, Donggyu Joo, Sehwan Chun, Junmo Kim.
 1. **[GPT](https://huggingface.co/docs/transformers/model_doc/openai-gpt)** (from OpenAI) released with the paper [Improving Language Understanding by Generative Pre-Training](https://blog.openai.com/language-unsupervised/) by Alec Radford, Karthik Narasimhan, Tim Salimans and Ilya Sutskever.
 1. **[GPT Neo](https://huggingface.co/docs/transformers/model_doc/gpt_neo)** (from EleutherAI) released in the repository [EleutherAI/gpt-neo](https://github.com/EleutherAI/gpt-neo) by Sid Black, Stella Biderman, Leo Gao, Phil Wang and Connor Leahy.
-1. **[GPT NeoX](https://huggingface.co/docs/transformers/model_doc/gpt_neox)** (from EleutherAI) released with the paper [GPT-NeoX-20B: An Open-Source Autoregressive Language Model](https://arxiv.org/abs/2204.06745) by Sid Black, Stella Biderman, Eric Hallahan, Quentin Anthony, Leo Gao, Laurence Golding, Horace He, Connor Leahy, Kyle McDonell, Jason Phang, Michael Pieler, USVSN Sai Prashanth, Shivanshu Purohit, Laria Reynolds, Jonathan Tow, Ben Wang, Samuel Weinbach 
+1. **[GPT NeoX](https://huggingface.co/docs/transformers/model_doc/gpt_neox)** (from EleutherAI) released with the paper [GPT-NeoX-20B: An Open-Source Autoregressive Language Model](https://arxiv.org/abs/2204.06745) by Sid Black, Stella Biderman, Eric Hallahan, Quentin Anthony, Leo Gao, Laurence Golding, Horace He, Connor Leahy, Kyle McDonell, Jason Phang, Michael Pieler, USVSN Sai Prashanth, Shivanshu Purohit, Laria Reynolds, Jonathan Tow, Ben Wang, Samuel Weinbach
+1. **[GPT NeoX Japanese](https://huggingface.co/docs/transformers/model_doc/gpt_neox_japanese)** (from ABEJA) released by Shinya Otani, Takayoshi Makabe, Anuj Arora, and Kyo Hattori.
 1. **[GPT-2](https://huggingface.co/docs/transformers/model_doc/gpt2)** (from OpenAI) released with the paper [Language Models are Unsupervised Multitask Learners](https://blog.openai.com/better-language-models/) by Alec Radford*, Jeffrey Wu*, Rewon Child, David Luan, Dario Amodei** and Ilya Sutskever**.
 1. **[GPT-J](https://huggingface.co/docs/transformers/model_doc/gptj)** (from EleutherAI) released with the paper [kingoflolz/mesh-transformer-jax](https://github.com/kingoflolz/mesh-transformer-jax/) by Ben Wang and Aran Komatsuzaki.
 1. **[GroupViT](https://huggingface.co/docs/transformers/model_doc/groupvit)** (from UCSD, NVIDIA) released with the paper [GroupViT: Semantic Segmentation Emerges from Text Supervision](https://arxiv.org/abs/2202.11094) by Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, Xiaolong Wang.
@@ -299,7 +305,7 @@ conda install -c huggingface transformers
 1. **[LayoutLM](https://huggingface.co/docs/transformers/model_doc/layoutlm)** (from Microsoft Research Asia) released with the paper [LayoutLM: Pre-training of Text and Layout for Document Image Understanding](https://arxiv.org/abs/1912.13318) by Yiheng Xu, Minghao Li, Lei Cui, Shaohan Huang, Furu Wei, Ming Zhou.
 1. **[LayoutLMv2](https://huggingface.co/docs/transformers/model_doc/layoutlmv2)** (from Microsoft Research Asia) released with the paper [LayoutLMv2: Multi-modal Pre-training for Visually-Rich Document Understanding](https://arxiv.org/abs/2012.14740) by Yang Xu, Yiheng Xu, Tengchao Lv, Lei Cui, Furu Wei, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Wanxiang Che, Min Zhang, Lidong Zhou.
 1. **[LayoutLMv3](https://huggingface.co/docs/transformers/model_doc/layoutlmv3)** (from Microsoft Research Asia) released with the paper [LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking](https://arxiv.org/abs/2204.08387) by Yupan Huang, Tengchao Lv, Lei Cui, Yutong Lu, Furu Wei.
-1. **[LayoutXLM](https://huggingface.co/docs/transformers/model_doc/layoutlmv2)** (from Microsoft Research Asia) released with the paper [LayoutXLM: Multimodal Pre-training for Multilingual Visually-rich Document Understanding](https://arxiv.org/abs/2104.08836) by Yiheng Xu, Tengchao Lv, Lei Cui, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Furu Wei.
+1. **[LayoutXLM](https://huggingface.co/docs/transformers/model_doc/layoutxlm)** (from Microsoft Research Asia) released with the paper [LayoutXLM: Multimodal Pre-training for Multilingual Visually-rich Document Understanding](https://arxiv.org/abs/2104.08836) by Yiheng Xu, Tengchao Lv, Lei Cui, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Furu Wei.
 1. **[LED](https://huggingface.co/docs/transformers/model_doc/led)** (from AllenAI) released with the paper [Longformer: The Long-Document Transformer](https://arxiv.org/abs/2004.05150) by Iz Beltagy, Matthew E. Peters, Arman Cohan.
 1. **[LeViT](https://huggingface.co/docs/transformers/model_doc/levit)** (from Meta AI) released with the paper [LeViT: A Vision Transformer in ConvNet's Clothing for Faster Inference](https://arxiv.org/abs/2104.01136) by Ben Graham, Alaaeldin El-Nouby, Hugo Touvron, Pierre Stock, Armand Joulin, Hervé Jégou, Matthijs Douze.
 1. **[Longformer](https://huggingface.co/docs/transformers/model_doc/longformer)** (from AllenAI) released with the paper [Longformer: The Long-Document Transformer](https://arxiv.org/abs/2004.05150) by Iz Beltagy, Matthew E. Peters, Arman Cohan.
@@ -309,7 +315,8 @@ conda install -c huggingface transformers
 1. **[M-CTC-T](https://huggingface.co/docs/transformers/model_doc/mctct)** (from Facebook) released with the paper [Pseudo-Labeling For Massively Multilingual Speech Recognition](https://arxiv.org/abs/2111.00161) by Loren Lugosch, Tatiana Likhomanenko, Gabriel Synnaeve, and Ronan Collobert.
 1. **[M2M100](https://huggingface.co/docs/transformers/model_doc/m2m_100)** (from Facebook) released with the paper [Beyond English-Centric Multilingual Machine Translation](https://arxiv.org/abs/2010.11125) by Angela Fan, Shruti Bhosale, Holger Schwenk, Zhiyi Ma, Ahmed El-Kishky, Siddharth Goyal, Mandeep Baines, Onur Celebi, Guillaume Wenzek, Vishrav Chaudhary, Naman Goyal, Tom Birch, Vitaliy Liptchinsky, Sergey Edunov, Edouard Grave, Michael Auli, Armand Joulin.
 1. **[MarianMT](https://huggingface.co/docs/transformers/model_doc/marian)** Machine translation models trained using [OPUS](http://opus.nlpl.eu/) data by Jörg Tiedemann. The [Marian Framework](https://marian-nmt.github.io/) is being developed by the Microsoft Translator Team.
-1. **[MaskFormer](https://huggingface.co/docs/transformers/model_doc/maskformer)** (from Meta and UIUC) released with the paper [Per-Pixel Classification is Not All You Need for Semantic Segmentation](https://arxiv.org/abs/2107.06278) by Bowen Cheng, Alexander G. Schwing, Alexander Kirillov 
+1. **[MarkupLM](https://huggingface.co/docs/transformers/model_doc/markuplm)** (from Microsoft Research Asia) released with the paper [MarkupLM: Pre-training of Text and Markup Language for Visually-rich Document Understanding](https://arxiv.org/abs/2110.08518) by Junlong Li, Yiheng Xu, Lei Cui, Furu Wei.
+1. **[MaskFormer](https://huggingface.co/docs/transformers/model_doc/maskformer)** (from Meta and UIUC) released with the paper [Per-Pixel Classification is Not All You Need for Semantic Segmentation](https://arxiv.org/abs/2107.06278) by Bowen Cheng, Alexander G. Schwing, Alexander Kirillov
 1. **[mBART](https://huggingface.co/docs/transformers/model_doc/mbart)** (from Facebook) released with the paper [Multilingual Denoising Pre-training for Neural Machine Translation](https://arxiv.org/abs/2001.08210) by Yinhan Liu, Jiatao Gu, Naman Goyal, Xian Li, Sergey Edunov, Marjan Ghazvininejad, Mike Lewis, Luke Zettlemoyer.
 1. **[mBART-50](https://huggingface.co/docs/transformers/model_doc/mbart)** (from Facebook) released with the paper [Multilingual Translation with Extensible Multilingual Pretraining and Finetuning](https://arxiv.org/abs/2008.00401) by Yuqing Tang, Chau Tran, Xian Li, Peng-Jen Chen, Naman Goyal, Vishrav Chaudhary, Jiatao Gu, Angela Fan.
 1. **[Megatron-BERT](https://huggingface.co/docs/transformers/model_doc/megatron-bert)** (from NVIDIA) released with the paper [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) by Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro.
@@ -326,6 +333,7 @@ conda install -c huggingface transformers
 1. **[OPT](https://huggingface.co/docs/transformers/master/model_doc/opt)** (from Meta AI) released with the paper [OPT: Open Pre-trained Transformer Language Models](https://arxiv.org/abs/2205.01068) by Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen et al.
 1. **[OWL-ViT](https://huggingface.co/docs/transformers/model_doc/owlvit)** (from Google AI) released with the paper [Simple Open-Vocabulary Object Detection with Vision Transformers](https://arxiv.org/abs/2205.06230) by Matthias Minderer, Alexey Gritsenko, Austin Stone, Maxim Neumann, Dirk Weissenborn, Alexey Dosovitskiy, Aravindh Mahendran, Anurag Arnab, Mostafa Dehghani, Zhuoran Shen, Xiao Wang, Xiaohua Zhai, Thomas Kipf, and Neil Houlsby.
 1. **[Pegasus](https://huggingface.co/docs/transformers/model_doc/pegasus)** (from Google) released with the paper [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu.
+1. **[PEGASUS-X](https://huggingface.co/docs/transformers/model_doc/pegasus_x)** (from Google) released with the paper [Investigating Efficiently Extending Transformers for Long Input Summarization](https://arxiv.org/abs/2208.04347) by Jason Phang, Yao Zhao, Peter J. Liu.
 1. **[Perceiver IO](https://huggingface.co/docs/transformers/model_doc/perceiver)** (from Deepmind) released with the paper [Perceiver IO: A General Architecture for Structured Inputs & Outputs](https://arxiv.org/abs/2107.14795) by Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier Hénaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, João Carreira.
 1. **[PhoBERT](https://huggingface.co/docs/transformers/model_doc/phobert)** (from VinAI Research) released with the paper [PhoBERT: Pre-trained language models for Vietnamese](https://www.aclweb.org/anthology/2020.findings-emnlp.92/) by Dat Quoc Nguyen and Anh Tuan Nguyen.
 1. **[PLBart](https://huggingface.co/docs/transformers/model_doc/plbart)** (from UCLA NLP) released with the paper [Unified Pre-training for Program Understanding and Generation](https://arxiv.org/abs/2103.06333) by Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, Kai-Wei Chang.
@@ -335,9 +343,9 @@ conda install -c huggingface transformers
 1. **[RAG](https://huggingface.co/docs/transformers/model_doc/rag)** (from Facebook) released with the paper [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://arxiv.org/abs/2005.11401) by Patrick Lewis, Ethan Perez, Aleksandara Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, Douwe Kiela.
 1. **[REALM](https://huggingface.co/docs/transformers/model_doc/realm.html)** (from Google Research) released with the paper [REALM: Retrieval-Augmented Language Model Pre-Training](https://arxiv.org/abs/2002.08909) by Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat and Ming-Wei Chang.
 1. **[Reformer](https://huggingface.co/docs/transformers/model_doc/reformer)** (from Google Research) released with the paper [Reformer: The Efficient Transformer](https://arxiv.org/abs/2001.04451) by Nikita Kitaev, Łukasz Kaiser, Anselm Levskaya.
-1. **[RegNet](https://huggingface.co/docs/transformers/model_doc/regnet)** (from META Research) released with the paper [Designing Network Design Space](https://arxiv.org/abs/2003.13678) by Ilija Radosavovic, Raj Prateek Kosaraju, Ross Girshick, Kaiming He, Piotr Dollár. 
+1. **[RegNet](https://huggingface.co/docs/transformers/model_doc/regnet)** (from META Research) released with the paper [Designing Network Design Space](https://arxiv.org/abs/2003.13678) by Ilija Radosavovic, Raj Prateek Kosaraju, Ross Girshick, Kaiming He, Piotr Dollár.
 1. **[RemBERT](https://huggingface.co/docs/transformers/model_doc/rembert)** (from Google Research) released with the paper [Rethinking embedding coupling in pre-trained language models](https://arxiv.org/pdf/2010.12821.pdf) by Hyung Won Chung, Thibault Févry, Henry Tsai, M. Johnson, Sebastian Ruder.
-1. **[ResNet](https://huggingface.co/docs/transformers/model_doc/resnet)** (from Microsoft Research) released with the paper [Deep Residual Learning for Image Recognition](https://arxiv.org/abs/1512.03385) by Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun. 
+1. **[ResNet](https://huggingface.co/docs/transformers/model_doc/resnet)** (from Microsoft Research) released with the paper [Deep Residual Learning for Image Recognition](https://arxiv.org/abs/1512.03385) by Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun.
 1. **[RoBERTa](https://huggingface.co/docs/transformers/model_doc/roberta)** (from Facebook), released together with the paper a [Robustly Optimized BERT Pretraining Approach](https://arxiv.org/abs/1907.11692) by Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, Veselin Stoyanov.
 1. **[RoFormer](https://huggingface.co/docs/transformers/model_doc/roformer)** (from ZhuiyiTechnology), released together with the paper a [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://arxiv.org/pdf/2104.09864v1.pdf) by Jianlin Su and Yu Lu and Shengfeng Pan and Bo Wen and Yunfeng Liu.
 1. **[SegFormer](https://huggingface.co/docs/transformers/model_doc/segformer)** (from NVIDIA) released with the paper [SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers](https://arxiv.org/abs/2105.15203) by Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M. Alvarez, Ping Luo.
@@ -348,27 +356,32 @@ conda install -c huggingface transformers
 1. **[Splinter](https://huggingface.co/docs/transformers/model_doc/splinter)** (from Tel Aviv University) released with the paper [Few-Shot Question Answering by Pretraining Span Selection](https://arxiv.org/abs/2101.00438) by Ori Ram, Yuval Kirstain, Jonathan Berant, Amir Globerson, Omer Levy.
 1. **[SqueezeBERT](https://huggingface.co/docs/transformers/model_doc/squeezebert)** (from Berkeley) released with the paper [SqueezeBERT: What can computer vision teach NLP about efficient neural networks?](https://arxiv.org/abs/2006.11316) by Forrest N. Iandola, Albert E. Shaw, Ravi Krishna, and Kurt W. Keutzer.
 1. **[Swin Transformer](https://huggingface.co/docs/transformers/model_doc/swin)** (from Microsoft) released with the paper [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://arxiv.org/abs/2103.14030) by Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, Baining Guo.
-1. **[Swin Transformer V2](https://huggingface.co/docs/transformers/main/model_doc/swinv2)** (from Microsoft) released with the paper [Swin Transformer V2: Scaling Up Capacity and Resolution](https://arxiv.org/abs/2111.09883) by Ze Liu, Han Hu, Yutong Lin, Zhuliang Yao, Zhenda Xie, Yixuan Wei, Jia Ning, Yue Cao, Zheng Zhang, Li Dong, Furu Wei, Baining Guo.
+1. **[Swin Transformer V2](https://huggingface.co/docs/transformers/model_doc/swinv2)** (from Microsoft) released with the paper [Swin Transformer V2: Scaling Up Capacity and Resolution](https://arxiv.org/abs/2111.09883) by Ze Liu, Han Hu, Yutong Lin, Zhuliang Yao, Zhenda Xie, Yixuan Wei, Jia Ning, Yue Cao, Zheng Zhang, Li Dong, Furu Wei, Baining Guo.
 1. **[T5](https://huggingface.co/docs/transformers/model_doc/t5)** (from Google AI) released with the paper [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://arxiv.org/abs/1910.10683) by Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu.
 1. **[T5v1.1](https://huggingface.co/docs/transformers/model_doc/t5v1.1)** (from Google AI) released with the paper [google-research/text-to-text-transfer-transformer](https://github.com/google-research/text-to-text-transfer-transformer/blob/main/released_checkpoints.md#t511) by Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu.
 1. **[TAPAS](https://huggingface.co/docs/transformers/model_doc/tapas)** (from Google AI) released with the paper [TAPAS: Weakly Supervised Table Parsing via Pre-training](https://arxiv.org/abs/2004.02349) by Jonathan Herzig, Paweł Krzysztof Nowak, Thomas Müller, Francesco Piccinno and Julian Martin Eisenschlos.
 1. **[TAPEX](https://huggingface.co/docs/transformers/model_doc/tapex)** (from Microsoft Research) released with the paper [TAPEX: Table Pre-training via Learning a Neural SQL Executor](https://arxiv.org/abs/2107.07653) by Qian Liu, Bei Chen, Jiaqi Guo, Morteza Ziyadi, Zeqi Lin, Weizhu Chen, Jian-Guang Lou.
-1. **[Trajectory Transformer](https://huggingface.co/docs/transformers/model_doc/trajectory_transformers)** (from the University of California at Berkeley) released with the paper [Offline Reinforcement Learning as One Big Sequence Modeling Problem](https://arxiv.org/abs/2106.02039) by Michael Janner, Qiyang Li, Sergey Levine 
+1. **[Time Series Transformer](https://huggingface.co/docs/transformers/model_doc/time_series_transformer)** (from HuggingFace).
+1. **[Trajectory Transformer](https://huggingface.co/docs/transformers/model_doc/trajectory_transformers)** (from the University of California at Berkeley) released with the paper [Offline Reinforcement Learning as One Big Sequence Modeling Problem](https://arxiv.org/abs/2106.02039) by Michael Janner, Qiyang Li, Sergey Levine
 1. **[Transformer-XL](https://huggingface.co/docs/transformers/model_doc/transfo-xl)** (from Google/CMU) released with the paper [Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context](https://arxiv.org/abs/1901.02860) by Zihang Dai*, Zhilin Yang*, Yiming Yang, Jaime Carbonell, Quoc V. Le, Ruslan Salakhutdinov.
 1. **[TrOCR](https://huggingface.co/docs/transformers/model_doc/trocr)** (from Microsoft) released with the paper [TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models](https://arxiv.org/abs/2109.10282) by Minghao Li, Tengchao Lv, Lei Cui, Yijuan Lu, Dinei Florencio, Cha Zhang, Zhoujun Li, Furu Wei.
-1. **[UL2](https://huggingface.co/docs/transformers/model_doc/ul2)** (from Google Research) released with the paper [Unifying Language Learning Paradigms](https://arxiv.org/abs/2205.05131v1) by Yi Tay, Mostafa Dehghani, Vinh Q. Tran, Xavier Garcia, Dara Bahri, Tal Schuster, Huaixiu Steven Zheng, Neil Houlsby, Donald Metzler 
+1. **[UL2](https://huggingface.co/docs/transformers/model_doc/ul2)** (from Google Research) released with the paper [Unifying Language Learning Paradigms](https://arxiv.org/abs/2205.05131v1) by Yi Tay, Mostafa Dehghani, Vinh Q. Tran, Xavier Garcia, Dara Bahri, Tal Schuster, Huaixiu Steven Zheng, Neil Houlsby, Donald Metzler
 1. **[UniSpeech](https://huggingface.co/docs/transformers/model_doc/unispeech)** (from Microsoft Research) released with the paper [UniSpeech: Unified Speech Representation Learning with Labeled and Unlabeled Data](https://arxiv.org/abs/2101.07597) by Chengyi Wang, Yu Wu, Yao Qian, Kenichi Kumatani, Shujie Liu, Furu Wei, Michael Zeng, Xuedong Huang.
 1. **[UniSpeechSat](https://huggingface.co/docs/transformers/model_doc/unispeech-sat)** (from Microsoft Research) released with the paper [UNISPEECH-SAT: UNIVERSAL SPEECH REPRESENTATION LEARNING WITH SPEAKER AWARE PRE-TRAINING](https://arxiv.org/abs/2110.05752) by Sanyuan Chen, Yu Wu, Chengyi Wang, Zhengyang Chen, Zhuo Chen, Shujie Liu, Jian Wu, Yao Qian, Furu Wei, Jinyu Li, Xiangzhan Yu.
 1. **[VAN](https://huggingface.co/docs/transformers/model_doc/van)** (from Tsinghua University and Nankai University) released with the paper [Visual Attention Network](https://arxiv.org/pdf/2202.09741.pdf) by Meng-Hao Guo, Cheng-Ze Lu, Zheng-Ning Liu, Ming-Ming Cheng, Shi-Min Hu.
+1. **[VideoMAE](https://huggingface.co/docs/transformers/model_doc/videomae)** (from Multimedia Computing Group, Nanjing University) released with the paper [VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training](https://arxiv.org/abs/2203.12602) by Zhan Tong, Yibing Song, Jue Wang, Limin Wang.
 1. **[ViLT](https://huggingface.co/docs/transformers/model_doc/vilt)** (from NAVER AI Lab/Kakao Enterprise/Kakao Brain) released with the paper [ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision](https://arxiv.org/abs/2102.03334) by Wonjae Kim, Bokyung Son, Ildoo Kim.
-1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby. 
+1. **[Vision Transformer (ViT)](https://huggingface.co/docs/transformers/model_doc/vit)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.
 1. **[VisualBERT](https://huggingface.co/docs/transformers/model_doc/visual_bert)** (from UCLA NLP) released with the paper [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) by Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang.
 1. **[ViTMAE](https://huggingface.co/docs/transformers/model_doc/vit_mae)** (from Meta AI) released with the paper [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) by Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick.
+1. **[ViTMSN](https://huggingface.co/docs/transformers/model_doc/vit_msn)** (from Meta AI) released with the paper [Masked Siamese Networks for Label-Efficient Learning](https://arxiv.org/abs/2204.07141) by Mahmoud Assran, Mathilde Caron, Ishan Misra, Piotr Bojanowski, Florian Bordes, Pascal Vincent, Armand Joulin, Michael Rabbat, Nicolas Ballas.
 1. **[Wav2Vec2](https://huggingface.co/docs/transformers/model_doc/wav2vec2)** (from Facebook AI) released with the paper [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) by Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli.
 1. **[Wav2Vec2-Conformer](https://huggingface.co/docs/transformers/model_doc/wav2vec2-conformer)** (from Facebook AI) released with the paper [FAIRSEQ S2T: Fast Speech-to-Text Modeling with FAIRSEQ](https://arxiv.org/abs/2010.05171) by Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Sravya Popuri, Dmytro Okhonko, Juan Pino.
 1. **[Wav2Vec2Phoneme](https://huggingface.co/docs/transformers/model_doc/wav2vec2_phoneme)** (from Facebook AI) released with the paper [Simple and Effective Zero-shot Cross-lingual Phoneme Recognition](https://arxiv.org/abs/2109.11680) by Qiantong Xu, Alexei Baevski, Michael Auli.
 1. **[WavLM](https://huggingface.co/docs/transformers/model_doc/wavlm)** (from Microsoft Research) released with the paper [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://arxiv.org/abs/2110.13900) by Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng, Furu Wei.
-1. **[XGLM](https://huggingface.co/docs/transformers/model_doc/xglm)** (From Facebook AI) released with the paper [Few-shot Learning with Multilingual Language Models](https://arxiv.org/abs/2112.10668) by Xi Victoria Lin, Todor Mihaylov, Mikel Artetxe, Tianlu Wang, Shuohui Chen, Daniel Simig, Myle Ott, Naman Goyal, Shruti Bhosale, Jingfei Du, Ramakanth Pasunuru, Sam Shleifer, Punit Singh Koura, Vishrav Chaudhary, Brian O'Horo, Jeff Wang, Luke Zettlemoyer, Zornitsa Kozareva, Mona Diab, Veselin Stoyanov, Xian Li. 
+1. **[Whisper](https://huggingface.co/docs/transformers/model_doc/whisper)** (from OpenAI) released with the paper [Robust Speech Recognition via Large-Scale Weak Supervision](https://cdn.openai.com/papers/whisper.pdf) by Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, Ilya Sutskever.
+1. **[X-CLIP](https://huggingface.co/docs/transformers/model_doc/xclip)** (from Microsoft Research) released with the paper [Expanding Language-Image Pretrained Models for General Video Recognition](https://arxiv.org/abs/2208.02816) by Bolin Ni, Houwen Peng, Minghao Chen, Songyang Zhang, Gaofeng Meng, Jianlong Fu, Shiming Xiang, Haibin Ling.
+1. **[XGLM](https://huggingface.co/docs/transformers/model_doc/xglm)** (From Facebook AI) released with the paper [Few-shot Learning with Multilingual Language Models](https://arxiv.org/abs/2112.10668) by Xi Victoria Lin, Todor Mihaylov, Mikel Artetxe, Tianlu Wang, Shuohui Chen, Daniel Simig, Myle Ott, Naman Goyal, Shruti Bhosale, Jingfei Du, Ramakanth Pasunuru, Sam Shleifer, Punit Singh Koura, Vishrav Chaudhary, Brian O'Horo, Jeff Wang, Luke Zettlemoyer, Zornitsa Kozareva, Mona Diab, Veselin Stoyanov, Xian Li.
 1. **[XLM](https://huggingface.co/docs/transformers/model_doc/xlm)** (from Facebook) released together with the paper [Cross-lingual Language Model Pretraining](https://arxiv.org/abs/1901.07291) by Guillaume Lample and Alexis Conneau.
 1. **[XLM-ProphetNet](https://huggingface.co/docs/transformers/model_doc/xlm-prophetnet)** (from Microsoft Research) released with the paper [ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training](https://arxiv.org/abs/2001.04063) by Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou.
 1. **[XLM-RoBERTa](https://huggingface.co/docs/transformers/model_doc/xlm-roberta)** (from Facebook AI), released together with the paper [Unsupervised Cross-lingual Representation Learning at Scale](https://arxiv.org/abs/1911.02116) by Alexis Conneau*, Kartikay Khandelwal*, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer and Veselin Stoyanov.

conftest.py

@@ -32,7 +32,6 @@ warnings.simplefilter(action="ignore", category=FutureWarning)
 
 
 def pytest_configure(config):
-    config.addinivalue_line("markers", "is_pipeline_test: mark test to run only when pipeline are tested")
     config.addinivalue_line(
         "markers", "is_pt_tf_cross_test: mark test to run only when PT and TF interactions are tested"
     )

docker/transformers-all-latest-gpu/Dockerfile

@@ -9,7 +9,7 @@ SHELL ["sh", "-lc"]
 # The following `ARG` are mainly used to specify the versions explicitly & directly in this docker file, and not meant
 # to be used as arguments for docker build (so far).
 
-ARG PYTORCH='1.12.0'
+ARG PYTORCH='1.12.1'
 # (not always a valid torch version)
 ARG INTEL_TORCH_EXT='1.11.0'
 # Example: `cu102`, `cu113`, etc.
@@ -45,6 +45,11 @@ RUN python3 -m pip install -U "itsdangerous<2.1.0"
 
 RUN python3 -m pip install --no-cache-dir git+https://github.com/huggingface/accelerate@main#egg=accelerate
 
+# Add bitsandbytes for mixed int8 testing
+RUN python3 -m pip install --no-cache-dir bitsandbytes
+
+RUN python3 -m pip install --no-cache-dir decord
+
 # When installing in editable mode, `transformers` is not recognized as a package.
 # this line must be added in order for python to be aware of transformers.
 RUN cd transformers && python3 setup.py develop

docker/transformers-cpu/Dockerfile

@@ -23,4 +23,4 @@ COPY . transformers/
 RUN cd transformers/ && \
     python3 -m pip install --no-cache-dir .
 
-CMD ["/bin/bash"]
+CMD ["/bin/bash"]

docker/transformers-pytorch-deepspeed-latest-gpu/Dockerfile

@@ -3,7 +3,7 @@ LABEL maintainer="Hugging Face"
 
 ARG DEBIAN_FRONTEND=noninteractive
 
-ARG PYTORCH='1.12.0'
+ARG PYTORCH='1.12.1'
 # Example: `cu102`, `cu113`, etc.
 ARG CUDA='cu113'
 
@@ -26,7 +26,7 @@ RUN python3 -m pip uninstall -y deepspeed
 # This has to be run (again) inside the GPU VMs running the tests.
 # The installation works here, but some tests fail, if we don't pre-build deepspeed again in the VMs running the tests.
 # TODO: Find out why test fail.
-RUN DS_BUILD_CPU_ADAM=1 DS_BUILD_AIO=1 DS_BUILD_UTILS=1 python3 -m pip install deepspeed --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check 2>&1
+RUN DS_BUILD_CPU_ADAM=1 DS_BUILD_FUSED_ADAM=1 DS_BUILD_AIO=1 DS_BUILD_UTILS=1 python3 -m pip install deepspeed --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check 2>&1
 
 # When installing in editable mode, `transformers` is not recognized as a package.
 # this line must be added in order for python to be aware of transformers.

docker/transformers-pytorch-deepspeed-nightly-gpu/Dockerfile

@@ -25,7 +25,25 @@ RUN python3 -m pip uninstall -y deepspeed
 # This has to be run inside the GPU VMs running the tests. (So far, it fails here due to GPU checks during compilation.)
 # Issue: https://github.com/microsoft/DeepSpeed/issues/2010
 # RUN git clone https://github.com/microsoft/DeepSpeed && cd DeepSpeed && rm -rf build && \
-#    DS_BUILD_CPU_ADAM=1 DS_BUILD_AIO=1 DS_BUILD_UTILS=1 python3 -m pip install . --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check 2>&1
+#    DS_BUILD_CPU_ADAM=1 DS_BUILD_FUSED_ADAM=1 DS_BUILD_AIO=1 DS_BUILD_UTILS=1 python3 -m pip install . --global-option="build_ext" --global-option="-j8" --no-cache -v --disable-pip-version-check 2>&1
+
+# For `torchdynamo` tests
+# (see https://github.com/huggingface/transformers/pull/17765)
+RUN git clone https://github.com/pytorch/functorch
+RUN python3 -m pip install --no-cache-dir ./functorch[aot]
+RUN cd functorch && python3 setup.py develop
+
+RUN git clone https://github.com/pytorch/torchdynamo
+RUN python3 -m pip install -r ./torchdynamo/requirements.txt
+RUN cd torchdynamo && python3 setup.py develop
+
+# install TensorRT
+RUN python3 -m pip install --no-cache-dir -U nvidia-pyindex
+RUN python3 -m pip install --no-cache-dir -U nvidia-tensorrt==8.2.4.2
+
+# install torch_tensorrt (fx path)
+RUN git clone https://github.com/pytorch/TensorRT.git
+RUN cd TensorRT/py && python3 setup.py install --fx-only
 
 # When installing in editable mode, `transformers` is not recognized as a package.
 # this line must be added in order for python to be aware of transformers.

docker/transformers-pytorch-gpu/Dockerfile

@@ -12,7 +12,7 @@ RUN git clone https://github.com/huggingface/transformers && cd transformers &&
 RUN python3 -m pip install --no-cache-dir -e ./transformers[dev-torch,testing]
 
 # If set to nothing, will install the latest version
-ARG PYTORCH='1.12.0'
+ARG PYTORCH='1.12.1'
 ARG TORCH_VISION=''
 ARG TORCH_AUDIO=''
 

docker/transformers-tensorflow-cpu/Dockerfile

@@ -22,4 +22,4 @@ COPY . transformers/
 RUN cd transformers/ && \
     python3 -m pip install --no-cache-dir .
 
-CMD ["/bin/bash"]
+CMD ["/bin/bash"]

docs/README.md

@@ -16,7 +16,7 @@ limitations under the License.
 
 # Generating the documentation
 
-To generate the documentation, you first have to build it. Several packages are necessary to build the doc,
+To generate the documentation, you first have to build it. Several packages are necessary to build the doc, 
 you can install them with the following command, at the root of the code repository:
 
 ```bash
@@ -33,7 +33,7 @@ pip install git+https://github.com/huggingface/doc-builder
 **NOTE**
 
 You only need to generate the documentation to inspect it locally (if you're planning changes and want to
-check how they look like before committing for instance). You don't have to commit the built documentation.
+check how they look before committing for instance). You don't have to commit the built documentation.
 
 ---
 
@@ -43,7 +43,7 @@ Once you have setup the `doc-builder` and additional packages, you can generate
 typing the following command:
 
 ```bash
-doc-builder build transformers docs/source/ --build_dir ~/tmp/test-build
+doc-builder build transformers docs/source/en/ --build_dir ~/tmp/test-build
 ```
 
 You can adapt the `--build_dir` to set any temporary folder that you prefer. This command will create it and generate
@@ -88,7 +88,7 @@ the filename without the extension in the [`_toctree.yml`](https://github.com/hu
 
 ## Renaming section headers and moving sections
 
-It helps to keep the old links working when renaming section header and/or moving sections from one document to another. This is because the old links are likely to be used in Issues, Forums and Social media and it'd be make for a much more superior user experience if users reading those months later could still easily navigate to the originally intended information.
+It helps to keep the old links working when renaming the section header and/or moving sections from one document to another. This is because the old links are likely to be used in Issues, Forums, and Social media and it'd make for a much more superior user experience if users reading those months later could still easily navigate to the originally intended information.
 
 Therefore we simply keep a little map of moved sections at the end of the document where the original section was. The key is to preserve the original anchor.
 
@@ -99,7 +99,7 @@ Sections that were moved:
 
 [ <a href="#section-b">Section A</a><a id="section-a"></a> ]
 ```
-and of course if you moved it to another file, then:
+and of course, if you moved it to another file, then:
 
 ```
 Sections that were moved:
@@ -109,7 +109,7 @@ Sections that were moved:
 
 Use the relative style to link to the new file so that the versioned docs continue to work.
 
-For an example of a rich moved sections set please see the very end of [the Trainer doc](https://github.com/huggingface/transformers/blob/main/docs/source/main_classes/trainer.mdx).
+For an example of a rich moved section set please see the very end of [the Trainer doc](https://github.com/huggingface/transformers/blob/main/docs/source/en/main_classes/trainer.mdx).
 
 
 ## Writing Documentation - Specification
@@ -126,7 +126,7 @@ Adding a new tutorial or section is done in two steps:
 - Link that file in `./source/_toctree.yml` on the correct toc-tree.
 
 Make sure to put your new file under the proper section. It's unlikely to go in the first section (*Get Started*), so
-depending on the intended targets (beginners, more advanced users or researchers) it should go in section two, three or
+depending on the intended targets (beginners, more advanced users, or researchers) it should go in sections two, three, or
 four.
 
 ### Translating
@@ -177,8 +177,8 @@ not to be displayed in the documentation, you can do so by specifying which meth
     - save_vocabulary
 ```
 
-If you just want to add a method that is not documented (for instance magic method like `__call__` are not documented
-byt default) you can put the list of methods to add in a list that contains `all`:
+If you just want to add a method that is not documented (for instance magic methods like `__call__` are not documented
+by default) you can put the list of methods to add in a list that contains `all`:
 
 ```
 ## XXXTokenizer
@@ -191,9 +191,9 @@ byt default) you can put the list of methods to add in a list that contains `all
 ### Writing source documentation
 
 Values that should be put in `code` should either be surrounded by backticks: \`like so\`. Note that argument names
-and objects like True, None or any strings should usually be put in `code`.
+and objects like True, None, or any strings should usually be put in `code`.
 
-When mentioning a class, function or method, it is recommended to use our syntax for internal links so that our tool
+When mentioning a class, function, or method, it is recommended to use our syntax for internal links so that our tool
 adds a link to its documentation with this syntax: \[\`XXXClass\`\] or \[\`function\`\]. This requires the class or 
 function to be in the main package.
 
@@ -207,7 +207,7 @@ The same works for methods so you can either use \[\`XXXClass.method\`\] or \[~\
 #### Defining arguments in a method
 
 Arguments should be defined with the `Args:` (or `Arguments:` or `Parameters:`) prefix, followed by a line return and
-an indentation. The argument should be followed by its type, with its shape if it is a tensor, a colon and its
+an indentation. The argument should be followed by its type, with its shape if it is a tensor, a colon, and its
 description:
 
 ```
@@ -216,7 +216,7 @@ description:
 ```
 
 If the description is too long to fit in one line, another indentation is necessary before writing the description
-after th argument.
+after the argument.
 
 Here's an example showcasing everything so far:
 
@@ -266,7 +266,7 @@ Multi-line code blocks can be useful for displaying examples. They are done betw
 ````
 
 We follow the [doctest](https://docs.python.org/3/library/doctest.html) syntax for the examples to automatically test
-the results stay consistent with the library.
+the results to stay consistent with the library.
 
 #### Writing a return block
 
@@ -274,27 +274,27 @@ The return block should be introduced with the `Returns:` prefix, followed by a
 The first line should be the type of the return, followed by a line return. No need to indent further for the elements
 building the return.
 
-Here's an example for a single value return:
+Here's an example of a single value return:
 
 ```
     Returns:
         `List[int]`: A list of integers in the range [0, 1] --- 1 for a special token, 0 for a sequence token.
 ```
 
-Here's an example for tuple return, comprising several objects:
+Here's an example of a tuple return, comprising several objects:
 
 ```
     Returns:
         `tuple(torch.FloatTensor)` comprising various elements depending on the configuration ([`BertConfig`]) and inputs:
         - ** loss** (*optional*, returned when `masked_lm_labels` is provided) `torch.FloatTensor` of shape `(1,)` --
-          Total loss as the sum of the masked language modeling loss and the next sequence prediction (classification) loss.
+          Total loss is the sum of the masked language modeling loss and the next sequence prediction (classification) loss.
         - **prediction_scores** (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.vocab_size)`) --
           Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
 ```
 
 #### Adding an image
 
-Due to the rapidly growing repository, it is important to make sure that no files that would significantly weigh down the repository are added. This includes images, videos and other non-text files. We prefer to leverage a hf.co hosted `dataset` like
+Due to the rapidly growing repository, it is important to make sure that no files that would significantly weigh down the repository are added. This includes images, videos, and other non-text files. We prefer to leverage a hf.co hosted `dataset` like
 the ones hosted on [`hf-internal-testing`](https://huggingface.co/hf-internal-testing) in which to place these files and reference
 them by URL. We recommend putting them in the following dataset: [huggingface/documentation-images](https://huggingface.co/datasets/huggingface/documentation-images).
 If an external contribution, feel free to add the images to your PR and ask a Hugging Face member to migrate your images
@@ -312,13 +312,13 @@ easily.
 
 # Testing documentation examples
 
-Good documentation oftens comes with an example of how a specific function or class should be used. 
+Good documentation often comes with an example of how a specific function or class should be used. 
 Each model class should contain at least one example showcasing
 how to use this model class in inference. *E.g.* the class [Wav2Vec2ForCTC](https://huggingface.co/docs/transformers/model_doc/wav2vec2#transformers.Wav2Vec2ForCTC) 
 includes an example of how to transcribe speech to text in the 
 [docstring of its forward function](https://huggingface.co/docs/transformers/model_doc/wav2vec2#transformers.Wav2Vec2ForCTC.forward).
 
-## Writing documenation examples
+## Writing documentation examples
 
 The syntax for Example docstrings can look as follows:
 
@@ -364,7 +364,7 @@ We use pytests' [doctest integration](https://docs.pytest.org/doctest.html) to v
 For Transformers, the doctests are run on a daily basis via GitHub Actions as can be 
 seen [here](https://github.com/huggingface/transformers/actions/workflows/doctests.yml).
 
-To include your example in the daily doctests, you need add the filename that
+To include your example in the daily doctests, you need to add the filename that
 contains the example docstring to the [documentation_tests.txt](../utils/documentation_tests.txt).
 
 ### For Python files
@@ -426,6 +426,6 @@ Here are a few tips to help you debug the doctests and make them pass:
 
 - The outputs of the code need to match the expected output **exactly**, so make sure you have the same outputs. In particular doctest will see a difference between single quotes and double quotes, or a missing parenthesis. The only exceptions to that rule are:
   * whitespace: one give whitespace (space, tabulation, new line) is equivalent to any number of whitespace, so you can add new lines where there are spaces to make your output more readable.
-  * numerical values: you should never put more than 4 or 5 digits to expected results as different setups or library versions might get you slightly different results. `doctest` is configure to ignore any difference lower than the precision to which you wrote (so 1e-4 if you write 4 digits).
+  * numerical values: you should never put more than 4 or 5 digits to expected results as different setups or library versions might get you slightly different results. `doctest` is configured to ignore any difference lower than the precision to which you wrote (so 1e-4 if you write 4 digits).
 - Don't leave a block of code that is very long to execute. If you can't make it fast, you can either not use the doctest syntax on it (so that it's ignored), or if you want to use the doctest syntax to show the results, you can add a comment `# doctest: +SKIP` at the end of the lines of code too long to execute
 - Each line of code that produces a result needs to have that result written below. You can ignore an output if you don't want to show it in your code example by adding a comment ` # doctest: +IGNORE_RESULT` at the end of the line of code producing it.

docs/TRANSLATING.md

@@ -54,5 +54,4 @@ The fields you should add are `local` (with the name of the file containing the
 
 Once you have translated the `_toctree.yml` file, you can start translating the [MDX](https://mdxjs.com/) files associated with your docs chapter.
 
-> 🙋 If you'd like others to help you with the translation, you can either [open an issue](https://github.com/huggingface/transformers/issues) or tag @[espejelomar](https://twitter.com/espejelomar)
- on Twitter to gain some visibility.
+> 🙋 If you'd like others to help you with the translation, you should [open an issue](https://github.com/huggingface/transformers/issues) and tag @sgugger.

docs/source/de/_config.py

@@ -0,0 +1,14 @@
+# docstyle-ignore
+INSTALL_CONTENT = """
+# Transformers installation
+! pip install transformers datasets
+# To install from source instead of the last release, comment the command above and uncomment the following one.
+# ! pip install git+https://github.com/huggingface/transformers.git
+"""
+
+notebook_first_cells = [{"type": "code", "content": INSTALL_CONTENT}]
+black_avoid_patterns = {
+    "{processor_class}": "FakeProcessorClass",
+    "{model_class}": "FakeModelClass",
+    "{object_class}": "FakeObjectClass",
+}

docs/source/de/_toctree.yml

@@ -0,0 +1,22 @@
+- sections:
+  - local: index
+    title: 🤗 Transformers
+  - local: quicktour
+    title: Schnellstart
+  - local: installation
+    title: Installation
+  title: Erste Schritte
+- sections:
+  - local: pipeline_tutorial
+    title: Pipelines für Inferenzen
+  - local: autoclass_tutorial
+    title: Laden von vortrainierten Instanzen mit einer AutoClass
+  - local: preprocessing
+    title: Vorverarbeiten
+  - local: training
+    title: Optimierung eines vortrainierten Modells
+  - local: accelerate
+    title: Verteiltes Training mit 🤗 Accelerate
+  - local: model_sharing
+    title: Ein Modell teilen
+  title: Tutorials

docs/source/de/accelerate.mdx

@@ -0,0 +1,132 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# Verteiltes Training mit 🤗 Accelerate
+
+Da die Modelle immer größer werden, hat sich die Parallelität als Strategie zum Trainieren größerer Modelle auf begrenzter Hardware und zur Beschleunigung der Trainingsgeschwindigkeit um mehrere Größenordnungen erwiesen. Bei Hugging Face haben wir die Bibliothek [🤗 Accelerate](https://huggingface.co/docs/accelerate) entwickelt, um Nutzern zu helfen, ein 🤗 Transformers-Modell auf jeder Art von verteiltem Setup zu trainieren, egal ob es sich um mehrere GPUs auf einer Maschine oder mehrere GPUs auf mehreren Maschinen handelt. In diesem Tutorial lernen Sie, wie Sie Ihre native PyTorch-Trainingsschleife anpassen, um das Training in einer verteilten Umgebung zu ermöglichen.
+
+## Einrichtung
+
+Beginnen Sie mit der Installation von 🤗 Accelerate:
+
+```bash
+pip install accelerate
+```
+
+Dann importieren und erstellen Sie ein [`~accelerate.Accelerator`]-Objekt. Der [`~accelerate.Accelerator`] wird automatisch Ihre Art der verteilten Einrichtung erkennen und alle notwendigen Komponenten für das Training initialisieren. Sie müssen Ihr Modell nicht explizit auf einem Gerät platzieren.
+
+```py
+>>> from accelerate import Accelerator
+
+>>> accelerator = Accelerator()
+```
+
+## Vorbereiten auf die Beschleunigung
+
+Der nächste Schritt ist die Übergabe aller relevanten Trainingsobjekte an die Methode [`~accelerate.Accelerator.prepare`]. Dazu gehören Ihre Trainings- und Evaluierungs-DataLoader, ein Modell und ein Optimierer:
+
+```py
+>>> train_dataloader, eval_dataloader, model, optimizer = accelerator.prepare(
+...     train_dataloader, eval_dataloader, model, optimizer
+... )
+```
+
+## Rückwärts
+
+Die letzte Ergänzung besteht darin, das typische `loss.backward()` in der Trainingsschleife durch die 🤗 Accelerate-Methode [`~accelerate.Accelerator.backward`] zu ersetzen:
+
+```py
+>>> for epoch in range(num_epochs):
+...     for batch in train_dataloader:
+...         outputs = model(**batch)
+...         loss = outputs.loss
+...         accelerator.backward(loss)
+
+...         optimizer.step()
+...         lr_scheduler.step()
+...         optimizer.zero_grad()
+...         progress_bar.update(1)
+```
+
+Wie Sie im folgenden Code sehen können, müssen Sie nur vier zusätzliche Codezeilen zu Ihrer Trainingsschleife hinzufügen, um verteiltes Training zu ermöglichen!
+
+```diff
++ from accelerate import Accelerator
+  from transformers import AdamW, AutoModelForSequenceClassification, get_scheduler
+
++ accelerator = Accelerator()
+
+  model = AutoModelForSequenceClassification.from_pretrained(checkpoint, num_labels=2)
+  optimizer = AdamW(model.parameters(), lr=3e-5)
+
+- device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
+- model.to(device)
+
++ train_dataloader, eval_dataloader, model, optimizer = accelerator.prepare(
++     train_dataloader, eval_dataloader, model, optimizer
++ )
+
+  num_epochs = 3
+  num_training_steps = num_epochs * len(train_dataloader)
+  lr_scheduler = get_scheduler(
+      "linear",
+      optimizer=optimizer,
+      num_warmup_steps=0,
+      num_training_steps=num_training_steps
+  )
+
+  progress_bar = tqdm(range(num_training_steps))
+
+  model.train()
+  for epoch in range(num_epochs):
+      for batch in train_dataloader:
+-         batch = {k: v.to(device) for k, v in batch.items()}
+          outputs = model(**batch)
+          loss = outputs.loss
+-         loss.backward()
++         accelerator.backward(loss)
+
+          optimizer.step()
+          lr_scheduler.step()
+          optimizer.zero_grad()
+          progress_bar.update(1)
+```
+
+## Trainieren
+
+Sobald Sie die entsprechenden Codezeilen hinzugefügt haben, starten Sie Ihr Training in einem Skript oder einem Notebook wie Colaboratory.
+
+### Trainieren mit einem Skript
+
+Wenn Sie Ihr Training mit einem Skript durchführen, führen Sie den folgenden Befehl aus, um eine Konfigurationsdatei zu erstellen und zu speichern:
+
+```bash
+accelerate config
+```
+
+Dann starten Sie Ihr Training mit:
+
+```bash
+accelerate launch train.py
+```
+
+### Trainieren mit einem Notebook
+
+🤗 Accelerate kann auch in einem Notebook laufen, wenn Sie planen, die TPUs von Colaboratory zu verwenden. Verpacken Sie den gesamten Code, der für das Training verantwortlich ist, in eine Funktion und übergeben Sie diese an [`~accelerate.notebook_launcher`]:
+
+```py
+>>> from accelerate import notebook_launcher
+
+>>> notebook_launcher(training_function)
+```
+
+Weitere Informationen über 🤗 Accelerate und seine umfangreichen Funktionen finden Sie in der [Dokumentation](https://huggingface.co/docs/accelerate).

docs/source/de/autoclass_tutorial.mdx

@@ -0,0 +1,127 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# Vortrainierte Instanzen mit einer AutoClass laden
+
+Bei so vielen verschiedenen Transformator-Architekturen kann es eine Herausforderung sein, eine für Ihren Checkpoint zu erstellen. Als Teil der 🤗 Transformers Kernphilosophie, die Bibliothek leicht, einfach und flexibel nutzbar zu machen, leitet eine `AutoClass` automatisch die richtige Architektur aus einem gegebenen Checkpoint ab und lädt sie. Mit der Methode `from_pretrained()` kann man schnell ein vortrainiertes Modell für eine beliebige Architektur laden, so dass man keine Zeit und Ressourcen aufwenden muss, um ein Modell von Grund auf zu trainieren. Die Erstellung dieser Art von Checkpoint-agnostischem Code bedeutet, dass Ihr Code, wenn er für einen Checkpoint funktioniert, auch mit einem anderen Checkpoint funktionieren wird - solange er für eine ähnliche Aufgabe trainiert wurde - selbst wenn die Architektur unterschiedlich ist.
+
+<Tip>
+
+Denken Sie daran, dass sich die Architektur auf das Skelett des Modells bezieht und die Checkpoints die Gewichte für eine bestimmte Architektur sind. Zum Beispiel ist [BERT](https://huggingface.co/bert-base-uncased) eine Architektur, während `bert-base-uncased` ein Checkpoint ist. Modell ist ein allgemeiner Begriff, der entweder Architektur oder Prüfpunkt bedeuten kann.
+
+</Tip>
+
+In dieser Anleitung lernen Sie, wie man:
+
+* Einen vortrainierten Tokenizer lädt.
+* Einen vortrainierten Merkmalsextraktor lädt.
+* Einen vortrainierten Prozessor lädt.
+* Ein vortrainiertes Modell lädt.
+
+## AutoTokenizer
+
+Nahezu jede NLP-Aufgabe beginnt mit einem Tokenizer. Ein Tokenizer wandelt Ihre Eingabe in ein Format um, das vom Modell verarbeitet werden kann.
+
+Laden Sie einen Tokenizer mit [`AutoTokenizer.from_pretrained`]:
+
+```py
+>>> from transformers import AutoTokenizer
+
+>>> tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
+```
+
+Dann tokenisieren Sie Ihre Eingabe wie unten gezeigt:
+
+```py
+>>> sequence = "In a hole in the ground there lived a hobbit."
+>>> print(tokenizer(sequence))
+{'input_ids': [101, 1999, 1037, 4920, 1999, 1996, 2598, 2045, 2973, 1037, 7570, 10322, 4183, 1012, 102], 
+ 'token_type_ids': [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], 
+ 'attention_mask': [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]}
+```
+
+## AutoFeatureExtractor
+
+Für Audio- und Bildverarbeitungsaufgaben verarbeitet ein Merkmalsextraktor das Audiosignal oder Bild in das richtige Eingabeformat.
+
+Laden Sie einen Merkmalsextraktor mit [`AutoFeatureExtractor.from_pretrained`]:
+
+```py
+>>> from transformers import AutoFeatureExtractor
+
+>>> feature_extractor = AutoFeatureExtractor.from_pretrained(
+...     "ehcalabres/wav2vec2-lg-xlsr-en-speech-emotion-recognition"
+... )
+```
+
+## AutoProcessor
+
+Multimodale Aufgaben erfordern einen Prozessor, der zwei Arten von Vorverarbeitungswerkzeugen kombiniert. Das Modell [LayoutLMV2](model_doc/layoutlmv2) beispielsweise benötigt einen Feature-Extraktor für Bilder und einen Tokenizer für Text; ein Prozessor kombiniert beide.
+
+Laden Sie einen Prozessor mit [`AutoProcessor.from_pretrained`]:
+
+```py
+>>> from transformers import AutoProcessor
+
+>>> processor = AutoProcessor.from_pretrained("microsoft/layoutlmv2-base-uncased")
+```
+
+## AutoModel
+
+<frameworkcontent>
+<pt>
+Mit den `AutoModelFor`-Klassen können Sie schließlich ein vortrainiertes Modell für eine bestimmte Aufgabe laden (siehe [hier](model_doc/auto) für eine vollständige Liste der verfügbaren Aufgaben). Laden Sie zum Beispiel ein Modell für die Sequenzklassifikation mit [`AutoModelForSequenceClassification.from_pretrained`]:
+
+```py
+>>> from transformers import AutoModelForSequenceClassification
+
+>>> model = AutoModelForSequenceClassification.from_pretrained("distilbert-base-uncased")
+```
+
+Sie können denselben Prüfpunkt problemlos wiederverwenden, um eine Architektur für eine andere Aufgabe zu laden:
+
+```py
+>>> from transformers import AutoModelForTokenClassification
+
+>>> model = AutoModelForTokenClassification.from_pretrained("distilbert-base-uncased")
+```
+
+<Tip warning={true}>
+
+Für PyTorch-Modelle verwendet die Methode `from_pretrained()` `torch.load()`, die intern `pickle` verwendet und als unsicher bekannt ist. Generell sollte man niemals ein Modell laden, das aus einer nicht vertrauenswürdigen Quelle stammen könnte, oder das manipuliert worden sein könnte. Dieses Sicherheitsrisiko wird für öffentliche Modelle, die auf dem Hugging Face Hub gehostet werden, teilweise gemildert, da diese bei jeder Übertragung [auf Malware](https://huggingface.co/docs/hub/security-malware) gescannt werden. Siehe die [Hub-Dokumentation](https://huggingface.co/docs/hub/security) für Best Practices wie [signierte Commit-Verifizierung](https://huggingface.co/docs/hub/security-gpg#signing-commits-with-gpg) mit GPG.
+
+TensorFlow- und Flax-Checkpoints sind nicht betroffen und können in PyTorch-Architekturen mit den Kwargs `from_tf` und `from_flax` für die Methode `from_pretrained` geladen werden, um dieses Problem zu umgehen.
+
+</Tip>
+
+Im Allgemeinen empfehlen wir die Verwendung der Klasse "AutoTokenizer" und der Klasse "AutoModelFor", um trainierte Instanzen von Modellen zu laden. Dadurch wird sichergestellt, dass Sie jedes Mal die richtige Architektur laden. Im nächsten [Tutorial] (Vorverarbeitung) erfahren Sie, wie Sie Ihren neu geladenen Tokenizer, Feature Extractor und Prozessor verwenden, um einen Datensatz für die Feinabstimmung vorzuverarbeiten.
+</pt>
+<tf>
+Mit den Klassen `TFAutoModelFor` schließlich können Sie ein vortrainiertes Modell für eine bestimmte Aufgabe laden (siehe [hier](model_doc/auto) für eine vollständige Liste der verfügbaren Aufgaben). Laden Sie zum Beispiel ein Modell für die Sequenzklassifikation mit [`TFAutoModelForSequenceClassification.from_pretrained`]:
+
+```py
+>>> from transformers import TFAutoModelForSequenceClassification
+
+>>> model = TFAutoModelForSequenceClassification.from_pretrained("distilbert-base-uncased")
+```
+
+Sie können denselben Prüfpunkt problemlos wiederverwenden, um eine Architektur für eine andere Aufgabe zu laden:
+
+```py
+>>> from transformers import TFAutoModelForTokenClassification
+
+>>> model = TFAutoModelForTokenClassification.from_pretrained("distilbert-base-uncased")
+```
+
+Im Allgemeinen empfehlen wir, die Klasse "AutoTokenizer" und die Klasse "TFAutoModelFor" zu verwenden, um vortrainierte Instanzen von Modellen zu laden. Dadurch wird sichergestellt, dass Sie jedes Mal die richtige Architektur laden. Im nächsten [Tutorial] (Vorverarbeitung) erfahren Sie, wie Sie Ihren neu geladenen Tokenizer, Feature Extractor und Prozessor verwenden, um einen Datensatz für die Feinabstimmung vorzuverarbeiten.
+</tf>
+</frameworkcontent>

docs/source/de/index.mdx

@@ -0,0 +1,322 @@
+<!--Copyright 2020 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# 🤗 Transformers
+
+Maschinelles Lernen auf dem neuesten Stand der Technik für PyTorch, TensorFlow und JAX.
+
+🤗 Transformers bietet APIs zum einfachen Herunterladen und Trainieren von vortrainierten Modellen auf dem neuesten Stand der Technik. Die Verwendung von vortrainierten Modellen kann Rechenkosten sparen und den CO2-Fußabdruck reduzieren und Zeit sparen, die für das Training eines Modells von Grund auf benötigt wird. Die Modelle können für verschiedene Modalitäten verwendet werden, wie z. B.:
+
+* 📝 Text: Textklassifizierung, Informationsextrahierung, Beantwortung von Fragen, Zusammenfassung, Übersetzung und Texterstellung in über 100 Sprachen.
+* 🖼️ Bilder: Bildklassifizierung, Objekterkennung und Segmentierung.
+* 🗣️ Audio: Spracherkennung und Audioklassifizierung.
+* 🐙 Multimodal: Beantwortung von Tabellenfragen, optische Zeichenerkennung, Informationsextraktion aus gescannten Dokumenten, Videoklassifizierung und Beantwortung visueller Fragen.
+
+Unsere Bibliothek unterstützt die nahtlose Integration von drei der beliebtesten Deep-Learning-Bibliotheken: [PyTorch](https://pytorch.org/), [TensorFlow](https://www.tensorflow.org/) und [JAX](https://jax.readthedocs.io/en/latest/). Trainieren Sie Ihr Modell in drei Codezeilen in einem Framework und laden Sie es zur Inferenz mit einem anderen.
+
+Jede 🤗 Transformers-Architektur ist in einem eigenständigen Python-Modul definiert, so dass sie leicht für Forschung und Experimente angepasst werden kann.
+
+## Wenn Sie auf der Suche nach individueller Unterstützung durch das Hugging Face-Team sind
+
+<a target="_blank" href="https://huggingface.co/support">
+    <img alt="HuggingFace Expert Acceleration Program" src="https://cdn-media.huggingface.co/marketing/transformers/new-support-improved.png" style="max-width: 600px; border: 1px solid #eee; border-radius: 4px; box-shadow: 0 1px 2px 0 rgba(0, 0, 0, 0.05);">
+</a><br>
+
+## Inhalt
+
+Die Dokumentation ist in fünf Teile gegliedert:
+
+- **GET STARTED** enthält eine kurze Tour und Installationsanweisungen, um mit 🤗 Transformers loszulegen.
+- **TUTORIALS** sind ein hervorragender Ausgangspunkt, wenn Sie neu in unserer Bibliothek sind. Dieser Abschnitt hilft Ihnen, die grundlegenden Fähigkeiten zu erlangen, die Sie benötigen, um mit 🤗 Transformers zu arbeiten.
+- **HOW-TO GUIDES** zeigen Ihnen, wie Sie ein bestimmtes Ziel erreichen können, z. B. die Feinabstimmung eines vortrainierten Modells für die Sprachmodellierung oder die Erstellung eines benutzerdefinierten Modellkopfs.
+- **KONZEPTUELLE ANLEITUNGEN** bietet weitere Diskussionen und Erklärungen zu den zugrunde liegenden Konzepten und Ideen hinter Modellen, Aufgaben und der Designphilosophie von 🤗 Transformers. 
+- **API** beschreibt jede Klasse und Funktion, gruppiert in:
+
+  - **MAIN CLASSES** für die Hauptklassen, die die wichtigsten APIs der Bibliothek darstellen.
+  - MODELLE** für die Klassen und Funktionen, die zu jedem in der Bibliothek implementierten Modell gehören.
+  - **INTERNAL HELPERS** für die Klassen und Funktionen, die wir intern verwenden.
+
+Die Bibliothek enthält derzeit JAX-, PyTorch- und TensorFlow-Implementierungen, vortrainierte Modellgewichte, Nutzungsskripte und Konvertierungsprogramme für die folgenden Modelle.
+
+### Unterstütze Modelle
+
+<!--This list is updated automatically from the README with _make fix-copies_. Do not update manually! -->
+
+1. **[ALBERT](model_doc/albert)** (from Google Research and the Toyota Technological Institute at Chicago) released with the paper [ALBERT: A Lite BERT for Self-supervised Learning of Language Representations](https://arxiv.org/abs/1909.11942), by Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, Radu Soricut.
+1. **[BART](model_doc/bart)** (from Facebook) released with the paper [BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://arxiv.org/abs/1910.13461) by Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Ves Stoyanov and Luke Zettlemoyer.
+1. **[BARThez](model_doc/barthez)** (from École polytechnique) released with the paper [BARThez: a Skilled Pretrained French Sequence-to-Sequence Model](https://arxiv.org/abs/2010.12321) by Moussa Kamal Eddine, Antoine J.-P. Tixier, Michalis Vazirgiannis.
+1. **[BARTpho](model_doc/bartpho)** (from VinAI Research) released with the paper [BARTpho: Pre-trained Sequence-to-Sequence Models for Vietnamese](https://arxiv.org/abs/2109.09701) by Nguyen Luong Tran, Duong Minh Le and Dat Quoc Nguyen.
+1. **[BEiT](model_doc/beit)** (from Microsoft) released with the paper [BEiT: BERT Pre-Training of Image Transformers](https://arxiv.org/abs/2106.08254) by Hangbo Bao, Li Dong, Furu Wei.
+1. **[BERT](model_doc/bert)** (from Google) released with the paper [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://arxiv.org/abs/1810.04805) by Jacob Devlin, Ming-Wei Chang, Kenton Lee and Kristina Toutanova.
+1. **[BERT For Sequence Generation](model_doc/bert-generation)** (from Google) released with the paper [Leveraging Pre-trained Checkpoints for Sequence Generation Tasks](https://arxiv.org/abs/1907.12461) by Sascha Rothe, Shashi Narayan, Aliaksei Severyn.
+1. **[BERTweet](model_doc/bertweet)** (from VinAI Research) released with the paper [BERTweet: A pre-trained language model for English Tweets](https://aclanthology.org/2020.emnlp-demos.2/) by Dat Quoc Nguyen, Thanh Vu and Anh Tuan Nguyen.
+1. **[BigBird-Pegasus](model_doc/bigbird_pegasus)** (from Google Research) released with the paper [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed.
+1. **[BigBird-RoBERTa](model_doc/big_bird)** (from Google Research) released with the paper [Big Bird: Transformers for Longer Sequences](https://arxiv.org/abs/2007.14062) by Manzil Zaheer, Guru Guruganesh, Avinava Dubey, Joshua Ainslie, Chris Alberti, Santiago Ontanon, Philip Pham, Anirudh Ravula, Qifan Wang, Li Yang, Amr Ahmed.
+1. **[Blenderbot](model_doc/blenderbot)** (from Facebook) released with the paper [Recipes for building an open-domain chatbot](https://arxiv.org/abs/2004.13637) by Stephen Roller, Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Kurt Shuster, Eric M. Smith, Y-Lan Boureau, Jason Weston.
+1. **[BlenderbotSmall](model_doc/blenderbot-small)** (from Facebook) released with the paper [Recipes for building an open-domain chatbot](https://arxiv.org/abs/2004.13637) by Stephen Roller, Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Kurt Shuster, Eric M. Smith, Y-Lan Boureau, Jason Weston.
+1. **[BLOOM](model_doc/bloom)** (from BigScience workshop) released by the [BigSicence Workshop](https://bigscience.huggingface.co/).
+1. **[BORT](model_doc/bort)** (from Alexa) released with the paper [Optimal Subarchitecture Extraction For BERT](https://arxiv.org/abs/2010.10499) by Adrian de Wynter and Daniel J. Perry.
+1. **[ByT5](model_doc/byt5)** (from Google Research) released with the paper [ByT5: Towards a token-free future with pre-trained byte-to-byte models](https://arxiv.org/abs/2105.13626) by Linting Xue, Aditya Barua, Noah Constant, Rami Al-Rfou, Sharan Narang, Mihir Kale, Adam Roberts, Colin Raffel.
+1. **[CamemBERT](model_doc/camembert)** (from Inria/Facebook/Sorbonne) released with the paper [CamemBERT: a Tasty French Language Model](https://arxiv.org/abs/1911.03894) by Louis Martin*, Benjamin Muller*, Pedro Javier Ortiz Suárez*, Yoann Dupont, Laurent Romary, Éric Villemonte de la Clergerie, Djamé Seddah and Benoît Sagot.
+1. **[CANINE](model_doc/canine)** (from Google Research) released with the paper [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) by Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting.
+1. **[CLIP](model_doc/clip)** (from OpenAI) released with the paper [Learning Transferable Visual Models From Natural Language Supervision](https://arxiv.org/abs/2103.00020) by Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, Ilya Sutskever.
+1. **[CodeGen](model_doc/codegen)** (from Salesforce) released with the paper [A Conversational Paradigm for Program Synthesis](https://arxiv.org/abs/2203.13474) by Erik Nijkamp, Bo Pang, Hiroaki Hayashi, Lifu Tu, Huan Wang, Yingbo Zhou, Silvio Savarese, Caiming Xiong.
+1. **[ConvBERT](model_doc/convbert)** (from YituTech) released with the paper [ConvBERT: Improving BERT with Span-based Dynamic Convolution](https://arxiv.org/abs/2008.02496) by Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan.
+1. **[ConvNeXT](model_doc/convnext)** (from Facebook AI) released with the paper [A ConvNet for the 2020s](https://arxiv.org/abs/2201.03545) by Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie.
+1. **[CPM](model_doc/cpm)** (from Tsinghua University) released with the paper [CPM: A Large-scale Generative Chinese Pre-trained Language Model](https://arxiv.org/abs/2012.00413) by Zhengyan Zhang, Xu Han, Hao Zhou, Pei Ke, Yuxian Gu, Deming Ye, Yujia Qin, Yusheng Su, Haozhe Ji, Jian Guan, Fanchao Qi, Xiaozhi Wang, Yanan Zheng, Guoyang Zeng, Huanqi Cao, Shengqi Chen, Daixuan Li, Zhenbo Sun, Zhiyuan Liu, Minlie Huang, Wentao Han, Jie Tang, Juanzi Li, Xiaoyan Zhu, Maosong Sun.
+1. **[CTRL](model_doc/ctrl)** (from Salesforce) released with the paper [CTRL: A Conditional Transformer Language Model for Controllable Generation](https://arxiv.org/abs/1909.05858) by Nitish Shirish Keskar*, Bryan McCann*, Lav R. Varshney, Caiming Xiong and Richard Socher.
+1. **[CvT](model_doc/cvt)** (from Microsoft) released with the paper [CvT: Introducing Convolutions to Vision Transformers](https://arxiv.org/abs/2103.15808) by Haiping Wu, Bin Xiao, Noel Codella, Mengchen Liu, Xiyang Dai, Lu Yuan, Lei Zhang.
+1. **[Data2Vec](model_doc/data2vec)** (from Facebook) released with the paper [Data2Vec:  A General Framework for Self-supervised Learning in Speech, Vision and Language](https://arxiv.org/abs/2202.03555) by Alexei Baevski, Wei-Ning Hsu, Qiantong Xu, Arun Babu, Jiatao Gu, Michael Auli.
+1. **[DeBERTa](model_doc/deberta)** (from Microsoft) released with the paper [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://arxiv.org/abs/2006.03654) by Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen.
+1. **[DeBERTa-v2](model_doc/deberta-v2)** (from Microsoft) released with the paper [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://arxiv.org/abs/2006.03654) by Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen.
+1. **[Decision Transformer](model_doc/decision_transformer)** (from Berkeley/Facebook/Google) released with the paper [Decision Transformer: Reinforcement Learning via Sequence Modeling](https://arxiv.org/abs/2106.01345) by Lili Chen, Kevin Lu, Aravind Rajeswaran, Kimin Lee, Aditya Grover, Michael Laskin, Pieter Abbeel, Aravind Srinivas, Igor Mordatch.
+1. **[DeiT](model_doc/deit)** (from Facebook) released with the paper [Training data-efficient image transformers & distillation through attention](https://arxiv.org/abs/2012.12877) by Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, Hervé Jégou.
+1. **[DETR](model_doc/detr)** (from Facebook) released with the paper [End-to-End Object Detection with Transformers](https://arxiv.org/abs/2005.12872) by Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, Sergey Zagoruyko.
+1. **[DialoGPT](model_doc/dialogpt)** (from Microsoft Research) released with the paper [DialoGPT: Large-Scale Generative Pre-training for Conversational Response Generation](https://arxiv.org/abs/1911.00536) by Yizhe Zhang, Siqi Sun, Michel Galley, Yen-Chun Chen, Chris Brockett, Xiang Gao, Jianfeng Gao, Jingjing Liu, Bill Dolan.
+1. **[DistilBERT](model_doc/distilbert)** (from HuggingFace), released together with the paper [DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter](https://arxiv.org/abs/1910.01108) by Victor Sanh, Lysandre Debut and Thomas Wolf. The same method has been applied to compress GPT2 into [DistilGPT2](https://github.com/huggingface/transformers/tree/main/examples/research_projects/distillation), RoBERTa into [DistilRoBERTa](https://github.com/huggingface/transformers/tree/main/examples/research_projects/distillation), Multilingual BERT into [DistilmBERT](https://github.com/huggingface/transformers/tree/main/examples/research_projects/distillation) and a German version of DistilBERT.
+1. **[DiT](model_doc/dit)** (from Microsoft Research) released with the paper [DiT: Self-supervised Pre-training for Document Image Transformer](https://arxiv.org/abs/2203.02378) by Junlong Li, Yiheng Xu, Tengchao Lv, Lei Cui, Cha Zhang, Furu Wei.
+1. **[DPR](model_doc/dpr)** (from Facebook) released with the paper [Dense Passage Retrieval for Open-Domain Question Answering](https://arxiv.org/abs/2004.04906) by Vladimir Karpukhin, Barlas Oğuz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih.
+1. **[DPT](master/model_doc/dpt)** (from Intel Labs) released with the paper [Vision Transformers for Dense Prediction](https://arxiv.org/abs/2103.13413) by René Ranftl, Alexey Bochkovskiy, Vladlen Koltun.
+1. **[ELECTRA](model_doc/electra)** (from Google Research/Stanford University) released with the paper [ELECTRA: Pre-training text encoders as discriminators rather than generators](https://arxiv.org/abs/2003.10555) by Kevin Clark, Minh-Thang Luong, Quoc V. Le, Christopher D. Manning.
+1. **[EncoderDecoder](model_doc/encoder-decoder)** (from Google Research) released with the paper [Leveraging Pre-trained Checkpoints for Sequence Generation Tasks](https://arxiv.org/abs/1907.12461) by Sascha Rothe, Shashi Narayan, Aliaksei Severyn.
+1. **[FlauBERT](model_doc/flaubert)** (from CNRS) released with the paper [FlauBERT: Unsupervised Language Model Pre-training for French](https://arxiv.org/abs/1912.05372) by Hang Le, Loïc Vial, Jibril Frej, Vincent Segonne, Maximin Coavoux, Benjamin Lecouteux, Alexandre Allauzen, Benoît Crabbé, Laurent Besacier, Didier Schwab.
+1. **[FLAVA](model_doc/flava)** (from Facebook AI) released with the paper [FLAVA: A Foundational Language And Vision Alignment Model](https://arxiv.org/abs/2112.04482) by Amanpreet Singh, Ronghang Hu, Vedanuj Goswami, Guillaume Couairon, Wojciech Galuba, Marcus Rohrbach, and Douwe Kiela.
+1. **[FNet](model_doc/fnet)** (from Google Research) released with the paper [FNet: Mixing Tokens with Fourier Transforms](https://arxiv.org/abs/2105.03824) by James Lee-Thorp, Joshua Ainslie, Ilya Eckstein, Santiago Ontanon.
+1. **[Funnel Transformer](model_doc/funnel)** (from CMU/Google Brain) released with the paper [Funnel-Transformer: Filtering out Sequential Redundancy for Efficient Language Processing](https://arxiv.org/abs/2006.03236) by Zihang Dai, Guokun Lai, Yiming Yang, Quoc V. Le.
+1. **[GLPN](model_doc/glpn)** (from KAIST) released with the paper [Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth](https://arxiv.org/abs/2201.07436) by Doyeon Kim, Woonghyun Ga, Pyungwhan Ahn, Donggyu Joo, Sehwan Chun, Junmo Kim.
+1. **[GPT](model_doc/openai-gpt)** (from OpenAI) released with the paper [Improving Language Understanding by Generative Pre-Training](https://blog.openai.com/language-unsupervised/) by Alec Radford, Karthik Narasimhan, Tim Salimans and Ilya Sutskever.
+1. **[GPT Neo](model_doc/gpt_neo)** (from EleutherAI) released in the repository [EleutherAI/gpt-neo](https://github.com/EleutherAI/gpt-neo) by Sid Black, Stella Biderman, Leo Gao, Phil Wang and Connor Leahy.
+1. **[GPT NeoX](model_doc/gpt_neox)** (from EleutherAI) released with the paper [GPT-NeoX-20B: An Open-Source Autoregressive Language Model](https://arxiv.org/abs/2204.06745) by Sid Black, Stella Biderman, Eric Hallahan, Quentin Anthony, Leo Gao, Laurence Golding, Horace He, Connor Leahy, Kyle McDonell, Jason Phang, Michael Pieler, USVSN Sai Prashanth, Shivanshu Purohit, Laria Reynolds, Jonathan Tow, Ben Wang, Samuel Weinbach
+1. **[GPT-2](model_doc/gpt2)** (from OpenAI) released with the paper [Language Models are Unsupervised Multitask Learners](https://blog.openai.com/better-language-models/) by Alec Radford*, Jeffrey Wu*, Rewon Child, David Luan, Dario Amodei** and Ilya Sutskever**.
+1. **[GPT-J](model_doc/gptj)** (from EleutherAI) released in the repository [kingoflolz/mesh-transformer-jax](https://github.com/kingoflolz/mesh-transformer-jax/) by Ben Wang and Aran Komatsuzaki.
+1. **[GroupViT](model_doc/groupvit)** (from UCSD, NVIDIA) released with the paper [GroupViT: Semantic Segmentation Emerges from Text Supervision](https://arxiv.org/abs/2202.11094) by Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, Xiaolong Wang.
+1. **[Hubert](model_doc/hubert)** (from Facebook) released with the paper [HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://arxiv.org/abs/2106.07447) by Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Ruslan Salakhutdinov, Abdelrahman Mohamed.
+1. **[I-BERT](model_doc/ibert)** (from Berkeley) released with the paper [I-BERT: Integer-only BERT Quantization](https://arxiv.org/abs/2101.01321) by Sehoon Kim, Amir Gholami, Zhewei Yao, Michael W. Mahoney, Kurt Keutzer.
+1. **[ImageGPT](model_doc/imagegpt)** (from OpenAI) released with the paper [Generative Pretraining from Pixels](https://openai.com/blog/image-gpt/) by Mark Chen, Alec Radford, Rewon Child, Jeffrey Wu, Heewoo Jun, David Luan, Ilya Sutskever.
+1. **[LayoutLM](model_doc/layoutlm)** (from Microsoft Research Asia) released with the paper [LayoutLM: Pre-training of Text and Layout for Document Image Understanding](https://arxiv.org/abs/1912.13318) by Yiheng Xu, Minghao Li, Lei Cui, Shaohan Huang, Furu Wei, Ming Zhou.
+1. **[LayoutLMv2](model_doc/layoutlmv2)** (from Microsoft Research Asia) released with the paper [LayoutLMv2: Multi-modal Pre-training for Visually-Rich Document Understanding](https://arxiv.org/abs/2012.14740) by Yang Xu, Yiheng Xu, Tengchao Lv, Lei Cui, Furu Wei, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Wanxiang Che, Min Zhang, Lidong Zhou.
+1. **[LayoutLMv3](model_doc/layoutlmv3)** (from Microsoft Research Asia) released with the paper [LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking](https://arxiv.org/abs/2204.08387) by Yupan Huang, Tengchao Lv, Lei Cui, Yutong Lu, Furu Wei.
+1. **[LayoutXLM](model_doc/layoutxlm)** (from Microsoft Research Asia) released with the paper [LayoutXLM: Multimodal Pre-training for Multilingual Visually-rich Document Understanding](https://arxiv.org/abs/2104.08836) by Yiheng Xu, Tengchao Lv, Lei Cui, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Furu Wei.
+1. **[LED](model_doc/led)** (from AllenAI) released with the paper [Longformer: The Long-Document Transformer](https://arxiv.org/abs/2004.05150) by Iz Beltagy, Matthew E. Peters, Arman Cohan.
+1. **[LeViT](model_doc/levit)** (from Meta AI) released with the paper [LeViT: A Vision Transformer in ConvNet's Clothing for Faster Inference](https://arxiv.org/abs/2104.01136) by Ben Graham, Alaaeldin El-Nouby, Hugo Touvron, Pierre Stock, Armand Joulin, Hervé Jégou, Matthijs Douze.
+1. **[Longformer](model_doc/longformer)** (from AllenAI) released with the paper [Longformer: The Long-Document Transformer](https://arxiv.org/abs/2004.05150) by Iz Beltagy, Matthew E. Peters, Arman Cohan.
+1. **[LongT5](model_doc/longt5)** (from Google AI) released with the paper [LongT5: Efficient Text-To-Text Transformer for Long Sequences](https://arxiv.org/abs/2112.07916) by Mandy Guo, Joshua Ainslie, David Uthus, Santiago Ontanon, Jianmo Ni, Yun-Hsuan Sung, Yinfei Yang.
+1. **[LUKE](model_doc/luke)** (from Studio Ousia) released with the paper [LUKE: Deep Contextualized Entity Representations with Entity-aware Self-attention](https://arxiv.org/abs/2010.01057) by Ikuya Yamada, Akari Asai, Hiroyuki Shindo, Hideaki Takeda, Yuji Matsumoto.
+1. **[LXMERT](model_doc/lxmert)** (from UNC Chapel Hill) released with the paper [LXMERT: Learning Cross-Modality Encoder Representations from Transformers for Open-Domain Question Answering](https://arxiv.org/abs/1908.07490) by Hao Tan and Mohit Bansal.
+1. **[M-CTC-T](model_doc/mctct)** (from Facebook) released with the paper [Pseudo-Labeling For Massively Multilingual Speech Recognition](https://arxiv.org/abs/2111.00161) by Loren Lugosch, Tatiana Likhomanenko, Gabriel Synnaeve, and Ronan Collobert.
+1. **[M2M100](model_doc/m2m_100)** (from Facebook) released with the paper [Beyond English-Centric Multilingual Machine Translation](https://arxiv.org/abs/2010.11125) by Angela Fan, Shruti Bhosale, Holger Schwenk, Zhiyi Ma, Ahmed El-Kishky, Siddharth Goyal, Mandeep Baines, Onur Celebi, Guillaume Wenzek, Vishrav Chaudhary, Naman Goyal, Tom Birch, Vitaliy Liptchinsky, Sergey Edunov, Edouard Grave, Michael Auli, Armand Joulin.
+1. **[MarianMT](model_doc/marian)** Machine translation models trained using [OPUS](http://opus.nlpl.eu/) data by Jörg Tiedemann. The [Marian Framework](https://marian-nmt.github.io/) is being developed by the Microsoft Translator Team.
+1. **[MaskFormer](model_doc/maskformer)** (from Meta and UIUC) released with the paper [Per-Pixel Classification is Not All You Need for Semantic Segmentation](https://arxiv.org/abs/2107.06278) by Bowen Cheng, Alexander G. Schwing, Alexander Kirillov.
+1. **[mBART](model_doc/mbart)** (from Facebook) released with the paper [Multilingual Denoising Pre-training for Neural Machine Translation](https://arxiv.org/abs/2001.08210) by Yinhan Liu, Jiatao Gu, Naman Goyal, Xian Li, Sergey Edunov, Marjan Ghazvininejad, Mike Lewis, Luke Zettlemoyer.
+1. **[mBART-50](model_doc/mbart)** (from Facebook) released with the paper [Multilingual Translation with Extensible Multilingual Pretraining and Finetuning](https://arxiv.org/abs/2008.00401) by Yuqing Tang, Chau Tran, Xian Li, Peng-Jen Chen, Naman Goyal, Vishrav Chaudhary, Jiatao Gu, Angela Fan.
+1. **[Megatron-BERT](model_doc/megatron-bert)** (from NVIDIA) released with the paper [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) by Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro.
+1. **[Megatron-GPT2](model_doc/megatron_gpt2)** (from NVIDIA) released with the paper [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://arxiv.org/abs/1909.08053) by Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro.
+1. **[mLUKE](model_doc/mluke)** (from Studio Ousia) released with the paper [mLUKE: The Power of Entity Representations in Multilingual Pretrained Language Models](https://arxiv.org/abs/2110.08151) by Ryokan Ri, Ikuya Yamada, and Yoshimasa Tsuruoka.
+1. **[MobileBERT](model_doc/mobilebert)** (from CMU/Google Brain) released with the paper [MobileBERT: a Compact Task-Agnostic BERT for Resource-Limited Devices](https://arxiv.org/abs/2004.02984) by Zhiqing Sun, Hongkun Yu, Xiaodan Song, Renjie Liu, Yiming Yang, and Denny Zhou.
+1. **[MobileViT](model_doc/mobilevit)** (from Apple) released with the paper [MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer](https://arxiv.org/abs/2110.02178) by Sachin Mehta and Mohammad Rastegari.
+1. **[MPNet](model_doc/mpnet)** (from Microsoft Research) released with the paper [MPNet: Masked and Permuted Pre-training for Language Understanding](https://arxiv.org/abs/2004.09297) by Kaitao Song, Xu Tan, Tao Qin, Jianfeng Lu, Tie-Yan Liu.
+1. **[MT5](model_doc/mt5)** (from Google AI) released with the paper [mT5: A massively multilingual pre-trained text-to-text transformer](https://arxiv.org/abs/2010.11934) by Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, Colin Raffel.
+1. **[MVP](model_doc/mvp)** (from RUC AI Box) released with the paper [MVP: Multi-task Supervised Pre-training for Natural Language Generation](https://arxiv.org/abs/2206.12131) by Tianyi Tang, Junyi Li, Wayne Xin Zhao and Ji-Rong Wen.
+1. **[Nezha](model_doc/nezha)** (from Huawei Noah’s Ark Lab) released with the paper [NEZHA: Neural Contextualized Representation for Chinese Language Understanding](https://arxiv.org/abs/1909.00204) by Junqiu Wei, Xiaozhe Ren, Xiaoguang Li, Wenyong Huang, Yi Liao, Yasheng Wang, Jiashu Lin, Xin Jiang, Xiao Chen and Qun Liu.
+1. **[NLLB](model_doc/nllb)** (from Meta) released with the paper [No Language Left Behind: Scaling Human-Centered Machine Translation](https://arxiv.org/abs/2207.04672) by the NLLB team.
+1. **[Nyströmformer](model_doc/nystromformer)** (from the University of Wisconsin - Madison) released with the paper [Nyströmformer: A Nyström-Based Algorithm for Approximating Self-Attention](https://arxiv.org/abs/2102.03902) by Yunyang Xiong, Zhanpeng Zeng, Rudrasis Chakraborty, Mingxing Tan, Glenn Fung, Yin Li, Vikas Singh.
+1. **[OPT](master/model_doc/opt)** (from Meta AI) released with the paper [OPT: Open Pre-trained Transformer Language Models](https://arxiv.org/abs/2205.01068) by Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen et al.
+1. **[OWL-ViT](model_doc/owlvit)** (from Google AI) released with the paper [Simple Open-Vocabulary Object Detection with Vision Transformers](https://arxiv.org/abs/2205.06230) by Matthias Minderer, Alexey Gritsenko, Austin Stone, Maxim Neumann, Dirk Weissenborn, Alexey Dosovitskiy, Aravindh Mahendran, Anurag Arnab, Mostafa Dehghani, Zhuoran Shen, Xiao Wang, Xiaohua Zhai, Thomas Kipf, and Neil Houlsby.
+1. **[Pegasus](model_doc/pegasus)** (from Google) released with the paper [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu.
+1. **[Perceiver IO](model_doc/perceiver)** (from Deepmind) released with the paper [Perceiver IO: A General Architecture for Structured Inputs & Outputs](https://arxiv.org/abs/2107.14795) by Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier Hénaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, João Carreira.
+1. **[PhoBERT](model_doc/phobert)** (from VinAI Research) released with the paper [PhoBERT: Pre-trained language models for Vietnamese](https://www.aclweb.org/anthology/2020.findings-emnlp.92/) by Dat Quoc Nguyen and Anh Tuan Nguyen.
+1. **[PLBart](model_doc/plbart)** (from UCLA NLP) released with the paper [Unified Pre-training for Program Understanding and Generation](https://arxiv.org/abs/2103.06333) by Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, Kai-Wei Chang.
+1. **[PoolFormer](model_doc/poolformer)** (from Sea AI Labs) released with the paper [MetaFormer is Actually What You Need for Vision](https://arxiv.org/abs/2111.11418) by Yu, Weihao and Luo, Mi and Zhou, Pan and Si, Chenyang and Zhou, Yichen and Wang, Xinchao and Feng, Jiashi and Yan, Shuicheng.
+1. **[ProphetNet](model_doc/prophetnet)** (from Microsoft Research) released with the paper [ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training](https://arxiv.org/abs/2001.04063) by Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou.
+1. **[QDQBert](model_doc/qdqbert)** (from NVIDIA) released with the paper [Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluation](https://arxiv.org/abs/2004.09602) by Hao Wu, Patrick Judd, Xiaojie Zhang, Mikhail Isaev and Paulius Micikevicius.
+1. **[RAG](model_doc/rag)** (from Facebook) released with the paper [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://arxiv.org/abs/2005.11401) by Patrick Lewis, Ethan Perez, Aleksandara Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, Douwe Kiela.
+1. **[REALM](model_doc/realm.html)** (from Google Research) released with the paper [REALM: Retrieval-Augmented Language Model Pre-Training](https://arxiv.org/abs/2002.08909) by Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat and Ming-Wei Chang.
+1. **[Reformer](model_doc/reformer)** (from Google Research) released with the paper [Reformer: The Efficient Transformer](https://arxiv.org/abs/2001.04451) by Nikita Kitaev, Łukasz Kaiser, Anselm Levskaya.
+1. **[RegNet](model_doc/regnet)** (from META Platforms) released with the paper [Designing Network Design Space](https://arxiv.org/abs/2003.13678) by Ilija Radosavovic, Raj Prateek Kosaraju, Ross Girshick, Kaiming He, Piotr Dollár.
+1. **[RemBERT](model_doc/rembert)** (from Google Research) released with the paper [Rethinking embedding coupling in pre-trained language models](https://arxiv.org/abs/2010.12821) by Hyung Won Chung, Thibault Févry, Henry Tsai, M. Johnson, Sebastian Ruder.
+1. **[ResNet](model_doc/resnet)** (from Microsoft Research) released with the paper [Deep Residual Learning for Image Recognition](https://arxiv.org/abs/1512.03385) by Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun.
+1. **[RoBERTa](model_doc/roberta)** (from Facebook), released together with the paper [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://arxiv.org/abs/1907.11692) by Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, Veselin Stoyanov.
+1. **[RoFormer](model_doc/roformer)** (from ZhuiyiTechnology), released together with the paper [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://arxiv.org/abs/2104.09864) by Jianlin Su and Yu Lu and Shengfeng Pan and Bo Wen and Yunfeng Liu.
+1. **[SegFormer](model_doc/segformer)** (from NVIDIA) released with the paper [SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers](https://arxiv.org/abs/2105.15203) by Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M. Alvarez, Ping Luo.
+1. **[SEW](model_doc/sew)** (from ASAPP) released with the paper [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870) by Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi.
+1. **[SEW-D](model_doc/sew_d)** (from ASAPP) released with the paper [Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition](https://arxiv.org/abs/2109.06870) by Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q. Weinberger, Yoav Artzi.
+1. **[SpeechToTextTransformer](model_doc/speech_to_text)** (from Facebook), released together with the paper [fairseq S2T: Fast Speech-to-Text Modeling with fairseq](https://arxiv.org/abs/2010.05171) by Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Dmytro Okhonko, Juan Pino.
+1. **[SpeechToTextTransformer2](model_doc/speech_to_text_2)** (from Facebook), released together with the paper [Large-Scale Self- and Semi-Supervised Learning for Speech Translation](https://arxiv.org/abs/2104.06678) by Changhan Wang, Anne Wu, Juan Pino, Alexei Baevski, Michael Auli, Alexis Conneau.
+1. **[Splinter](model_doc/splinter)** (from Tel Aviv University), released together with the paper [Few-Shot Question Answering by Pretraining Span Selection](https://arxiv.org/abs/2101.00438) by Ori Ram, Yuval Kirstain, Jonathan Berant, Amir Globerson, Omer Levy.
+1. **[SqueezeBERT](model_doc/squeezebert)** (from Berkeley) released with the paper [SqueezeBERT: What can computer vision teach NLP about efficient neural networks?](https://arxiv.org/abs/2006.11316) by Forrest N. Iandola, Albert E. Shaw, Ravi Krishna, and Kurt W. Keutzer.
+1. **[Swin Transformer](model_doc/swin)** (from Microsoft) released with the paper [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://arxiv.org/abs/2103.14030) by Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, Baining Guo.
+1. **[Swin Transformer V2](model_doc/swinv2)** (from Microsoft) released with the paper [Swin Transformer V2: Scaling Up Capacity and Resolution](https://arxiv.org/abs/2111.09883) by Ze Liu, Han Hu, Yutong Lin, Zhuliang Yao, Zhenda Xie, Yixuan Wei, Jia Ning, Yue Cao, Zheng Zhang, Li Dong, Furu Wei, Baining Guo.
+1. **[T5](model_doc/t5)** (from Google AI) released with the paper [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://arxiv.org/abs/1910.10683) by Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu.
+1. **[T5v1.1](model_doc/t5v1.1)** (from Google AI) released in the repository [google-research/text-to-text-transfer-transformer](https://github.com/google-research/text-to-text-transfer-transformer/blob/main/released_checkpoints.md#t511) by Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu.
+1. **[TAPAS](model_doc/tapas)** (from Google AI) released with the paper [TAPAS: Weakly Supervised Table Parsing via Pre-training](https://arxiv.org/abs/2004.02349) by Jonathan Herzig, Paweł Krzysztof Nowak, Thomas Müller, Francesco Piccinno and Julian Martin Eisenschlos.
+1. **[TAPEX](model_doc/tapex)** (from Microsoft Research) released with the paper [TAPEX: Table Pre-training via Learning a Neural SQL Executor](https://arxiv.org/abs/2107.07653) by Qian Liu, Bei Chen, Jiaqi Guo, Morteza Ziyadi, Zeqi Lin, Weizhu Chen, Jian-Guang Lou.
+1. **[Trajectory Transformer](model_doc/trajectory_transformers)** (from the University of California at Berkeley) released with the paper [Offline Reinforcement Learning as One Big Sequence Modeling Problem](https://arxiv.org/abs/2106.02039) by Michael Janner, Qiyang Li, Sergey Levine
+1. **[Transformer-XL](model_doc/transfo-xl)** (from Google/CMU) released with the paper [Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context](https://arxiv.org/abs/1901.02860) by Zihang Dai*, Zhilin Yang*, Yiming Yang, Jaime Carbonell, Quoc V. Le, Ruslan Salakhutdinov.
+1. **[TrOCR](model_doc/trocr)** (from Microsoft), released together with the paper [TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models](https://arxiv.org/abs/2109.10282) by Minghao Li, Tengchao Lv, Lei Cui, Yijuan Lu, Dinei Florencio, Cha Zhang, Zhoujun Li, Furu Wei.
+1. **[UL2](model_doc/ul2)** (from Google Research) released with the paper [Unifying Language Learning Paradigms](https://arxiv.org/abs/2205.05131v1) by Yi Tay, Mostafa Dehghani, Vinh Q. Tran, Xavier Garcia, Dara Bahri, Tal Schuster, Huaixiu Steven Zheng, Neil Houlsby, Donald Metzler
+1. **[UniSpeech](model_doc/unispeech)** (from Microsoft Research) released with the paper [UniSpeech: Unified Speech Representation Learning with Labeled and Unlabeled Data](https://arxiv.org/abs/2101.07597) by Chengyi Wang, Yu Wu, Yao Qian, Kenichi Kumatani, Shujie Liu, Furu Wei, Michael Zeng, Xuedong Huang.
+1. **[UniSpeechSat](model_doc/unispeech-sat)** (from Microsoft Research) released with the paper [UNISPEECH-SAT: UNIVERSAL SPEECH REPRESENTATION LEARNING WITH SPEAKER AWARE PRE-TRAINING](https://arxiv.org/abs/2110.05752) by Sanyuan Chen, Yu Wu, Chengyi Wang, Zhengyang Chen, Zhuo Chen, Shujie Liu, Jian Wu, Yao Qian, Furu Wei, Jinyu Li, Xiangzhan Yu.
+1. **[VAN](model_doc/van)** (from Tsinghua University and Nankai University) released with the paper [Visual Attention Network](https://arxiv.org/abs/2202.09741) by Meng-Hao Guo, Cheng-Ze Lu, Zheng-Ning Liu, Ming-Ming Cheng, Shi-Min Hu.
+1. **[VideoMAE](model_doc/videomae)** (from Multimedia Computing Group, Nanjing University) released with the paper [VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training](https://arxiv.org/abs/2203.12602) by Zhan Tong, Yibing Song, Jue Wang, Limin Wang.
+1. **[ViLT](model_doc/vilt)** (from NAVER AI Lab/Kakao Enterprise/Kakao Brain) released with the paper [ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision](https://arxiv.org/abs/2102.03334) by Wonjae Kim, Bokyung Son, Ildoo Kim.
+1. **[Vision Transformer (ViT)](model_doc/vit)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.
+1. **[VisualBERT](model_doc/visual_bert)** (from UCLA NLP) released with the paper [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) by Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang.
+1. **[ViTMAE](model_doc/vit_mae)** (from Meta AI) released with the paper [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) by Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick.
+1. **[Wav2Vec2](model_doc/wav2vec2)** (from Facebook AI) released with the paper [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) by Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli.
+1. **[Wav2Vec2-Conformer](model_doc/wav2vec2-conformer)** (from Facebook AI) released with the paper [FAIRSEQ S2T: Fast Speech-to-Text Modeling with FAIRSEQ](https://arxiv.org/abs/2010.05171) by Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Sravya Popuri, Dmytro Okhonko, Juan Pino.
+1. **[Wav2Vec2Phoneme](model_doc/wav2vec2_phoneme)** (from Facebook AI) released with the paper [Simple and Effective Zero-shot Cross-lingual Phoneme Recognition](https://arxiv.org/abs/2109.11680) by Qiantong Xu, Alexei Baevski, Michael Auli.
+1. **[WavLM](model_doc/wavlm)** (from Microsoft Research) released with the paper [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://arxiv.org/abs/2110.13900) by Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng, Furu Wei.
+1. **[XGLM](model_doc/xglm)** (From Facebook AI) released with the paper [Few-shot Learning with Multilingual Language Models](https://arxiv.org/abs/2112.10668) by Xi Victoria Lin, Todor Mihaylov, Mikel Artetxe, Tianlu Wang, Shuohui Chen, Daniel Simig, Myle Ott, Naman Goyal, Shruti Bhosale, Jingfei Du, Ramakanth Pasunuru, Sam Shleifer, Punit Singh Koura, Vishrav Chaudhary, Brian O'Horo, Jeff Wang, Luke Zettlemoyer, Zornitsa Kozareva, Mona Diab, Veselin Stoyanov, Xian Li.
+1. **[XLM](model_doc/xlm)** (from Facebook) released together with the paper [Cross-lingual Language Model Pretraining](https://arxiv.org/abs/1901.07291) by Guillaume Lample and Alexis Conneau.
+1. **[XLM-ProphetNet](model_doc/xlm-prophetnet)** (from Microsoft Research) released with the paper [ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training](https://arxiv.org/abs/2001.04063) by Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou.
+1. **[XLM-RoBERTa](model_doc/xlm-roberta)** (from Facebook AI), released together with the paper [Unsupervised Cross-lingual Representation Learning at Scale](https://arxiv.org/abs/1911.02116) by Alexis Conneau*, Kartikay Khandelwal*, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer and Veselin Stoyanov.
+1. **[XLM-RoBERTa-XL](model_doc/xlm-roberta-xl)** (from Facebook AI), released together with the paper [Larger-Scale Transformers for Multilingual Masked Language Modeling](https://arxiv.org/abs/2105.00572) by Naman Goyal, Jingfei Du, Myle Ott, Giri Anantharaman, Alexis Conneau.
+1. **[XLNet](model_doc/xlnet)** (from Google/CMU) released with the paper [​XLNet: Generalized Autoregressive Pretraining for Language Understanding](https://arxiv.org/abs/1906.08237) by Zhilin Yang*, Zihang Dai*, Yiming Yang, Jaime Carbonell, Ruslan Salakhutdinov, Quoc V. Le.
+1. **[XLS-R](model_doc/xls_r)** (from Facebook AI) released with the paper [XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale](https://arxiv.org/abs/2111.09296) by Arun Babu, Changhan Wang, Andros Tjandra, Kushal Lakhotia, Qiantong Xu, Naman Goyal, Kritika Singh, Patrick von Platen, Yatharth Saraf, Juan Pino, Alexei Baevski, Alexis Conneau, Michael Auli.
+1. **[XLSR-Wav2Vec2](model_doc/xlsr_wav2vec2)** (from Facebook AI) released with the paper [Unsupervised Cross-Lingual Representation Learning For Speech Recognition](https://arxiv.org/abs/2006.13979) by Alexis Conneau, Alexei Baevski, Ronan Collobert, Abdelrahman Mohamed, Michael Auli.
+1. **[YOLOS](model_doc/yolos)** (from Huazhong University of Science & Technology) released with the paper [You Only Look at One Sequence: Rethinking Transformer in Vision through Object Detection](https://arxiv.org/abs/2106.00666) by Yuxin Fang, Bencheng Liao, Xinggang Wang, Jiemin Fang, Jiyang Qi, Rui Wu, Jianwei Niu, Wenyu Liu.
+1. **[YOSO](model_doc/yoso)** (from the University of Wisconsin - Madison) released with the paper [You Only Sample (Almost) Once: Linear Cost Self-Attention Via Bernoulli Sampling](https://arxiv.org/abs/2111.09714) by Zhanpeng Zeng, Yunyang Xiong, Sathya N. Ravi, Shailesh Acharya, Glenn Fung, Vikas Singh.
+
+
+### Unterstützte Frameworks
+
+Die folgende Tabelle zeigt die derzeitige Unterstützung in der Bibliothek für jedes dieser Modelle, unabhängig davon, ob sie einen Python
+Tokenizer haben (als "langsam" bezeichnet), ein "schneller" Tokenizer, der von der 🤗 Tokenizers Bibliothek unterstützt wird, ob sie Unterstützung in Jax (via
+Flax), PyTorch, und/oder TensorFlow haben.
+
+<!--This table is updated automatically from the auto modules with _make fix-copies_. Do not update manually!-->
+
+|            Model            | Tokenizer slow | Tokenizer fast | PyTorch support | TensorFlow support | Flax Support |
+|:---------------------------:|:--------------:|:--------------:|:---------------:|:------------------:|:------------:|
+|           ALBERT            |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|            BART             |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|            BEiT             |       ❌       |       ❌       |       ✅        |         ❌         |      ✅      |
+|            BERT             |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|       Bert Generation       |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
+|           BigBird           |       ✅       |       ✅       |       ✅        |         ❌         |      ✅      |
+|       BigBird-Pegasus       |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|         Blenderbot          |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|       BlenderbotSmall       |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|            BLOOM            |       ❌       |       ✅       |       ✅        |         ❌         |      ❌      |
+|          CamemBERT          |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|           CANINE            |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
+|            CLIP             |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|           CodeGen           |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
+|          ConvBERT           |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|          ConvNeXT           |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
+|            CTRL             |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
+|             CvT             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|        Data2VecAudio        |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|        Data2VecText         |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|       Data2VecVision        |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
+|           DeBERTa           |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|         DeBERTa-v2          |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|    Decision Transformer     |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|            DeiT             |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
+|            DETR             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|         DistilBERT          |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|             DPR             |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|             DPT             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|           ELECTRA           |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|       Encoder decoder       |       ❌       |       ❌       |       ✅        |         ✅         |      ✅      |
+| FairSeq Machine-Translation |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
+|          FlauBERT           |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
+|            FLAVA            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|            FNet             |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
+|     Funnel Transformer      |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|            GLPN             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|           GPT Neo           |       ❌       |       ❌       |       ✅        |         ❌         |      ✅      |
+|          GPT NeoX           |       ❌       |       ✅       |       ✅        |         ❌         |      ❌      |
+|            GPT-J            |       ❌       |       ❌       |       ✅        |         ✅         |      ✅      |
+|          GroupViT           |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|           Hubert            |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
+|           I-BERT            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|          ImageGPT           |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|          LayoutLM           |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|         LayoutLMv2          |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
+|         LayoutLMv3          |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
+|             LED             |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|            LeViT            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|         Longformer          |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|           LongT5            |       ❌       |       ❌       |       ✅        |         ❌         |      ✅      |
+|            LUKE             |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
+|           LXMERT            |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|           M-CTC-T           |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|           M2M100            |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
+|           Marian            |       ✅       |       ❌       |       ✅        |         ✅         |      ✅      |
+|         MaskFormer          |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|            mBART            |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|        Megatron-BERT        |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|         MobileBERT          |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|          MobileViT          |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|            MPNet            |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|             MT5             |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|             MVP             |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
+|            Nezha            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|        Nyströmformer        |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|         OpenAI GPT          |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|        OpenAI GPT-2         |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|             OPT             |       ❌       |       ❌       |       ✅        |         ✅         |      ✅      |
+|           OWL-ViT           |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|           Pegasus           |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|          Perceiver          |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
+|           PLBart            |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
+|         PoolFormer          |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|         ProphetNet          |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
+|           QDQBert           |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|             RAG             |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
+|            REALM            |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
+|          Reformer           |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
+|           RegNet            |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
+|           RemBERT           |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|           ResNet            |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
+|          RetriBERT          |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
+|           RoBERTa           |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|          RoFormer           |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|          SegFormer          |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
+|             SEW             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|            SEW-D            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|   Speech Encoder decoder    |       ❌       |       ❌       |       ✅        |         ❌         |      ✅      |
+|         Speech2Text         |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
+|        Speech2Text2         |       ✅       |       ❌       |       ❌        |         ❌         |      ❌      |
+|          Splinter           |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
+|         SqueezeBERT         |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
+|      Swin Transformer       |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
+|     Swin Transformer V2     |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|             T5              |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|            TAPAS            |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
+|   Trajectory Transformer    |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|       Transformer-XL        |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
+|            TrOCR            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|          UniSpeech          |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|        UniSpeechSat         |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|             VAN             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|          VideoMAE           |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|            ViLT             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|   Vision Encoder decoder    |       ❌       |       ❌       |       ✅        |         ✅         |      ✅      |
+|    VisionTextDualEncoder    |       ❌       |       ❌       |       ✅        |         ❌         |      ✅      |
+|         VisualBERT          |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|             ViT             |       ❌       |       ❌       |       ✅        |         ✅         |      ✅      |
+|           ViTMAE            |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
+|          Wav2Vec2           |       ✅       |       ❌       |       ✅        |         ✅         |      ✅      |
+|     Wav2Vec2-Conformer      |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|            WavLM            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|            XGLM             |       ✅       |       ✅       |       ✅        |         ❌         |      ✅      |
+|             XLM             |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
+|       XLM-ProphetNet        |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
+|         XLM-RoBERTa         |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|       XLM-RoBERTa-XL        |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|            XLNet            |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
+|            YOLOS            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|            YOSO             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+
+<!-- End table-->

docs/source/de/installation.mdx

@@ -0,0 +1,246 @@
+<!---
+Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+-->
+
+# Installation
+
+Installieren Sie 🤗 Transformers für die Deep-Learning-Bibliothek, mit der Sie arbeiten, richten Sie Ihren Cache ein und konfigurieren Sie 🤗 Transformers optional für den Offline-Betrieb.
+
+🤗 Transformers wurde unter Python 3.6+, PyTorch 1.1.0+, TensorFlow 2.0+, und Flax getestet. Folgen Sie den Installationsanweisungen unten für die von Ihnen verwendete Deep-Learning-Bibliothek:
+
+* [PyTorch](https://pytorch.org/get-started/locally/) installation instructions.
+* [TensorFlow 2.0](https://www.tensorflow.org/install/pip) installation instructions.
+* [Flax](https://flax.readthedocs.io/en/latest/) installation instructions.
+
+## Installation mit pip
+
+Sie sollten 🤗 Transformers in einer [virtuellen Umgebung](https://docs.python.org/3/library/venv.html) installieren. Wenn Sie mit virtuellen Python-Umgebungen nicht vertraut sind, werfen Sie einen Blick auf diese [Anleitung](https://packaging.python.org/guides/installing-using-pip-and-virtual-environments/). Eine virtuelle Umgebung macht es einfacher, verschiedene Projekte zu verwalten und Kompatibilitätsprobleme zwischen Abhängigkeiten zu vermeiden.
+
+Beginnen wir mit der Erstellung einer virtuellen Umgebung in Ihrem Projektverzeichnis:
+
+
+```bash
+python -m venv .env
+```
+
+Aktivieren wir die virtuelle Umgebung. Unter Linux und MacOs:
+
+```bash
+source .env/bin/activate
+```
+Aktivieren wir die virtuelle Umgebung unter Windows
+
+```bash
+.env/Scripts/activate
+```
+
+Jetzt können wir die 🤗 Transformers mit dem folgenden Befehl installieren:
+
+```bash
+pip install transformers
+```
+
+Bei reiner CPU-Unterstützung können wir 🤗 Transformers und eine Deep-Learning-Bibliothek bequem in einer Zeile installieren. Installieren wir zum Beispiel 🤗 Transformers und PyTorch mit:
+
+```bash
+pip install transformers[torch]
+```
+
+🤗 Transformers und TensorFlow 2.0:
+
+```bash
+pip install transformers[tf-cpu]
+```
+
+🤗 Transformers und Flax:
+
+```bash
+pip install transformers[flax]
+```
+
+Überprüfen wir abschließend, ob 🤗 Transformers ordnungsgemäß installiert wurde, indem wir den folgenden Befehl ausführen. Es wird ein vortrainiertes Modell heruntergeladen:
+
+```bash
+python -c "from transformers import pipeline; print(pipeline('sentiment-analysis')('we love you'))"
+```
+
+Dann wird die Kategorie und die Wahrscheinlichkeit ausgegeben:
+
+```bash
+[{'label': 'POSITIVE', 'score': 0.9998704791069031}]
+```
+
+## Installation aus dem Code
+
+Installieren wir 🤗 Transformers aus dem Quellcode mit dem folgenden Befehl:
+
+```bash
+pip install git+https://github.com/huggingface/transformers
+```
+
+Dieser Befehl installiert die aktuelle `main` Version und nicht die neueste `stable` Version. Die `main`-Version ist nützlich, um mit den neuesten Entwicklungen Schritt zu halten. Zum Beispiel, wenn ein Fehler seit der letzten offiziellen Version behoben wurde, aber eine neue Version noch nicht veröffentlicht wurde. Das bedeutet jedoch, dass die "Hauptversion" nicht immer stabil ist. Wir bemühen uns, die Hauptversion einsatzbereit zu halten, und die meisten Probleme werden normalerweise innerhalb weniger Stunden oder eines Tages behoben. Wenn Sie auf ein Problem stoßen, öffnen Sie bitte ein [Issue] (https://github.com/huggingface/transformers/issues), damit wir es noch schneller beheben können!
+
+Überprüfen wir, ob 🤗 Transformers richtig installiert wurde, indem Sie den folgenden Befehl ausführen:
+
+
+```bash
+python -c "from transformers import pipeline; print(pipeline('sentiment-analysis')('I love you'))"
+```
+
+## Editierbare Installation
+
+Sie benötigen eine bearbeitbare Installation, wenn Sie:
+
+* die "Haupt"-Version des Quellcodes verwenden möchten.
+* Zu 🤗 Transformers beitragen und Änderungen am Code testen wollen.
+
+Klonen Sie das Repository und installieren 🤗 Transformers mit den folgenden Befehlen:
+
+```bash
+git clone https://github.com/huggingface/transformers.git
+cd transformers
+pip install -e .
+```
+
+Diese Befehle verknüpfen den Ordner, in den Sie das Repository geklont haben, mit den Pfaden Ihrer Python-Bibliotheken. Python wird nun in dem Ordner suchen, in den Sie geklont haben, zusätzlich zu den normalen Bibliothekspfaden. Wenn zum Beispiel Ihre Python-Pakete normalerweise in `~/anaconda3/envs/main/lib/python3.7/site-packages/` installiert sind, wird Python auch den Ordner durchsuchen, in den Sie geklont haben: `~/transformers/`.
+
+
+<Tip warning={true}>
+
+Sie müssen den Ordner `transformers` behalten, wenn Sie die Bibliothek weiter verwenden wollen.
+
+</Tip>
+
+Jetzt können Sie Ihren Klon mit dem folgenden Befehl ganz einfach auf die neueste Version von 🤗 Transformers aktualisieren:
+
+
+```bash
+cd ~/transformers/
+git pull
+```
+
+Ihre Python-Umgebung wird beim nächsten Ausführen die `main`-Version von 🤗 Transformers finden.
+
+## Installation mit conda
+
+Installation von dem conda Kanal `huggingface`:
+
+```bash
+conda install -c huggingface transformers
+```
+
+## Cache Einrichtung
+
+Vorgefertigte Modelle werden heruntergeladen und lokal zwischengespeichert unter: `~/.cache/huggingface/hub`. Dies ist das Standardverzeichnis, das durch die Shell-Umgebungsvariable "TRANSFORMERS_CACHE" vorgegeben ist. Unter Windows wird das Standardverzeichnis durch `C:\Benutzer\Benutzername\.cache\huggingface\hub` angegeben. Sie können die unten aufgeführten Shell-Umgebungsvariablen - in der Reihenfolge ihrer Priorität - ändern, um ein anderes Cache-Verzeichnis anzugeben:
+
+1. Shell-Umgebungsvariable (Standard): `HUGGINGFACE_HUB_CACHE` oder `TRANSFORMERS_CACHE`.
+2. Shell-Umgebungsvariable: `HF_HOME`.
+3. Shell-Umgebungsvariable: `XDG_CACHE_HOME` + `/huggingface`.
+
+
+<Tip>
+
+Transformers verwendet die Shell-Umgebungsvariablen `PYTORCH_TRANSFORMERS_CACHE` oder `PYTORCH_PRETRAINED_BERT_CACHE`, wenn Sie von einer früheren Iteration dieser Bibliothek kommen und diese Umgebungsvariablen gesetzt haben, sofern Sie nicht die Shell-Umgebungsvariable `TRANSFORMERS_CACHE` angeben.
+  
+</Tip>
+
+## Offline Modus
+
+Transformers ist in der Lage, in einer Firewall- oder Offline-Umgebung zu laufen, indem es nur lokale Dateien verwendet. Setzen Sie die Umgebungsvariable `TRANSFORMERS_OFFLINE=1`, um dieses Verhalten zu aktivieren.
+
+<Tip>
+
+Fügen sie [🤗 Datasets](https://huggingface.co/docs/datasets/) zu Ihrem Offline-Trainingsworkflow hinzufügen, indem Sie die Umgebungsvariable `HF_DATASETS_OFFLINE=1` setzen.
+
+</Tip>
+
+So würden Sie beispielsweise ein Programm in einem normalen Netzwerk mit einer Firewall für externe Instanzen mit dem folgenden Befehl ausführen:
+
+```bash
+python examples/pytorch/translation/run_translation.py --model_name_or_path t5-small --dataset_name wmt16 --dataset_config ro-en ...
+```
+
+Führen Sie das gleiche Programm in einer Offline-Instanz mit aus:
+
+```bash
+HF_DATASETS_OFFLINE=1 TRANSFORMERS_OFFLINE=1 \
+python examples/pytorch/translation/run_translation.py --model_name_or_path t5-small --dataset_name wmt16 --dataset_config ro-en ...
+```
+
+Das Skript sollte nun laufen, ohne sich aufzuhängen oder eine Zeitüberschreitung abzuwarten, da es weiß, dass es nur nach lokalen Dateien suchen soll.
+
+
+### Abrufen von Modellen und Tokenizern zur Offline-Verwendung
+
+Eine andere Möglichkeit, 🤗 Transformers offline zu verwenden, besteht darin, die Dateien im Voraus herunterzuladen und dann auf ihren lokalen Pfad zu verweisen, wenn Sie sie offline verwenden müssen. Es gibt drei Möglichkeiten, dies zu tun:
+
+* Laden Sie eine Datei über die Benutzeroberfläche des [Model Hub](https://huggingface.co/models) herunter, indem Sie auf das ↓-Symbol klicken.
+
+    ![download-icon](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/download-icon.png)
+
+* Verwenden Sie den [PreTrainedModel.from_pretrained] und [PreTrainedModel.save_pretrained] Workflow:
+
+    1. Laden Sie Ihre Dateien im Voraus mit [`PreTrainedModel.from_pretrained`] herunter:
+
+    ```py
+    >>> from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
+
+    >>> tokenizer = AutoTokenizer.from_pretrained("bigscience/T0_3B")
+    >>> model = AutoModelForSeq2SeqLM.from_pretrained("bigscience/T0_3B")
+    ```
+
+    2. Speichern Sie Ihre Dateien in einem bestimmten Verzeichnis mit [`PreTrainedModel.save_pretrained`]:
+
+    ```py
+    >>> tokenizer.save_pretrained("./your/path/bigscience_t0")
+    >>> model.save_pretrained("./your/path/bigscience_t0")
+    ```
+
+    3. Wenn Sie nun offline sind, laden Sie Ihre Dateien mit [`PreTrainedModel.from_pretrained`] aus dem bestimmten Verzeichnis:
+
+    ```py
+    >>> tokenizer = AutoTokenizer.from_pretrained("./your/path/bigscience_t0")
+    >>> model = AutoModel.from_pretrained("./your/path/bigscience_t0")
+    ```
+
+* Programmatisches Herunterladen von Dateien mit der [huggingface_hub](https://github.com/huggingface/huggingface_hub/tree/main/src/huggingface_hub) Bibliothek:
+
+    1. Installieren Sie die "huggingface_hub"-Bibliothek in Ihrer virtuellen Umgebung:
+
+    ```bash
+    python -m pip install huggingface_hub
+    ```
+
+    2. Verwenden Sie die Funktion [`hf_hub_download`](https://huggingface.co/docs/hub/adding-a-library#download-files-from-the-hub), um eine Datei in einen bestimmten Pfad herunterzuladen. Der folgende Befehl lädt zum Beispiel die Datei "config.json" aus dem Modell [T0](https://huggingface.co/bigscience/T0_3B) in den gewünschten Pfad herunter:
+
+    ```py
+    >>> from huggingface_hub import hf_hub_download
+
+    >>> hf_hub_download(repo_id="bigscience/T0_3B", filename="config.json", cache_dir="./your/path/bigscience_t0")
+    ```
+
+Sobald Ihre Datei heruntergeladen und lokal zwischengespeichert ist, geben Sie den lokalen Pfad an, um sie zu laden und zu verwenden:
+
+```py
+>>> from transformers import AutoConfig
+
+>>> config = AutoConfig.from_pretrained("./your/path/bigscience_t0/config.json")
+```
+
+<Tip>
+
+Weitere Informationen zum Herunterladen von Dateien, die auf dem Hub gespeichert sind, finden Sie im Abschnitt [Wie man Dateien vom Hub herunterlädt] (https://huggingface.co/docs/hub/how-to-downstream).
+  
+</Tip>

docs/source/de/model_sharing.mdx

@@ -0,0 +1,228 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# Ein Modell teilen
+
+Die letzten beiden Tutorials haben gezeigt, wie man ein Modell mit PyTorch, Keras und 🤗 Accelerate für verteilte Setups feinabstimmen kann. Der nächste Schritt besteht darin, Ihr Modell mit der Community zu teilen! Bei Hugging Face glauben wir an den offenen Austausch von Wissen und Ressourcen, um künstliche Intelligenz für alle zu demokratisieren. Wir ermutigen Sie, Ihr Modell mit der Community zu teilen, um anderen zu helfen, Zeit und Ressourcen zu sparen.
+
+In diesem Tutorial lernen Sie zwei Methoden kennen, wie Sie ein trainiertes oder verfeinertes Modell auf dem [Model Hub](https://huggingface.co/models) teilen können:
+
+- Programmgesteuertes Übertragen Ihrer Dateien auf den Hub.
+- Ziehen Sie Ihre Dateien per Drag-and-Drop über die Weboberfläche in den Hub.
+
+<iframe width="560" height="315" src="https://www.youtube.com/embed/XvSGPZFEjDY" title="YouTube video player"
+frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope;
+picture-in-picture" allowfullscreen></iframe>
+
+<Tip>
+
+Um ein Modell mit der Öffentlichkeit zu teilen, benötigen Sie ein Konto auf [huggingface.co](https://huggingface.co/join). Sie können auch einer bestehenden Organisation beitreten oder eine neue Organisation gründen.
+
+</Tip>
+
+## Repository-Funktionen
+
+Jedes Repository im Model Hub verhält sich wie ein typisches GitHub-Repository. Unsere Repositorys bieten Versionierung, Commit-Historie und die Möglichkeit, Unterschiede zu visualisieren.
+
+Die integrierte Versionierung des Model Hub basiert auf Git und [git-lfs](https://git-lfs.github.com/). Mit anderen Worten: Sie können ein Modell als ein Repository behandeln, was eine bessere Zugriffskontrolle und Skalierbarkeit ermöglicht. Die Versionskontrolle ermöglicht *Revisionen*, eine Methode zum Anheften einer bestimmten Version eines Modells mit einem Commit-Hash, Tag oder Branch.
+
+Folglich können Sie eine bestimmte Modellversion mit dem Parameter "Revision" laden:
+
+```py
+>>> model = AutoModel.from_pretrained(
+...     "julien-c/EsperBERTo-small", revision="v2.0.1"  # tag name, or branch name, or commit hash
+... )
+```
+
+Dateien lassen sich auch in einem Repository leicht bearbeiten, und Sie können die Commit-Historie sowie die Unterschiede einsehen:
+
+![vis_diff](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/vis_diff.png)
+
+## Einrichtung
+
+Bevor Sie ein Modell für den Hub freigeben, benötigen Sie Ihre Hugging Face-Anmeldedaten. Wenn Sie Zugang zu einem Terminal haben, führen Sie den folgenden Befehl in der virtuellen Umgebung aus, in der 🤗 Transformers installiert ist. Dadurch werden Ihre Zugangsdaten in Ihrem Hugging Face-Cache-Ordner (standardmäßig `~/.cache/`) gespeichert:
+
+```bash
+huggingface-cli login
+```
+
+Wenn Sie ein Notebook wie Jupyter oder Colaboratory verwenden, stellen Sie sicher, dass Sie die [`huggingface_hub`](https://huggingface.co/docs/hub/adding-a-library) Bibliothek installiert haben. Diese Bibliothek ermöglicht Ihnen die programmatische Interaktion mit dem Hub.
+
+```bash
+pip install huggingface_hub
+```
+
+Verwenden Sie dann `notebook_login`, um sich beim Hub anzumelden, und folgen Sie dem Link [hier](https://huggingface.co/settings/token), um ein Token für die Anmeldung zu generieren:
+
+```py
+>>> from huggingface_hub import notebook_login
+
+>>> notebook_login()
+```
+
+## Ein Modell für alle Frameworks konvertieren
+
+Um sicherzustellen, dass Ihr Modell von jemandem verwendet werden kann, der mit einem anderen Framework arbeitet, empfehlen wir Ihnen, Ihr Modell sowohl mit PyTorch- als auch mit TensorFlow-Checkpoints zu konvertieren und hochzuladen. Während Benutzer immer noch in der Lage sind, Ihr Modell von einem anderen Framework zu laden, wenn Sie diesen Schritt überspringen, wird es langsamer sein, weil 🤗 Transformers den Checkpoint on-the-fly konvertieren müssen.
+
+Die Konvertierung eines Checkpoints für ein anderes Framework ist einfach. Stellen Sie sicher, dass Sie PyTorch und TensorFlow installiert haben (siehe [hier](installation) für Installationsanweisungen), und finden Sie dann das spezifische Modell für Ihre Aufgabe in dem anderen Framework. 
+
+<frameworkcontent>
+<pt>
+Geben Sie `from_tf=True` an, um einen Prüfpunkt von TensorFlow nach PyTorch zu konvertieren:
+
+```py
+>>> pt_model = DistilBertForSequenceClassification.from_pretrained("path/to/awesome-name-you-picked", from_tf=True)
+>>> pt_model.save_pretrained("path/to/awesome-name-you-picked")
+```
+</pt>
+<tf>
+Geben Sie `from_pt=True` an, um einen Prüfpunkt von PyTorch nach TensorFlow zu konvertieren:
+
+```py
+>>> tf_model = TFDistilBertForSequenceClassification.from_pretrained("path/to/awesome-name-you-picked", from_pt=True)
+```
+
+Dann können Sie Ihr neues TensorFlow-Modell mit seinem neuen Checkpoint speichern:
+
+```py
+>>> tf_model.save_pretrained("path/to/awesome-name-you-picked")
+```
+</tf>
+<jax>
+Wenn ein Modell in Flax verfügbar ist, können Sie auch einen Kontrollpunkt von PyTorch nach Flax konvertieren:
+
+```py
+>>> flax_model = FlaxDistilBertForSequenceClassification.from_pretrained(
+...     "path/to/awesome-name-you-picked", from_pt=True
+... )
+```
+</jax>
+</frameworkcontent>
+
+## Ein Modell während des Trainings hochladen
+
+<frameworkcontent>
+<pt>
+<Youtube id="Z1-XMy-GNLQ"/>
+
+Die Weitergabe eines Modells an den Hub ist so einfach wie das Hinzufügen eines zusätzlichen Parameters oder Rückrufs. Erinnern Sie sich an das [Feinabstimmungs-Tutorial](training), in der Klasse [`TrainingArguments`] geben Sie Hyperparameter und zusätzliche Trainingsoptionen an. Eine dieser Trainingsoptionen beinhaltet die Möglichkeit, ein Modell direkt an den Hub zu pushen. Setzen Sie `push_to_hub=True` in Ihrer [`TrainingArguments`]:
+
+```py
+>>> training_args = TrainingArguments(output_dir="my-awesome-model", push_to_hub=True)
+```
+
+Übergeben Sie Ihre Trainingsargumente wie gewohnt an [`Trainer`]:
+
+```py
+>>> trainer = Trainer(
+...     model=model,
+...     args=training_args,
+...     train_dataset=small_train_dataset,
+...     eval_dataset=small_eval_dataset,
+...     compute_metrics=compute_metrics,
+... )
+```
+
+Nach der Feinabstimmung Ihres Modells rufen Sie [`~transformers.Trainer.push_to_hub`] auf [`Trainer`] auf, um das trainierte Modell an den Hub zu übertragen. Transformers fügt sogar automatisch Trainings-Hyperparameter, Trainingsergebnisse und Framework-Versionen zu Ihrer Modellkarte hinzu!
+
+```py
+>>> trainer.push_to_hub()
+```
+</pt>
+<tf>
+Geben Sie ein Modell mit [`PushToHubCallback`] an den Hub weiter. In der [`PushToHubCallback`] Funktion, fügen Sie hinzu:
+
+- Ein Ausgabeverzeichnis für Ihr Modell.
+- Einen Tokenizer.
+- Die `hub_model_id`, die Ihr Hub-Benutzername und Modellname ist.
+
+```py
+>>> from transformers.keras.callbacks import PushToHubCallback
+
+>>> push_to_hub_callback = PushToHubCallback(
+...     output_dir="./your_model_save_path", tokenizer=tokenizer, hub_model_id="your-username/my-awesome-model"
+... )
+```
+
+Fügen Sie den Callback zu [`fit`](https://keras.io/api/models/model_training_apis/) hinzu, und 🤗 Transformers wird das trainierte Modell an den Hub weiterleiten:
+
+```py
+>>> model.fit(tf_train_dataset, validation_data=tf_validation_dataset, epochs=3, callbacks=push_to_hub_callback)
+```
+</tf>
+</frameworkcontent>
+
+## Verwenden Sie die Funktion `push_to_hub`.
+
+Sie können `push_to_hub` auch direkt für Ihr Modell aufrufen, um es in den Hub hochzuladen.
+
+Geben Sie den Namen Ihres Modells in "push_to_hub" an:
+
+```py
+>>> pt_model.push_to_hub("my-awesome-model")
+```
+
+Dadurch wird ein Repository unter Ihrem Benutzernamen mit dem Modellnamen `my-awesome-model` erstellt. Benutzer können nun Ihr Modell mit der Funktion `from_pretrained` laden:
+
+```py
+>>> from transformers import AutoModel
+
+>>> model = AutoModel.from_pretrained("your_username/my-awesome-model")
+```
+
+Wenn Sie zu einer Organisation gehören und Ihr Modell stattdessen unter dem Namen der Organisation pushen wollen, fügen Sie diesen einfach zur `repo_id` hinzu:
+
+```py
+>>> pt_model.push_to_hub("my-awesome-org/my-awesome-model")
+```
+
+Die Funktion "push_to_hub" kann auch verwendet werden, um andere Dateien zu einem Modell-Repository hinzuzufügen. Zum Beispiel kann man einen Tokenizer zu einem Modell-Repository hinzufügen:
+
+```py
+>>> tokenizer.push_to_hub("my-awesome-model")
+```
+
+Oder vielleicht möchten Sie die TensorFlow-Version Ihres fein abgestimmten PyTorch-Modells hinzufügen:
+
+```py
+>>> tf_model.push_to_hub("my-awesome-model")
+```
+
+Wenn Sie nun zu Ihrem Hugging Face-Profil navigieren, sollten Sie Ihr neu erstelltes Modell-Repository sehen. Wenn Sie auf die Registerkarte **Dateien** klicken, werden alle Dateien angezeigt, die Sie in das Repository hochgeladen haben.
+
+Weitere Einzelheiten zum Erstellen und Hochladen von Dateien in ein Repository finden Sie in der Hub-Dokumentation [hier](https://huggingface.co/docs/hub/how-to-upstream).
+
+## Hochladen mit der Weboberfläche
+
+Benutzer, die einen no-code Ansatz bevorzugen, können ein Modell über das Webinterface des Hubs hochladen. Besuchen Sie [huggingface.co/new](https://huggingface.co/new) um ein neues Repository zu erstellen:
+
+![new_model_repo](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/new_model_repo.png)
+
+Fügen Sie von hier aus einige Informationen über Ihr Modell hinzu:
+
+- Wählen Sie den **Besitzer** des Repositorys. Dies können Sie selbst oder eine der Organisationen sein, denen Sie angehören.
+- Wählen Sie einen Namen für Ihr Modell, der auch der Name des Repositorys sein wird.
+- Wählen Sie, ob Ihr Modell öffentlich oder privat ist.
+- Geben Sie die Lizenzverwendung für Ihr Modell an.
+
+Klicken Sie nun auf die Registerkarte **Dateien** und klicken Sie auf die Schaltfläche **Datei hinzufügen**, um eine neue Datei in Ihr Repository hochzuladen. Ziehen Sie dann eine Datei per Drag-and-Drop hoch und fügen Sie eine Übergabemeldung hinzu.
+
+![upload_file](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/upload_file.png)
+
+## Hinzufügen einer Modellkarte
+
+Um sicherzustellen, dass die Benutzer die Fähigkeiten, Grenzen, möglichen Verzerrungen und ethischen Aspekte Ihres Modells verstehen, fügen Sie bitte eine Modellkarte zu Ihrem Repository hinzu. Die Modellkarte wird in der Datei `README.md` definiert. Sie können eine Modellkarte hinzufügen, indem Sie:
+
+* Manuelles Erstellen und Hochladen einer "README.md"-Datei.
+* Klicken Sie auf die Schaltfläche **Modellkarte bearbeiten** in Ihrem Modell-Repository.
+
+Werfen Sie einen Blick auf die DistilBert [model card](https://huggingface.co/distilbert-base-uncased) als gutes Beispiel für die Art von Informationen, die eine Modellkarte enthalten sollte. Weitere Details über andere Optionen, die Sie in der Datei "README.md" einstellen können, wie z.B. den Kohlenstoff-Fußabdruck eines Modells oder Beispiele für Widgets, finden Sie in der Dokumentation [hier](https://huggingface.co/docs/hub/models-cards).

docs/source/de/pipeline_tutorial.mdx

@@ -0,0 +1,171 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# Pipelines für Inferenzen
+
+Die [`pipeline`] macht es einfach, jedes beliebige Modell aus dem [Hub](https://huggingface.co/models) für die Inferenz auf jede Sprache, Computer Vision, Sprache und multimodale Aufgaben zu verwenden. Selbst wenn Sie keine Erfahrung mit einer bestimmten Modalität haben oder nicht mit dem zugrundeliegenden Code hinter den Modellen vertraut sind, können Sie sie mit der [`pipeline`] für Inferenzen verwenden! In diesem Beispiel lernen Sie, wie:
+
+* Eine [`pipeline`] für Inferenz zu verwenden.
+* Einen bestimmten Tokenizer oder ein bestimmtes Modell zu verwenden.
+* Eine [`pipeline`] für Audio-, Vision- und multimodale Aufgaben zu verwenden.
+
+<Tip>
+
+Eine vollständige Liste der unterstützten Aufgaben und verfügbaren Parameter finden Sie in der [`pipeline`]-Dokumentation.
+
+</Tip>
+
+## Verwendung von Pipelines
+
+Obwohl jede Aufgabe eine zugehörige [`pipeline`] hat, ist es einfacher, die allgemeine [`pipeline`]-Abstraktion zu verwenden, die alle aufgabenspezifischen Pipelines enthält. Die [`pipeline`] lädt automatisch ein Standardmodell und eine Vorverarbeitungsklasse, die für Ihre Aufgabe inferenzfähig ist.
+
+1. Beginnen Sie mit der Erstellung einer [`pipeline`] und geben Sie eine Inferenzaufgabe an:
+
+```py
+>>> from transformers import pipeline
+
+>>> generator = pipeline(task="text-generation")
+```
+
+2. Übergeben Sie Ihren Eingabetext an die [`pipeline`]:
+
+```py
+>>> generator(
+...     "Three Rings for the Elven-kings under the sky, Seven for the Dwarf-lords in their halls of stone"
+... )  # doctest: +SKIP
+[{'generated_text': 'Three Rings for the Elven-kings under the sky, Seven for the Dwarf-lords in their halls of stone, Seven for the Iron-priests at the door to the east, and thirteen for the Lord Kings at the end of the mountain'}]
+```
+
+Wenn Sie mehr als eine Eingabe haben, übergeben Sie die Eingabe als Liste:
+
+```py
+>>> generator(
+...     [
+...         "Three Rings for the Elven-kings under the sky, Seven for the Dwarf-lords in their halls of stone",
+...         "Nine for Mortal Men, doomed to die, One for the Dark Lord on his dark throne",
+...     ]
+... )  # doctest: +SKIP
+```
+
+Alle zusätzlichen Parameter für Ihre Aufgabe können auch in die [`pipeline`] aufgenommen werden. Die Aufgabe `Text-Generierung` hat eine [`~generation_utils.GenerationMixin.generate`]-Methode mit mehreren Parametern zur Steuerung der Ausgabe. Wenn Sie zum Beispiel mehr als eine Ausgabe erzeugen wollen, setzen Sie den Parameter `num_return_sequences`:
+
+```py
+>>> generator(
+...     "Three Rings for the Elven-kings under the sky, Seven for the Dwarf-lords in their halls of stone",
+...     num_return_sequences=2,
+... )  # doctest: +SKIP
+```
+
+### Wählen Sie ein Modell und einen Tokenizer
+
+Die [`pipeline`] akzeptiert jedes Modell aus dem [Hub] (https://huggingface.co/models). Auf dem Hub gibt es Tags, mit denen Sie nach einem Modell filtern können, das Sie für Ihre Aufgabe verwenden möchten. Sobald Sie ein passendes Modell ausgewählt haben, laden Sie es mit der entsprechenden `AutoModelFor` und [`AutoTokenizer`] Klasse. Laden Sie zum Beispiel die Klasse [`AutoModelForCausalLM`] für eine kausale Sprachmodellierungsaufgabe:
+
+```py
+>>> from transformers import AutoTokenizer, AutoModelForCausalLM
+
+>>> tokenizer = AutoTokenizer.from_pretrained("distilgpt2")
+>>> model = AutoModelForCausalLM.from_pretrained("distilgpt2")
+```
+
+Erstellen Sie eine [`pipeline`] für Ihre Aufgabe, und geben Sie das Modell und den Tokenizer an, die Sie geladen haben:
+
+```py
+>>> from transformers import pipeline
+
+>>> generator = pipeline(task="text-generation", model=model, tokenizer=tokenizer)
+```
+
+Übergeben Sie Ihren Eingabetext an die [`pipeline`] , um einen Text zu erzeugen:
+
+```py
+>>> generator(
+...     "Three Rings for the Elven-kings under the sky, Seven for the Dwarf-lords in their halls of stone"
+... )  # doctest: +SKIP
+[{'generated_text': 'Three Rings for the Elven-kings under the sky, Seven for the Dwarf-lords in their halls of stone, Seven for the Dragon-lords (for them to rule in a world ruled by their rulers, and all who live within the realm'}]
+```
+
+## Audio-Pipeline
+
+Die [`pipeline`] unterstützt auch Audioaufgaben wie Audioklassifizierung und automatische Spracherkennung.
+
+Lassen Sie uns zum Beispiel die Emotion in diesem Audioclip klassifizieren:
+
+```py
+>>> from datasets import load_dataset
+>>> import torch
+
+>>> torch.manual_seed(42)  # doctest: +IGNORE_RESULT
+>>> ds = load_dataset("hf-internal-testing/librispeech_asr_demo", "clean", split="validation")
+>>> audio_file = ds[0]["audio"]["path"]
+```
+
+Finden Sie ein [Audioklassifikation](https://huggingface.co/models?pipeline_tag=audio-classification) Modell auf dem Model Hub für Emotionserkennung und laden Sie es in die [`pipeline`]:
+
+```py
+>>> from transformers import pipeline
+
+>>> audio_classifier = pipeline(
+...     task="audio-classification", model="ehcalabres/wav2vec2-lg-xlsr-en-speech-emotion-recognition"
+... )
+```
+
+Übergeben Sie die Audiodatei an die [`pipeline`]:
+
+```py
+>>> preds = audio_classifier(audio_file)
+>>> preds = [{"score": round(pred["score"], 4), "label": pred["label"]} for pred in preds]
+>>> preds
+[{'score': 0.1315, 'label': 'calm'}, {'score': 0.1307, 'label': 'neutral'}, {'score': 0.1274, 'label': 'sad'}, {'score': 0.1261, 'label': 'fearful'}, {'score': 0.1242, 'label': 'happy'}]
+```
+
+## Bildverarbeitungs-Pipeline
+
+Die Verwendung einer [`pipeline`] für Bildverarbeitungsaufgaben ist praktisch identisch.
+
+Geben Sie Ihre Aufgabe an und übergeben Sie Ihr Bild an den Klassifikator. Das Bild kann ein Link oder ein lokaler Pfad zu dem Bild sein. Zum Beispiel: Welche Katzenart ist unten abgebildet?
+
+![pipeline-cat-chonk](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/pipeline-cat-chonk.jpeg)
+
+```py
+>>> from transformers import pipeline
+
+>>> vision_classifier = pipeline(task="image-classification")
+>>> preds = vision_classifier(
+...     images="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/pipeline-cat-chonk.jpeg"
+... )
+>>> preds = [{"score": round(pred["score"], 4), "label": pred["label"]} for pred in preds]
+>>> preds
+[{'score': 0.4335, 'label': 'lynx, catamount'}, {'score': 0.0348, 'label': 'cougar, puma, catamount, mountain lion, painter, panther, Felis concolor'}, {'score': 0.0324, 'label': 'snow leopard, ounce, Panthera uncia'}, {'score': 0.0239, 'label': 'Egyptian cat'}, {'score': 0.0229, 'label': 'tiger cat'}]
+```
+
+## Multimodale Pipeline
+
+Die [`pipeline`] unterstützt mehr als eine Modalität. Eine Aufgabe zur Beantwortung visueller Fragen (VQA) kombiniert zum Beispiel Text und Bild. Verwenden Sie einen beliebigen Bildlink und eine Frage, die Sie zu dem Bild stellen möchten. Das Bild kann eine URL oder ein lokaler Pfad zu dem Bild sein.
+
+Wenn Sie zum Beispiel das gleiche Bild wie in der obigen Vision-Pipeline verwenden:
+
+```py
+>>> image = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/pipeline-cat-chonk.jpeg"
+>>> question = "Where is the cat?"
+```
+
+Erstellen Sie eine Pipeline für "vqa" und übergeben Sie ihr das Bild und die Frage:
+
+```py
+>>> from transformers import pipeline
+
+>>> vqa = pipeline(task="vqa")
+>>> preds = vqa(image=image, question=question)
+>>> preds = [{"score": round(pred["score"], 4), "answer": pred["answer"]} for pred in preds]
+>>> preds
+[{'score': 0.9112, 'answer': 'snow'}, {'score': 0.8796, 'answer': 'in snow'}, {'score': 0.6717, 'answer': 'outside'}, {'score': 0.0291, 'answer': 'on ground'}, {'score': 0.027, 'answer': 'ground'}]
+```

docs/source/de/preprocessing.mdx

@@ -0,0 +1,502 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# Vorverarbeiten
+
+[[open-in-colab]]
+
+Bevor Sie Ihre Daten in einem Modell verwenden können, müssen die Daten in ein für das Modell akzeptables Format gebracht werden. Ein Modell versteht keine Rohtexte, Bilder oder Audiodaten. Diese Eingaben müssen in Zahlen umgewandelt und zu Tensoren zusammengesetzt werden. In dieser Anleitung werden Sie:
+
+* Textdaten mit einem Tokenizer vorverarbeiten.
+* Bild- oder Audiodaten mit einem Feature Extractor vorverarbeiten.
+* Daten für eine multimodale Aufgabe mit einem Prozessor vorverarbeiten.
+
+## NLP
+
+<Youtube id="Yffk5aydLzg"/>
+
+Das wichtigste Werkzeug zur Verarbeitung von Textdaten ist ein [Tokenizer](main_classes/tokenizer). Ein Tokenizer zerlegt Text zunächst nach einer Reihe von Regeln in *Token*. Die Token werden in Zahlen umgewandelt, die zum Aufbau von Tensoren als Eingabe für ein Modell verwendet werden. Alle zusätzlichen Eingaben, die ein Modell benötigt, werden ebenfalls vom Tokenizer hinzugefügt.
+
+<Tip>
+
+Wenn Sie ein vortrainiertes Modell verwenden möchten, ist es wichtig, den zugehörigen vortrainierten Tokenizer zu verwenden. Dadurch wird sichergestellt, dass der Text auf die gleiche Weise aufgeteilt wird wie das Pretraining-Korpus und die gleichen entsprechenden Token-zu-Index (in der Regel als *vocab* bezeichnet) während des Pretrainings verwendet werden.
+
+</Tip>
+
+Laden Sie einen vortrainierten Tokenizer mit der Klasse [AutoTokenizer], um schnell loszulegen. Damit wird das *vocab* heruntergeladen, das verwendet wird, wenn ein Modell vortrainiert wird.
+
+### Tokenize
+
+Laden Sie einen vortrainierten Tokenizer mit [`AutoTokenizer.from_pretrained`]:
+
+```py
+>>> from transformers import AutoTokenizer
+
+>>> tokenizer = AutoTokenizer.from_pretrained("bert-base-cased")
+```
+
+Dann übergeben Sie Ihren Satz an den Tokenizer:
+
+```py
+>>> encoded_input = tokenizer("Do not meddle in the affairs of wizards, for they are subtle and quick to anger.")
+>>> print(encoded_input)
+{'input_ids': [101, 2079, 2025, 19960, 10362, 1999, 1996, 3821, 1997, 16657, 1010, 2005, 2027, 2024, 11259, 1998, 4248, 2000, 4963, 1012, 102], 
+ 'token_type_ids': [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], 
+ 'attention_mask': [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]}
+```
+
+Der Tokenizer gibt ein Wörterbuch mit drei wichtigen Elementen zurück:
+
+* [input_ids](glossary#input-ids) sind die Indizes, die den einzelnen Token im Satz entsprechen.
+* [attention_mask](glossary#attention-mask) gibt an, ob ein Token beachtet werden soll oder nicht.
+* [token_type_ids](glossary#token-type-ids) gibt an, zu welcher Sequenz ein Token gehört, wenn es mehr als eine Sequenz gibt.
+
+Sie können die `input_ids` dekodieren, um die ursprüngliche Eingabe zurückzugeben:
+
+```py
+>>> tokenizer.decode(encoded_input["input_ids"])
+'[CLS] Do not meddle in the affairs of wizards, for they are subtle and quick to anger. [SEP]'
+```
+
+Wie Sie sehen können, hat der Tokenisierer zwei spezielle Token - `CLS` und `SEP` (Klassifikator und Separator) - zum Satz hinzugefügt. Nicht alle Modelle benötigen
+spezielle Token, aber wenn dies der Fall ist, fügt der Tokenisierer sie automatisch für Sie hinzu.
+
+Wenn Sie mehrere Sätze verarbeiten wollen, übergeben Sie die Sätze als Liste an den Tokenizer:
+
+```py
+>>> batch_sentences = [
+...     "But what about second breakfast?",
+...     "Don't think he knows about second breakfast, Pip.",
+...     "What about elevensies?",
+... ]
+>>> encoded_inputs = tokenizer(batch_sentences)
+>>> print(encoded_inputs)
+{'input_ids': [[101, 1252, 1184, 1164, 1248, 6462, 136, 102], 
+               [101, 1790, 112, 189, 1341, 1119, 3520, 1164, 1248, 6462, 117, 21902, 1643, 119, 102], 
+               [101, 1327, 1164, 5450, 23434, 136, 102]], 
+ 'token_type_ids': [[0, 0, 0, 0, 0, 0, 0, 0], 
+                    [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], 
+                    [0, 0, 0, 0, 0, 0, 0]], 
+ 'attention_mask': [[1, 1, 1, 1, 1, 1, 1, 1], 
+                    [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], 
+                    [1, 1, 1, 1, 1, 1, 1]]}
+```
+
+### Pad
+
+Dies bringt uns zu einem wichtigen Thema. Wenn Sie einen Haufen von Sätzen verarbeiten, sind diese nicht immer gleich lang. Das ist ein Problem, weil Tensoren, die Eingabe für das Modell, eine einheitliche Form haben müssen. Padding ist eine Strategie, die sicherstellt, dass Tensoren rechteckig sind, indem ein spezielles *Padding-Token* zu Sätzen mit weniger Token hinzugefügt wird.
+
+Setzen Sie den Parameter "padding" auf "true", um die kürzeren Sequenzen im Stapel so aufzufüllen, dass sie der längsten Sequenz entsprechen:
+
+```py
+>>> batch_sentences = [
+...     "But what about second breakfast?",
+...     "Don't think he knows about second breakfast, Pip.",
+...     "What about elevensies?",
+... ]
+>>> encoded_input = tokenizer(batch_sentences, padding=True)
+>>> print(encoded_input)
+{'input_ids': [[101, 1252, 1184, 1164, 1248, 6462, 136, 102, 0, 0, 0, 0, 0, 0, 0], 
+               [101, 1790, 112, 189, 1341, 1119, 3520, 1164, 1248, 6462, 117, 21902, 1643, 119, 102], 
+               [101, 1327, 1164, 5450, 23434, 136, 102, 0, 0, 0, 0, 0, 0, 0, 0]], 
+ 'token_type_ids': [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], 
+                    [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], 
+                    [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], 
+ 'attention_mask': [[1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0], 
+                    [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], 
+                    [1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0]]}
+```
+
+Beachten Sie, dass der Tokenizer den ersten und den dritten Satz mit einer "0" aufgefüllt hat, weil sie kürzer sind!
+
+### Kürzung
+
+Auf der anderen Seite des Spektrums kann es vorkommen, dass eine Sequenz zu lang für ein Modell ist. In diesem Fall müssen Sie die Sequenz auf eine kürzere Länge kürzen.
+
+Setzen Sie den Parameter "truncation" auf "true", um eine Sequenz auf die vom Modell akzeptierte Höchstlänge zu kürzen:
+
+```py
+>>> batch_sentences = [
+...     "But what about second breakfast?",
+...     "Don't think he knows about second breakfast, Pip.",
+...     "What about elevensies?",
+... ]
+>>> encoded_input = tokenizer(batch_sentences, padding=True, truncation=True)
+>>> print(encoded_input)
+{'input_ids': [[101, 1252, 1184, 1164, 1248, 6462, 136, 102, 0, 0, 0, 0, 0, 0, 0], 
+               [101, 1790, 112, 189, 1341, 1119, 3520, 1164, 1248, 6462, 117, 21902, 1643, 119, 102], 
+               [101, 1327, 1164, 5450, 23434, 136, 102, 0, 0, 0, 0, 0, 0, 0, 0]], 
+ 'token_type_ids': [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], 
+                    [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], 
+                    [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], 
+ 'attention_mask': [[1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0], 
+                    [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1], 
+                    [1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0]]}
+```
+
+### Tensoren erstellen
+
+Schließlich möchten Sie, dass der Tokenizer die tatsächlichen Tensoren zurückgibt, die dem Modell zugeführt werden.
+
+Setzen Sie den Parameter `return_tensors` entweder auf `pt` für PyTorch, oder `tf` für TensorFlow:
+
+<frameworkcontent>
+<pt>
+
+```py
+>>> batch_sentences = [
+...     "But what about second breakfast?",
+...     "Don't think he knows about second breakfast, Pip.",
+...     "What about elevensies?",
+... ]
+>>> encoded_input = tokenizer(batch_sentences, padding=True, truncation=True, return_tensors="pt")
+>>> print(encoded_input)
+{'input_ids': tensor([[101, 1252, 1184, 1164, 1248, 6462, 136, 102, 0, 0, 0, 0, 0, 0, 0],
+                      [101, 1790, 112, 189, 1341, 1119, 3520, 1164, 1248, 6462, 117, 21902, 1643, 119, 102],
+                      [101, 1327, 1164, 5450, 23434, 136, 102, 0, 0, 0, 0, 0, 0, 0, 0]]), 
+ 'token_type_ids': tensor([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+                           [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+                           [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]), 
+ 'attention_mask': tensor([[1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0],
+                           [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
+                           [1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0]])}
+```
+</pt>
+<tf>
+```py
+>>> batch_sentences = [
+...     "But what about second breakfast?",
+...     "Don't think he knows about second breakfast, Pip.",
+...     "What about elevensies?",
+... ]
+>>> encoded_input = tokenizer(batch_sentences, padding=True, truncation=True, return_tensors="tf")
+>>> print(encoded_input)
+{'input_ids': <tf.Tensor: shape=(2, 9), dtype=int32, numpy=
+array([[101, 1252, 1184, 1164, 1248, 6462, 136, 102, 0, 0, 0, 0, 0, 0, 0],
+       [101, 1790, 112, 189, 1341, 1119, 3520, 1164, 1248, 6462, 117, 21902, 1643, 119, 102],
+       [101, 1327, 1164, 5450, 23434, 136, 102, 0, 0, 0, 0, 0, 0, 0, 0]],
+      dtype=int32)>, 
+ 'token_type_ids': <tf.Tensor: shape=(2, 9), dtype=int32, numpy=
+array([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+       [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], dtype=int32)>, 
+ 'attention_mask': <tf.Tensor: shape=(2, 9), dtype=int32, numpy=
+array([[1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0],
+       [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
+       [1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0]], dtype=int32)>}
+```
+</tf>
+</frameworkcontent>
+
+## Audio
+
+Audioeingaben werden anders vorverarbeitet als Texteingaben, aber das Endziel bleibt dasselbe: numerische Sequenzen zu erstellen, die das Modell verstehen kann. Ein [feature extractor](main_classes/feature_extractor) dient dem ausdrücklichen Zweck, Merkmale aus Rohbild- oder Audiodaten zu extrahieren und in Tensoren zu konvertieren. Bevor Sie beginnen, installieren Sie 🤗 Datasets, um einen Audio-Datensatz zu laden, mit dem Sie experimentieren können:
+
+```bash
+pip install datasets
+```
+
+Laden Sie den [MInDS-14](https://huggingface.co/datasets/PolyAI/minds14) Datensatz (weitere Informationen zum Laden eines Datensatzes finden Sie im 🤗 [Datasets tutorial](https://huggingface.co/docs/datasets/load_hub.html)):
+
+```py
+>>> from datasets import load_dataset, Audio
+
+>>> dataset = load_dataset("PolyAI/minds14", name="en-US", split="train")
+```
+
+Greifen Sie auf das erste Element der `audio`-Spalte zu, um einen Blick auf die Eingabe zu werfen. Durch den Aufruf der Spalte "audio" wird die Audiodatei automatisch geladen und neu gesampelt:
+
+```py
+>>> dataset[0]["audio"]
+{'array': array([ 0.        ,  0.00024414, -0.00024414, ..., -0.00024414,
+         0.        ,  0.        ], dtype=float32),
+ 'path': '/root/.cache/huggingface/datasets/downloads/extracted/f14948e0e84be638dd7943ac36518a4cf3324e8b7aa331c5ab11541518e9368c/en-US~JOINT_ACCOUNT/602ba55abb1e6d0fbce92065.wav',
+ 'sampling_rate': 8000}
+```
+
+Dies gibt drei Elemente zurück:
+
+* "array" ist das Sprachsignal, das als 1D-Array geladen - und möglicherweise neu gesampelt - wurde.
+* Pfad" zeigt auf den Speicherort der Audiodatei.
+* `sampling_rate` bezieht sich darauf, wie viele Datenpunkte im Sprachsignal pro Sekunde gemessen werden.
+
+### Resample
+
+Für dieses Tutorial werden Sie das Modell [Wav2Vec2](https://huggingface.co/facebook/wav2vec2-base) verwenden. Wie Sie aus der Modellkarte ersehen können, ist das Wav2Vec2-Modell auf 16kHz abgetastetes Sprachaudio vortrainiert. Es ist wichtig, dass die Abtastrate Ihrer Audiodaten mit der Abtastrate des Datensatzes übereinstimmt, der für das Pre-Training des Modells verwendet wurde. Wenn die Abtastrate Ihrer Daten nicht dieselbe ist, müssen Sie Ihre Audiodaten neu abtasten. 
+
+Der Datensatz [MInDS-14](https://huggingface.co/datasets/PolyAI/minds14) hat zum Beispiel eine Abtastrate von 8000 kHz. Um das Wav2Vec2-Modell mit diesem Datensatz verwenden zu können, müssen Sie die Abtastrate auf 16 kHz erhöhen:
+
+```py
+>>> dataset = load_dataset("PolyAI/minds14", name="en-US", split="train")
+>>> dataset[0]["audio"]
+{'array': array([ 0.        ,  0.00024414, -0.00024414, ..., -0.00024414,
+         0.        ,  0.        ], dtype=float32),
+ 'path': '/root/.cache/huggingface/datasets/downloads/extracted/f14948e0e84be638dd7943ac36518a4cf3324e8b7aa331c5ab11541518e9368c/en-US~JOINT_ACCOUNT/602ba55abb1e6d0fbce92065.wav',
+ 'sampling_rate': 8000}
+```
+
+1. Verwenden Sie die Methode [~datasets.Dataset.cast_column] von 🤗 Datasets, um die Abtastrate auf 16kHz zu erhöhen:
+
+```py
+>>> dataset = dataset.cast_column("audio", Audio(sampling_rate=16_000))
+```
+
+2. Laden Sie die Audiodatei:
+
+```py
+>>> dataset[0]["audio"]
+{'array': array([ 2.3443763e-05,  2.1729663e-04,  2.2145823e-04, ...,
+         3.8356509e-05, -7.3497440e-06, -2.1754686e-05], dtype=float32),
+ 'path': '/root/.cache/huggingface/datasets/downloads/extracted/f14948e0e84be638dd7943ac36518a4cf3324e8b7aa331c5ab11541518e9368c/en-US~JOINT_ACCOUNT/602ba55abb1e6d0fbce92065.wav',
+ 'sampling_rate': 16000}
+```
+
+Wie Sie sehen können, ist die Abtastrate jetzt 16kHz!
+
+### Merkmalsextraktor
+
+Der nächste Schritt ist das Laden eines Merkmalsextraktors, um die Eingabe zu normalisieren und aufzufüllen. Beim Auffüllen von Textdaten wird für kürzere Sequenzen ein `0` hinzugefügt. Die gleiche Idee gilt für Audiodaten, und der Audio-Feature-Extraktor fügt eine `0` - interpretiert als Stille - zu `array` hinzu.
+
+Laden Sie den Merkmalsextraktor mit [`AutoFeatureExtractor.from_pretrained`]:
+
+```py
+>>> from transformers import AutoFeatureExtractor
+
+>>> feature_extractor = AutoFeatureExtractor.from_pretrained("facebook/wav2vec2-base")
+```
+
+Übergeben Sie das Audio-"Array" an den Feature-Extraktor. Wir empfehlen auch, das Argument `sampling_rate` im Feature Extractor hinzuzufügen, um eventuell auftretende stille Fehler besser zu beheben.
+
+```py
+>>> audio_input = [dataset[0]["audio"]["array"]]
+>>> feature_extractor(audio_input, sampling_rate=16000)
+{'input_values': [array([ 3.8106556e-04,  2.7506407e-03,  2.8015103e-03, ...,
+        5.6335266e-04,  4.6588284e-06, -1.7142107e-04], dtype=float32)]}
+```
+
+### Auffüllen und Kürzen
+
+Genau wie beim Tokenizer können Sie variable Sequenzen in einem Stapel durch Auffüllen oder Abschneiden behandeln. Werfen Sie einen Blick auf die Sequenzlänge dieser beiden Audiobeispiele:
+
+```py
+>>> dataset[0]["audio"]["array"].shape
+(173398,)
+
+>>> dataset[1]["audio"]["array"].shape
+(106496,)
+```
+
+Wie Sie sehen können, hat das erste Beispiel eine längere Sequenz als das zweite Beispiel. Lassen Sie uns eine Funktion erstellen, die den Datensatz vorverarbeitet. Geben Sie eine maximale Länge der Probe an, und der Feature-Extraktor wird die Sequenzen entweder auffüllen oder abschneiden, damit sie dieser Länge entsprechen:
+
+```py
+>>> def preprocess_function(examples):
+...     audio_arrays = [x["array"] for x in examples["audio"]]
+...     inputs = feature_extractor(
+...         audio_arrays,
+...         sampling_rate=16000,
+...         padding=True,
+...         max_length=100000,
+...         truncation=True,
+...     )
+...     return inputs
+```
+
+Wenden Sie die Funktion auf die ersten paar Beispiele im Datensatz an:
+
+```py
+>>> processed_dataset = preprocess_function(dataset[:5])
+```
+
+Schauen Sie sich nun noch einmal die verarbeiteten Beispiel-Längen an:
+
+```py
+>>> processed_dataset["input_values"][0].shape
+(100000,)
+
+>>> processed_dataset["input_values"][1].shape
+(100000,)
+```
+
+Die Länge der ersten beiden Beispiele entspricht nun der von Ihnen angegebenen Maximallänge.
+
+## Bildverarbeitung
+
+Ein Merkmalsextraktor wird auch verwendet, um Bilder für Bildverarbeitungsaufgaben zu verarbeiten. Auch hier besteht das Ziel darin, das Rohbild in eine Reihe von Tensoren als Eingabe zu konvertieren.
+
+Laden wir den [food101](https://huggingface.co/datasets/food101) Datensatz für dieses Tutorial. Verwenden Sie den Parameter 🤗 Datasets `split`, um nur eine kleine Stichprobe aus dem Trainingssplit zu laden, da der Datensatz recht groß ist:
+
+```py
+>>> from datasets import load_dataset
+
+>>> dataset = load_dataset("food101", split="train[:100]")
+```
+
+Als Nächstes sehen Sie sich das Bild mit dem Merkmal 🤗 Datensätze [Bild] (https://huggingface.co/docs/datasets/package_reference/main_classes.html?highlight=image#datasets.Image) an:
+
+```py
+>>> dataset[0]["image"]
+```
+
+![vision-preprocess-tutorial.png](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/vision-preprocess-tutorial.png)
+
+### Merkmalsextraktor
+
+Laden Sie den Merkmalsextraktor mit [`AutoFeatureExtractor.from_pretrained`]:
+
+```py
+>>> from transformers import AutoFeatureExtractor
+
+>>> feature_extractor = AutoFeatureExtractor.from_pretrained("google/vit-base-patch16-224")
+```
+
+### Datenerweiterung
+
+Bei Bildverarbeitungsaufgaben ist es üblich, den Bildern als Teil der Vorverarbeitung eine Art von Datenerweiterung hinzuzufügen. Sie können Erweiterungen mit jeder beliebigen Bibliothek hinzufügen, aber in diesem Tutorial werden Sie das Modul [`transforms`](https://pytorch.org/vision/stable/transforms.html) von torchvision verwenden.
+
+1. Normalisieren Sie das Bild und verwenden Sie [`Compose`](https://pytorch.org/vision/master/generated/torchvision.transforms.Compose.html), um einige Transformationen - [`RandomResizedCrop`](https://pytorch.org/vision/main/generated/torchvision.transforms.RandomResizedCrop.html) und [`ColorJitter`](https://pytorch.org/vision/main/generated/torchvision.transforms.ColorJitter.html) - miteinander zu verknüpfen:
+
+```py
+>>> from torchvision.transforms import Compose, Normalize, RandomResizedCrop, ColorJitter, ToTensor
+
+>>> normalize = Normalize(mean=feature_extractor.image_mean, std=feature_extractor.image_std)
+>>> _transforms = Compose(
+...     [RandomResizedCrop(feature_extractor.size), ColorJitter(brightness=0.5, hue=0.5), ToTensor(), normalize]
+... )
+```
+
+2. Das Modell akzeptiert [`pixel_values`](model_doc/visionencoderdecoder#transformers.VisionEncoderDecoderModel.forward.pixel_values) als Eingabe. Dieser Wert wird vom Merkmalsextraktor erzeugt. Erstellen Sie eine Funktion, die `pixel_values` aus den Transformationen erzeugt:
+
+```py
+>>> def transforms(examples):
+...     examples["pixel_values"] = [_transforms(image.convert("RGB")) for image in examples["image"]]
+...     return examples
+```
+
+3. Dann verwenden Sie 🤗 Datasets [`set_transform`](https://huggingface.co/docs/datasets/process.html#format-transform), um die Transformationen im laufenden Betrieb anzuwenden:
+
+```py
+>>> dataset.set_transform(transforms)
+```
+
+4. Wenn Sie nun auf das Bild zugreifen, werden Sie feststellen, dass der Feature Extractor die Modelleingabe "pixel_values" hinzugefügt hat:
+
+```py
+>>> dataset[0]["image"]
+{'image': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=384x512 at 0x7F1A7B0630D0>,
+ 'label': 6,
+ 'pixel_values': tensor([[[ 0.0353,  0.0745,  0.1216,  ..., -0.9922, -0.9922, -0.9922],
+          [-0.0196,  0.0667,  0.1294,  ..., -0.9765, -0.9843, -0.9922],
+          [ 0.0196,  0.0824,  0.1137,  ..., -0.9765, -0.9686, -0.8667],
+          ...,
+          [ 0.0275,  0.0745,  0.0510,  ..., -0.1137, -0.1216, -0.0824],
+          [ 0.0667,  0.0824,  0.0667,  ..., -0.0588, -0.0745, -0.0980],
+          [ 0.0353,  0.0353,  0.0431,  ..., -0.0039, -0.0039, -0.0588]],
+ 
+         [[ 0.2078,  0.2471,  0.2863,  ..., -0.9451, -0.9373, -0.9451],
+          [ 0.1608,  0.2471,  0.3098,  ..., -0.9373, -0.9451, -0.9373],
+          [ 0.2078,  0.2706,  0.3020,  ..., -0.9608, -0.9373, -0.8275],
+          ...,
+          [-0.0353,  0.0118, -0.0039,  ..., -0.2392, -0.2471, -0.2078],
+          [ 0.0196,  0.0353,  0.0196,  ..., -0.1843, -0.2000, -0.2235],
+          [-0.0118, -0.0039, -0.0039,  ..., -0.0980, -0.0980, -0.1529]],
+ 
+         [[ 0.3961,  0.4431,  0.4980,  ..., -0.9216, -0.9137, -0.9216],
+          [ 0.3569,  0.4510,  0.5216,  ..., -0.9059, -0.9137, -0.9137],
+          [ 0.4118,  0.4745,  0.5216,  ..., -0.9137, -0.8902, -0.7804],
+          ...,
+          [-0.2314, -0.1922, -0.2078,  ..., -0.4196, -0.4275, -0.3882],
+          [-0.1843, -0.1686, -0.2000,  ..., -0.3647, -0.3804, -0.4039],
+          [-0.1922, -0.1922, -0.1922,  ..., -0.2941, -0.2863, -0.3412]]])}
+```
+
+Hier sehen Sie, wie das Bild nach der Vorverarbeitung aussieht. Wie von den angewandten Transformationen zu erwarten, wurde das Bild willkürlich beschnitten und seine Farbeigenschaften sind anders.
+
+```py
+>>> import numpy as np
+>>> import matplotlib.pyplot as plt
+
+>>> img = dataset[0]["pixel_values"]
+>>> plt.imshow(img.permute(1, 2, 0))
+```
+
+![preprocessed_image](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/preprocessed_image.png)
+
+## Multimodal
+
+Für multimodale Aufgaben werden Sie eine Kombination aus allem, was Sie bisher gelernt haben, verwenden und Ihre Fähigkeiten auf eine Aufgabe der automatischen Spracherkennung (ASR) anwenden. Dies bedeutet, dass Sie einen:
+
+* Feature Extractor zur Vorverarbeitung der Audiodaten.
+* Tokenizer, um den Text zu verarbeiten.
+
+Kehren wir zum [LJ Speech](https://huggingface.co/datasets/lj_speech) Datensatz zurück:
+
+```py
+>>> from datasets import load_dataset
+
+>>> lj_speech = load_dataset("lj_speech", split="train")
+```
+
+Da Sie hauptsächlich an den Spalten "Audio" und "Text" interessiert sind, entfernen Sie die anderen Spalten:
+
+```py
+>>> lj_speech = lj_speech.map(remove_columns=["file", "id", "normalized_text"])
+```
+
+Schauen Sie sich nun die Spalten "Audio" und "Text" an:
+
+```py
+>>> lj_speech[0]["audio"]
+{'array': array([-7.3242188e-04, -7.6293945e-04, -6.4086914e-04, ...,
+         7.3242188e-04,  2.1362305e-04,  6.1035156e-05], dtype=float32),
+ 'path': '/root/.cache/huggingface/datasets/downloads/extracted/917ece08c95cf0c4115e45294e3cd0dee724a1165b7fc11798369308a465bd26/LJSpeech-1.1/wavs/LJ001-0001.wav',
+ 'sampling_rate': 22050}
+
+>>> lj_speech[0]["text"]
+'Printing, in the only sense with which we are at present concerned, differs from most if not from all the arts and crafts represented in the Exhibition'
+```
+
+Erinnern Sie sich an den früheren Abschnitt über die Verarbeitung von Audiodaten: Sie sollten immer die Abtastrate Ihrer Audiodaten [resample](preprocessing#audio), damit sie mit der Abtastrate des Datensatzes übereinstimmt, der für das Vortraining eines Modells verwendet wird:
+
+```py
+>>> lj_speech = lj_speech.cast_column("audio", Audio(sampling_rate=16_000))
+```
+
+### Prozessor
+
+Ein Processor kombiniert einen Feature-Extraktor und einen Tokenizer. Laden Sie einen Processor mit [`AutoProcessor.from_pretrained]:
+
+```py
+>>> from transformers import AutoProcessor
+
+>>> processor = AutoProcessor.from_pretrained("facebook/wav2vec2-base-960h")
+```
+
+1. Erstellen Sie eine Funktion, die die Audiodaten zu `input_values` verarbeitet und den Text zu `labels` tokenisiert. Dies sind Ihre Eingaben für das Modell:
+
+```py
+>>> def prepare_dataset(example):
+...     audio = example["audio"]
+
+...     example.update(processor(audio=audio["array"], text=example["text"], sampling_rate=16000))
+
+...     return example
+```
+
+2. Wenden Sie die Funktion "prepare_dataset" auf ein Beispiel an:
+
+```py
+>>> prepare_dataset(lj_speech[0])
+```
+
+Beachten Sie, dass der Processor `input_values` und `labels` hinzugefügt hat. Auch die Abtastrate wurde korrekt auf 16kHz heruntergerechnet.
+
+Toll, Sie sollten jetzt in der Lage sein, Daten für jede Modalität vorzuverarbeiten und sogar verschiedene Modalitäten zu kombinieren! Im nächsten Kurs lernen Sie, wie Sie ein Modell mit Ihren neu aufbereiteten Daten feinabstimmen können.

docs/source/de/quicktour.mdx

@@ -0,0 +1,428 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# Schnellstart
+
+[[open-in-colab]]
+
+Mit 🤗 Transformers können Sie sofort loslegen! Verwenden Sie die [`pipeline`] für schnelle Inferenz und laden Sie schnell ein vortrainiertes Modell und einen Tokenizer mit einer [AutoClass](./model_doc/auto), um Ihre Text-, Bild- oder Audioaufgabe zu lösen.
+
+<Tip>
+
+Alle in der Dokumentation vorgestellten Codebeispiele haben oben links einen Umschalter für PyTorch und TensorFlow. Wenn
+nicht, wird erwartet, dass der Code für beide Backends ohne Änderungen funktioniert.
+
+</Tip>
+
+## Pipeline
+
+[`pipeline`] ist der einfachste Weg, ein vortrainiertes Modell für eine bestimmte Aufgabe zu verwenden.
+
+<Youtube id="tiZFewofSLM"/>
+
+Die [`pipeline`] unterstützt viele gängige Aufgaben:
+
+**Text**:
+* Stimmungsanalyse: Klassifizierung der Polarität eines gegebenen Textes.
+* Textgenerierung (auf Englisch): Generierung von Text aus einer gegebenen Eingabe.
+* Name-Entity-Recognition (NER): Kennzeichnung jedes Worts mit der Entität, die es repräsentiert (Person, Datum, Ort usw.).
+* Beantwortung von Fragen: Extrahieren der Antwort aus dem Kontext, wenn ein gewisser Kontext und eine Frage gegeben sind.
+* Fill-mask: Ausfüllen von Lücken in einem Text mit maskierten Wörtern.
+* Zusammenfassung: Erstellung einer Zusammenfassung einer langen Text- oder Dokumentensequenz.
+* Übersetzung: Übersetzen eines Textes in eine andere Sprache.
+* Merkmalsextraktion: Erstellen einer Tensordarstellung des Textes.
+
+**Bild**:
+* Bildklassifizierung: Klassifizierung eines Bildes.
+* Bildsegmentierung: Klassifizierung jedes Pixels in einem Bild.
+* Objekterkennung: Erkennen von Objekten innerhalb eines Bildes.
+
+**Audio**:
+* Audioklassifizierung: Zuweisung eines Labels zu einem bestimmten Audiosegment.
+* Automatische Spracherkennung (ASR): Transkription von Audiodaten in Text.
+
+<Tip>
+
+Für mehr Details über die [`pipeline`] und assoziierte Aufgaben, schauen Sie in die Dokumentation [hier](./main_classes/pipelines).
+
+</Tip>
+
+### Verwendung der Pipeline
+
+Im folgenden Beispiel werden Sie die [`pipeline`] für die Stimmungsanalyse verwenden.
+
+Installieren Sie die folgenden Abhängigkeiten, falls Sie dies nicht bereits getan haben:
+
+<frameworkcontent>
+<pt>
+```bash
+pip install torch
+```
+</pt>
+<tf>
+```bash
+pip install tensorflow
+```
+</tf>
+</frameworkcontent>
+
+Importieren sie die [`pipeline`] und spezifizieren sie die Aufgabe, welche sie lösen möchten:
+
+```py
+>>> from transformers import pipeline
+
+>>> classifier = pipeline("sentiment-analysis")
+```
+
+Die Pipeline lädt ein standardmäßiges [vortrainiertes Modell] (https://huggingface.co/distilbert-base-uncased-finetuned-sst-2-english) und einen Tokenizer für die Stimmungs-Analyse herunter und speichert sie. Jetzt können Sie den "Klassifikator" auf Ihren Zieltext anwenden:
+
+```py
+>>> classifier("We are very happy to show you the 🤗 Transformers library.")
+[{'label': 'POSITIVE', 'score': 0.9998}]
+```
+
+For more than one sentence, pass a list of sentences to the [`pipeline`] which returns a list of dictionaries:
+
+```py
+>>> results = classifier(["We are very happy to show you the 🤗 Transformers library.", "We hope you don't hate it."])
+>>> for result in results:
+...     print(f"label: {result['label']}, with score: {round(result['score'], 4)}")
+label: POSITIVE, with score: 0.9998
+label: NEGATIVE, with score: 0.5309
+```
+
+Die [`pipeline`] kann auch über einen ganzen Datensatz iterieren. Starten wir mit der Installation der [🤗 Datasets](https://huggingface.co/docs/datasets/) Bibliothek:
+
+```bash
+pip install datasets 
+```
+
+Erstellen wir eine [`pipeline`] mit der Aufgabe die wir lösen und dem Modell welches wir nutzen möchten.
+
+```py
+>>> import torch
+>>> from transformers import pipeline
+
+>>> speech_recognizer = pipeline("automatic-speech-recognition", model="facebook/wav2vec2-base-960h")
+```
+
+Als nächstes laden wir den Datensatz (siehe 🤗 Datasets [Quick Start](https://huggingface.co/docs/datasets/quickstart.html) für mehr Details) welches wir nutzen möchten. Zum Beispiel laden wir den [MInDS-14](https://huggingface.co/datasets/PolyAI/minds14) Datensatz:
+
+```py
+>>> from datasets import load_dataset, Audio
+
+>>> dataset = load_dataset("PolyAI/minds14", name="en-US", split="train")  # doctest: +IGNORE_RESULT
+```
+
+Wir müssen sicherstellen, dass die Abtastrate des Datensatzes der Abtastrate entspricht, mit der `facebook/wav2vec2-base-960h` trainiert wurde.
+
+```py
+>>> dataset = dataset.cast_column("audio", Audio(sampling_rate=speech_recognizer.feature_extractor.sampling_rate))
+```
+
+Audiodateien werden automatisch geladen und neu abgetastet, wenn die Spalte "audio" aufgerufen wird.
+Extrahieren wir die rohen Wellenform-Arrays der ersten 4 Beispiele und übergeben wir sie als Liste an die Pipeline:
+
+```py
+>>> result = speech_recognizer(dataset[:4]["audio"])
+>>> print([d["text"] for d in result])
+['I WOULD LIKE TO SET UP A JOINT ACCOUNT WITH MY PARTNER HOW DO I PROCEED WITH DOING THAT', "FODING HOW I'D SET UP A JOIN TO HET WITH MY WIFE AND WHERE THE AP MIGHT BE", "I I'D LIKE TOY SET UP A JOINT ACCOUNT WITH MY PARTNER I'M NOT SEEING THE OPTION TO DO IT ON THE AP SO I CALLED IN TO GET SOME HELP CAN I JUST DO IT OVER THE PHONE WITH YOU AND GIVE YOU THE INFORMATION OR SHOULD I DO IT IN THE AP AND I'M MISSING SOMETHING UQUETTE HAD PREFERRED TO JUST DO IT OVER THE PHONE OF POSSIBLE THINGS", 'HOW DO I THURN A JOIN A COUNT']
+```
+
+Bei einem größeren Datensatz mit vielen Eingaben (wie bei Sprache oder Bildverarbeitung) sollten Sie einen Generator anstelle einer Liste übergeben, der alle Eingaben in den Speicher lädt. Weitere Informationen finden Sie in der [Pipeline-Dokumentation](./main_classes/pipelines).
+
+### Ein anderes Modell und einen anderen Tokenizer in der Pipeline verwenden
+
+Die [`pipeline`] kann jedes Modell aus dem [Model Hub] (https://huggingface.co/models) verwenden, wodurch es einfach ist, die [`pipeline`] für andere Anwendungsfälle anzupassen. Wenn Sie beispielsweise ein Modell wünschen, das französischen Text verarbeiten kann, verwenden Sie die Tags im Model Hub, um nach einem geeigneten Modell zu filtern. Das oberste gefilterte Ergebnis liefert ein mehrsprachiges [BERT-Modell](https://huggingface.co/nlptown/bert-base-multilingual-uncased-sentiment), das auf die Stimmungsanalyse abgestimmt ist. Großartig, verwenden wir dieses Modell!
+
+```py
+>>> model_name = "nlptown/bert-base-multilingual-uncased-sentiment"
+```
+
+<frameworkcontent>
+<pt>
+Use the [`AutoModelForSequenceClassification`] and [`AutoTokenizer`] to load the pretrained model and it's associated tokenizer (more on an `AutoClass` below):
+
+```py
+>>> from transformers import AutoTokenizer, AutoModelForSequenceClassification
+
+>>> model = AutoModelForSequenceClassification.from_pretrained(model_name)
+>>> tokenizer = AutoTokenizer.from_pretrained(model_name)
+```
+</pt>
+<tf>
+Use the [`TFAutoModelForSequenceClassification`] and [`AutoTokenizer`] to load the pretrained model and it's associated tokenizer (more on an `TFAutoClass` below):
+
+```py
+>>> from transformers import AutoTokenizer, TFAutoModelForSequenceClassification
+
+>>> model = TFAutoModelForSequenceClassification.from_pretrained(model_name)
+>>> tokenizer = AutoTokenizer.from_pretrained(model_name)
+```
+</tf>
+</frameworkcontent>
+
+Dann können Sie das Modell und den Tokenizer in der [`pipeline`] angeben und den `Klassifikator` auf Ihren Zieltext anwenden:
+
+```py
+>>> classifier = pipeline("sentiment-analysis", model=model, tokenizer=tokenizer)
+>>> classifier("Nous sommes très heureux de vous présenter la bibliothèque 🤗 Transformers.")
+[{'label': '5 stars', 'score': 0.7273}]
+```
+
+Wenn Sie kein Modell für Ihren Anwendungsfall finden können, müssen Sie ein vortrainiertes Modell auf Ihren Daten feinabstimmen. Schauen Sie sich unser [Feinabstimmungs-Tutorial](./training) an, um zu erfahren, wie das geht. Und schließlich, nachdem Sie Ihr trainiertes Modell verfeinert haben, sollten Sie es mit der Community im Model Hub teilen (siehe Tutorial [hier](./model_sharing)), um NLP für alle zu demokratisieren! 🤗
+
+## AutoClass
+
+<Youtube id="AhChOFRegn4"/>
+
+Unter der Haube arbeiten die Klassen [`AutoModelForSequenceClassification`] und [`AutoTokenizer`] zusammen, um die [`pipeline`] zu betreiben. Eine [`AutoClass`](./model_doc/auto) ist eine Abkürzung, die automatisch die Architektur eines trainierten Modells aus dessen Namen oder Pfad abruft. Sie müssen nur die passende `AutoClass` für Ihre Aufgabe und den zugehörigen Tokenizer mit [`AutoTokenizer`] auswählen. 
+
+Kehren wir zu unserem Beispiel zurück und sehen wir uns an, wie Sie die `AutoClass` verwenden können, um die Ergebnisse der [`pipeline`] zu replizieren.
+
+### AutoTokenizer
+
+Ein Tokenizer ist für die Vorverarbeitung von Text in ein für das Modell verständliches Format zuständig. Zunächst zerlegt der Tokenisierer den Text in Wörter, die *Token* genannt werden. Es gibt mehrere Regeln für den Tokenisierungsprozess, z. B. wie und auf welcher Ebene ein Wort aufgespalten wird (weitere Informationen über Tokenisierung [hier](./tokenizer_summary)). Das Wichtigste ist jedoch, dass Sie den Tokenizer mit demselben Modellnamen instanziieren müssen, um sicherzustellen, dass Sie dieselben Tokenisierungsregeln verwenden, mit denen ein Modell zuvor trainiert wurde.
+Laden sie einen Tokenizer mit [`AutoTokenizer`]:
+
+```py
+>>> from transformers import AutoTokenizer
+
+>>> model_name = "nlptown/bert-base-multilingual-uncased-sentiment"
+>>> tokenizer = AutoTokenizer.from_pretrained(model_name)
+```
+
+Anschließend wandelt der Tokenizer die Token in Zahlen um, um einen Tensor als Eingabe für das Modell zu konstruieren. Dieser wird als *Vokabular* des Modells bezeichnet.
+
+Übergeben Sie Ihren Text an den Tokenizer:
+
+```py
+>>> encoding = tokenizer("We are very happy to show you the 🤗 Transformers library.")
+>>> print(encoding)
+{'input_ids': [101, 11312, 10320, 12495, 19308, 10114, 11391, 10855, 10103, 100, 58263, 13299, 119, 102],
+ 'token_type_ids': [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
+ 'attention_mask': [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]}
+```
+
+Der Tokenizer gibt ein Wörterbuch zurück, das Folgendes enthält:
+
+* [input_ids](./glossary#input-ids): numerische Repräsentationen Ihrer Token.
+* [atttention_mask](.glossary#attention-mask): gibt an, welche Token beachtet werden sollen.
+
+Genau wie die [`pipeline`] akzeptiert der Tokenizer eine Liste von Eingaben. Darüber hinaus kann der Tokenizer den Text auch auffüllen und kürzen, um einen Stapel mit einheitlicher Länge zurückzugeben:
+
+<frameworkcontent>
+<pt>
+```py
+>>> pt_batch = tokenizer(
+...     ["We are very happy to show you the 🤗 Transformers library.", "We hope you don't hate it."],
+...     padding=True,
+...     truncation=True,
+...     max_length=512,
+...     return_tensors="pt",
+... )
+```
+</pt>
+<tf>
+```py
+>>> tf_batch = tokenizer(
+...     ["We are very happy to show you the 🤗 Transformers library.", "We hope you don't hate it."],
+...     padding=True,
+...     truncation=True,
+...     max_length=512,
+...     return_tensors="tf",
+... )
+```
+</tf>
+</frameworkcontent>
+
+Lesen Sie das Tutorial [preprocessing](./preprocessing) für weitere Details zur Tokenisierung.
+
+### AutoModel
+
+<frameworkcontent>
+<pt>
+🤗 Transformers bietet eine einfache und einheitliche Möglichkeit, vortrainierte Instanzen zu laden. Das bedeutet, dass Sie ein [`AutoModel`] laden können, wie Sie einen [`AutoTokenizer`] laden würden. Der einzige Unterschied ist die Auswahl des richtigen [`AutoModel`] für die Aufgabe. Da Sie eine Text- oder Sequenzklassifizierung vornehmen, laden Sie [`AutoModelForSequenceClassification`]:
+
+```py
+>>> from transformers import AutoModelForSequenceClassification
+
+>>> model_name = "nlptown/bert-base-multilingual-uncased-sentiment"
+>>> pt_model = AutoModelForSequenceClassification.from_pretrained(model_name)
+```
+
+<Tip>
+
+In der [Aufgabenzusammenfassung](./task_summary) steht, welche [AutoModel]-Klasse für welche Aufgabe zu verwenden ist.
+
+</Tip>
+
+Jetzt können Sie Ihren vorverarbeiteten Stapel von Eingaben direkt an das Modell übergeben. Sie müssen nur das Wörterbuch entpacken, indem Sie `**` hinzufügen:
+
+```py
+>>> pt_outputs = pt_model(**pt_batch)
+```
+
+Das Modell gibt die endgültigen Aktivierungen in dem Attribut "logits" aus. Wenden Sie die Softmax-Funktion auf die "logits" an, um die Wahrscheinlichkeiten zu erhalten:
+  
+```py
+>>> from torch import nn
+
+>>> pt_predictions = nn.functional.softmax(pt_outputs.logits, dim=-1)
+>>> print(pt_predictions)
+tensor([[0.0021, 0.0018, 0.0115, 0.2121, 0.7725],
+        [0.2084, 0.1826, 0.1969, 0.1755, 0.2365]], grad_fn=<SoftmaxBackward0>)
+```
+</pt>
+<tf>
+🤗 Transformers bietet eine einfache und einheitliche Methode zum Laden von vortrainierten Instanzen. Das bedeutet, dass Sie ein [`TFAutoModel`] genauso laden können, wie Sie einen [`AutoTokenizer`] laden würden. Der einzige Unterschied ist die Auswahl des richtigen [`TFAutoModel`] für die Aufgabe. Da Sie Text - oder Sequenz - Klassifizierung machen, laden Sie [`TFAutoModelForSequenceClassification`]:
+
+```py
+>>> from transformers import TFAutoModelForSequenceClassification
+
+>>> model_name = "nlptown/bert-base-multilingual-uncased-sentiment"
+>>> tf_model = TFAutoModelForSequenceClassification.from_pretrained(model_name)
+```
+
+<Tip>
+
+In der [Aufgabenzusammenfassung](./task_summary) steht, welche [AutoModel]-Klasse für welche Aufgabe zu verwenden ist.
+
+</Tip>
+
+Jetzt können Sie Ihren vorverarbeiteten Stapel von Eingaben direkt an das Modell übergeben, indem Sie die Wörterbuchschlüssel direkt an die Tensoren übergeben:
+  
+```py
+>>> tf_outputs = tf_model(tf_batch)
+```
+
+Das Modell gibt die endgültigen Aktivierungen in dem Attribut "logits" aus. Wenden Sie die Softmax-Funktion auf die "logits" an, um die Wahrscheinlichkeiten zu erhalten:
+
+```py
+>>> import tensorflow as tf
+
+>>> tf_predictions = tf.nn.softmax(tf_outputs.logits, axis=-1)
+>>> tf_predictions  # doctest: +IGNORE_RESULT
+```
+</tf>
+</frameworkcontent>
+
+<Tip>
+
+Alle 🤗 Transformers-Modelle (PyTorch oder TensorFlow) geben die Tensoren *vor* der endgültigen Aktivierungsfunktion
+Funktion (wie Softmax) aus, da die endgültige Aktivierungsfunktion oft mit dem Verlusten verschmolzen ist.
+
+</Tip>
+
+Modelle sind ein standardmäßiges [`torch.nn.Module`](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) oder ein [`tf.keras.Model`](https://www.tensorflow.org/api_docs/python/tf/keras/Model), sodass Sie sie in Ihrer üblichen Trainingsschleife verwenden können. Um jedoch die Dinge einfacher zu machen, bietet 🤗 Transformers eine [`Trainer`]-Klasse für PyTorch, die Funktionalität für verteiltes Training, gemischte Präzision und mehr bietet. Für TensorFlow können Sie die Methode `fit` aus [Keras](https://keras.io/) verwenden. Siehe das [training tutorial](./training) für weitere Details.
+
+<Tip>
+
+Transformers-Modellausgaben sind spezielle Datenklassen, so dass ihre Attribute in einer IDE automatisch vervollständigt werden.
+Die Modellausgänge verhalten sich auch wie ein Tupel oder ein Wörterbuch (z.B. können Sie mit einem Integer, einem Slice oder einem String indexieren), wobei die Attribute, die "None" sind, ignoriert werden.
+
+</Tip>
+
+### Modell speichern
+
+<frameworkcontent>
+<pt>
+Sobald Ihr Modell feinabgestimmt ist, können Sie es mit seinem Tokenizer speichern, indem Sie [`PreTrainedModel.save_pretrained`] verwenden:
+
+```py
+>>> pt_save_directory = "./pt_save_pretrained"
+>>> tokenizer.save_pretrained(pt_save_directory)  # doctest: +IGNORE_RESULT
+>>> pt_model.save_pretrained(pt_save_directory)
+```
+
+Wenn Sie bereit sind, das Modell erneut zu verwenden, laden Sie es mit [`PreTrainedModel.from_pretrained`]:
+
+```py
+>>> pt_model = AutoModelForSequenceClassification.from_pretrained("./pt_save_pretrained")
+```
+</pt>
+<tf>
+Sobald Ihr Modell feinabgestimmt ist, können Sie es mit seinem Tokenizer unter Verwendung von [`TFPreTrainedModel.save_pretrained`] speichern:
+
+```py
+>>> tf_save_directory = "./tf_save_pretrained"
+>>> tokenizer.save_pretrained(tf_save_directory)  # doctest: +IGNORE_RESULT
+>>> tf_model.save_pretrained(tf_save_directory)
+```
+
+Wenn Sie bereit sind, das Modell wieder zu verwenden, laden Sie es mit [`TFPreTrainedModel.from_pretrained`]:
+
+```py
+>>> tf_model = TFAutoModelForSequenceClassification.from_pretrained("./tf_save_pretrained")
+```
+</tf>
+</frameworkcontent>
+
+Ein besonders cooles 🤗 Transformers-Feature ist die Möglichkeit, ein Modell zu speichern und es entweder als PyTorch- oder TensorFlow-Modell wieder zu laden. Der Parameter "from_pt" oder "from_tf" kann das Modell von einem Framework in das andere konvertieren:
+
+<frameworkcontent>
+<pt>
+```py
+>>> from transformers import AutoModel
+
+>>> tokenizer = AutoTokenizer.from_pretrained(tf_save_directory)
+>>> pt_model = AutoModelForSequenceClassification.from_pretrained(tf_save_directory, from_tf=True)
+```
+</pt>
+<tf>
+```py
+>>> from transformers import TFAutoModel
+
+>>> tokenizer = AutoTokenizer.from_pretrained(pt_save_directory)
+>>> tf_model = TFAutoModelForSequenceClassification.from_pretrained(pt_save_directory, from_pt=True)
+```
+</tf>
+</frameworkcontent>
+
+## Custom model builds
+
+Sie können die Konfigurationsklasse des Modells ändern, um zu bestimmen, wie ein Modell aufgebaut ist. Die Konfiguration legt die Attribute eines Modells fest, z. B. die Anzahl der verborgenen Schichten oder der Aufmerksamkeitsköpfe. Wenn Sie ein Modell aus einer benutzerdefinierten Konfigurationsklasse initialisieren, beginnen Sie bei Null. Die Modellattribute werden zufällig initialisiert, und Sie müssen das Modell trainieren, bevor Sie es verwenden können, um aussagekräftige Ergebnisse zu erhalten.
+
+Beginnen Sie mit dem Import von [`AutoConfig`] und laden Sie dann das trainierte Modell, das Sie ändern möchten. Innerhalb von [`AutoConfig.from_pretrained`] können Sie das Attribut angeben, das Sie ändern möchten, z. B. die Anzahl der Aufmerksamkeitsköpfe:
+
+```py
+>>> from transformers import AutoConfig
+
+>>> my_config = AutoConfig.from_pretrained("distilbert-base-uncased", n_heads=12)
+```
+
+<frameworkcontent>
+<pt>
+Create a model from your custom configuration with [`AutoModel.from_config`]:
+
+```py
+>>> from transformers import AutoModel
+
+>>> my_model = AutoModel.from_config(my_config)
+```
+</pt>
+<tf>
+Create a model from your custom configuration with [`TFAutoModel.from_config`]:
+
+```py
+>>> from transformers import TFAutoModel
+
+>>> my_model = TFAutoModel.from_config(my_config)
+```
+</tf>
+</frameworkcontent>
+
+Weitere Informationen zur Erstellung von benutzerdefinierten Konfigurationen finden Sie in der Anleitung [Erstellen einer benutzerdefinierten Architektur](./create_a_model).
+
+## Wie geht es weiter?
+
+Nachdem Sie nun die 🤗 Transformers-Kurztour abgeschlossen haben, schauen Sie sich unsere Anleitungen an und erfahren Sie, wie Sie spezifischere Dinge tun können, wie das Schreiben eines benutzerdefinierten Modells, die Feinabstimmung eines Modells für eine Aufgabe und wie man ein Modell mit einem Skript trainiert. Wenn Sie mehr über die Kernkonzepte von 🤗 Transformers erfahren möchten, nehmen Sie sich eine Tasse Kaffee und werfen Sie einen Blick auf unsere konzeptionellen Leitfäden!

docs/source/de/training.mdx

@@ -0,0 +1,427 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# Optimierung eines vortrainierten Modells
+
+[[open-in-colab]]
+
+Die Verwendung eines vorab trainierten Modells hat erhebliche Vorteile. Es reduziert die Rechenkosten und den CO2-Fußabdruck und ermöglicht Ihnen die Verwendung von Modellen, die dem neuesten Stand der Technik entsprechen, ohne dass Sie ein Modell von Grund auf neu trainieren müssen. Transformers bietet Zugang zu Tausenden von vortrainierten Modellen für eine Vielzahl von Aufgaben. Wenn Sie ein vorab trainiertes Modell verwenden, trainieren Sie es auf einem für Ihre Aufgabe spezifischen Datensatz. Dies wird als Feinabstimmung bezeichnet und ist eine unglaublich leistungsfähige Trainingstechnik. In diesem Tutorial werden Sie ein vortrainiertes Modell mit einem Deep-Learning-Framework Ihrer Wahl feinabstimmen:
+
+* Feinabstimmung eines vorab trainierten Modells mit 🤗 Transformers [`Trainer`].
+* Feinabstimmung eines vorab trainierten Modells in TensorFlow mit Keras.
+* Feinabstimmung eines vorab trainierten Modells in nativem PyTorch.
+
+<a id='data-processing'></a>
+
+## Vorbereitung eines Datensatzes
+
+<Youtube id="_BZearw7f0w"/>
+
+Bevor Sie die Feinabstimmung eines vortrainierten Modells vornehmen können, müssen Sie einen Datensatz herunterladen und für das Training vorbereiten. Im vorangegangenen Leitfaden haben Sie gelernt, wie man Daten für das Training aufbereitet, und jetzt haben Sie die Gelegenheit, diese Fähigkeiten zu testen!
+
+Laden Sie zunächst den Datensatz [Yelp Reviews](https://huggingface.co/datasets/yelp_review_full):
+
+```py
+>>> from datasets import load_dataset
+
+>>> dataset = load_dataset("yelp_review_full")
+>>> dataset["train"][100]
+{'label': 0,
+ 'text': 'My expectations for McDonalds are t rarely high. But for one to still fail so spectacularly...that takes something special!\\nThe cashier took my friends\'s order, then promptly ignored me. I had to force myself in front of a cashier who opened his register to wait on the person BEHIND me. I waited over five minutes for a gigantic order that included precisely one kid\'s meal. After watching two people who ordered after me be handed their food, I asked where mine was. The manager started yelling at the cashiers for \\"serving off their orders\\" when they didn\'t have their food. But neither cashier was anywhere near those controls, and the manager was the one serving food to customers and clearing the boards.\\nThe manager was rude when giving me my order. She didn\'t make sure that I had everything ON MY RECEIPT, and never even had the decency to apologize that I felt I was getting poor service.\\nI\'ve eaten at various McDonalds restaurants for over 30 years. I\'ve worked at more than one location. I expect bad days, bad moods, and the occasional mistake. But I have yet to have a decent experience at this store. It will remain a place I avoid unless someone in my party needs to avoid illness from low blood sugar. Perhaps I should go back to the racially biased service of Steak n Shake instead!'}
+```
+
+Wie Sie nun wissen, benötigen Sie einen Tokenizer, um den Text zu verarbeiten und eine Auffüll- und Abschneidungsstrategie einzubauen, um mit variablen Sequenzlängen umzugehen. Um Ihren Datensatz in einem Schritt zu verarbeiten, verwenden Sie die 🤗 Methode Datasets [`map`](https://huggingface.co/docs/datasets/process.html#map), um eine Vorverarbeitungsfunktion auf den gesamten Datensatz anzuwenden:
+
+```py
+>>> from transformers import AutoTokenizer
+
+>>> tokenizer = AutoTokenizer.from_pretrained("bert-base-cased")
+
+
+>>> def tokenize_function(examples):
+...     return tokenizer(examples["text"], padding="max_length", truncation=True)
+
+
+>>> tokenized_datasets = dataset.map(tokenize_function, batched=True)
+```
+
+Wenn Sie möchten, können Sie eine kleinere Teilmenge des gesamten Datensatzes für die Feinabstimmung erstellen, um den Zeitaufwand zu verringern:
+
+```py
+>>> small_train_dataset = tokenized_datasets["train"].shuffle(seed=42).select(range(1000))
+>>> small_eval_dataset = tokenized_datasets["test"].shuffle(seed=42).select(range(1000))
+```
+
+<a id='trainer'></a>
+
+## Training
+
+An dieser Stelle sollten Sie dem Abschnitt folgen, der dem Rahmen entspricht, den Sie verwenden möchten. Sie können über die Links
+in der rechten Seitenleiste können Sie zu dem gewünschten Abschnitt springen - und wenn Sie den gesamten Inhalt eines bestimmten Frameworks ausblenden möchten,
+klicken Sie einfach auf die Schaltfläche oben rechts im Block des jeweiligen Frameworks!
+
+<frameworkcontent>
+<pt>
+<Youtube id="nvBXf7s7vTI"/>
+
+## Trainieren mit PyTorch Trainer
+
+🤗 Transformers bietet eine [`Trainer`]-Klasse, die für das Training von 🤗 Transformers-Modellen optimiert ist und es einfacher macht, mit dem Training zu beginnen, ohne manuell eine eigene Trainingsschleife zu schreiben. Die [`Trainer`]-API unterstützt eine breite Palette von Trainingsoptionen und Funktionen wie Logging, Gradientenakkumulation und gemischte Präzision.
+
+Beginnen Sie mit dem Laden Ihres Modells und geben Sie die Anzahl der erwarteten Labels an. Aus dem Yelp Review [dataset card](https://huggingface.co/datasets/yelp_review_full#data-fields) wissen Sie, dass es fünf Labels gibt:
+
+```py
+>>> from transformers import AutoModelForSequenceClassification
+
+>>> model = AutoModelForSequenceClassification.from_pretrained("bert-base-cased", num_labels=5)
+```
+
+<Tip>
+
+Es wird eine Warnung angezeigt, dass einige der trainierten Parameter nicht verwendet werden und einige Parameter zufällig
+initialisiert werden. Machen Sie sich keine Sorgen, das ist völlig normal! Der vorher trainierte Kopf des BERT-Modells wird verworfen und durch einen zufällig initialisierten Klassifikationskopf ersetzt. Sie werden diesen neuen Modellkopf in Ihrer Sequenzklassifizierungsaufgabe feinabstimmen, indem Sie das Wissen des vortrainierten Modells auf ihn übertragen.
+
+</Tip>
+
+### Hyperparameter für das Training
+
+Als Nächstes erstellen Sie eine Klasse [`TrainingArguments`], die alle Hyperparameter enthält, die Sie einstellen können, sowie Flags zur Aktivierung verschiedener Trainingsoptionen. Für dieses Lernprogramm können Sie mit den Standard- [Hyperparametern](https://huggingface.co/docs/transformers/main_classes/trainer#transformers.TrainingArguments) beginnen, aber Sie können mit diesen experimentieren, um Ihre optimalen Einstellungen zu finden.
+
+Geben Sie an, wo die Kontrollpunkte Ihres Trainings gespeichert werden sollen:
+
+```py
+>>> from transformers import TrainingArguments
+
+>>> training_args = TrainingArguments(output_dir="test_trainer")
+```
+
+### Auswerten
+
+Der [`Trainer`] wertet die Leistung des Modells während des Trainings nicht automatisch aus. Sie müssen [`Trainer`] eine Funktion übergeben, um Metriken zu berechnen und zu berichten. Die [🤗 Evaluate](https://huggingface.co/docs/evaluate/index) Bibliothek bietet eine einfache [`accuracy`](https://huggingface.co/spaces/evaluate-metric/accuracy) Funktion, die Sie mit der [`evaluate.load`] Funktion laden können (siehe diese [quicktour](https://huggingface.co/docs/evaluate/a_quick_tour) für weitere Informationen):
+
+```py
+>>> import numpy as np
+>>> import evaluate
+
+>>> metric = evaluate.load("accuracy")
+```
+
+Rufen Sie [`~evaluate.compute`] auf `metric` auf, um die Genauigkeit Ihrer Vorhersagen zu berechnen. Bevor Sie Ihre Vorhersagen an `compute` übergeben, müssen Sie die Vorhersagen in Logits umwandeln (denken Sie daran, dass alle 🤗 Transformers-Modelle Logits zurückgeben):
+
+```py
+>>> def compute_metrics(eval_pred):
+...     logits, labels = eval_pred
+...     predictions = np.argmax(logits, axis=-1)
+...     return metric.compute(predictions=predictions, references=labels)
+```
+
+Wenn Sie Ihre Bewertungsmetriken während der Feinabstimmung überwachen möchten, geben Sie den Parameter `evaluation_strategy` in Ihren Trainingsargumenten an, um die Bewertungsmetrik am Ende jeder Epoche zu ermitteln:
+
+```py
+>>> from transformers import TrainingArguments, Trainer
+
+>>> training_args = TrainingArguments(output_dir="test_trainer", evaluation_strategy="epoch")
+```
+
+### Trainer
+
+Erstellen Sie ein [`Trainer`]-Objekt mit Ihrem Modell, Trainingsargumenten, Trainings- und Testdatensätzen und einer Evaluierungsfunktion:
+
+```py
+>>> trainer = Trainer(
+...     model=model,
+...     args=training_args,
+...     train_dataset=small_train_dataset,
+...     eval_dataset=small_eval_dataset,
+...     compute_metrics=compute_metrics,
+... )
+```
+
+Anschließend können Sie Ihr Modell durch den Aufruf von [`~transformers.Trainer.train`] optimieren:
+
+```py
+>>> trainer.train()
+```
+</pt>
+<tf>
+<a id='keras'></a>
+
+<Youtube id="rnTGBy2ax1c"/>
+
+## Trainieren Sie ein TensorFlow-Modell mit Keras
+
+Sie können auch 🤗 Transformers Modelle in TensorFlow mit der Keras API trainieren!
+
+### Laden von Daten für Keras
+
+Wenn Sie ein 🤗 Transformers Modell mit der Keras API trainieren wollen, müssen Sie Ihren Datensatz in ein Format konvertieren, das
+Keras versteht. Wenn Ihr Datensatz klein ist, können Sie das Ganze einfach in NumPy-Arrays konvertieren und an Keras übergeben.
+Probieren wir das zuerst aus, bevor wir etwas Komplizierteres tun.
+
+Laden Sie zunächst ein Dataset. Wir werden den CoLA-Datensatz aus dem [GLUE-Benchmark](https://huggingface.co/datasets/glue) verwenden,
+da es sich um eine einfache Aufgabe zur Klassifizierung von binärem Text handelt, und nehmen vorerst nur den Trainingssplit.
+
+```py
+from datasets import load_dataset
+
+dataset = load_dataset("glue", "cola")
+dataset = dataset["train"]  # Just take the training split for now
+```
+
+Als nächstes laden Sie einen Tokenizer und tokenisieren die Daten als NumPy-Arrays. Beachten Sie, dass die Beschriftungen bereits eine Liste von 0 und 1en sind,
+Wir können sie also ohne Tokenisierung direkt in ein NumPy-Array konvertieren!
+
+```py
+from transformers import AutoTokenizer
+
+tokenizer = AutoTokenizer.from_pretrained("bert-base-cased")
+tokenized_data = tokenizer(dataset["text"], return_tensors="np", padding=True)
+
+labels = np.array(dataset["label"])  # Label is already an array of 0 and 1
+```
+
+Schließlich laden, [`compile`](https://keras.io/api/models/model_training_apis/#compile-method) und [`fit`](https://keras.io/api/models/model_training_apis/#fit-method) Sie das Modell:
+
+```py
+from transformers import TFAutoModelForSequenceClassification
+from tensorflow.keras.optimizers import Adam
+
+# Load and compile our model
+model = TFAutoModelForSequenceClassification.from_pretrained("bert-base-cased")
+# Lower learning rates are often better for fine-tuning transformers
+model.compile(optimizer=Adam(3e-5))
+
+model.fit(tokenized_data, labels)
+```
+
+<Tip>
+
+Sie müssen Ihren Modellen kein Verlustargument übergeben, wenn Sie sie `compile()`! Hugging-Face-Modelle wählen automatisch
+einen Loss, der für ihre Aufgabe und Modellarchitektur geeignet ist, wenn dieses Argument leer gelassen wird. Sie können jederzeit außer Kraft setzen, indem Sie selbst einen Loss angeben, wenn Sie das möchten!
+
+</Tip>
+
+Dieser Ansatz eignet sich hervorragend für kleinere Datensätze, aber bei größeren Datensätzen kann er zu einem Problem werden. Warum?
+Weil das tokenisierte Array und die Beschriftungen vollständig in den Speicher geladen werden müssten, und weil NumPy nicht mit
+"gezackte" Arrays nicht verarbeiten kann, so dass jedes tokenisierte Sample auf die Länge des längsten Samples im gesamten Datensatz aufgefüllt werden müsste.
+Datensatzes aufgefüllt werden. Dadurch wird das Array noch größer, und all die aufgefüllten Token verlangsamen auch das Training!
+
+### Laden von Daten als tf.data.Dataset
+
+Wenn Sie eine Verlangsamung des Trainings vermeiden wollen, können Sie Ihre Daten stattdessen als `tf.data.Dataset` laden. Sie können zwar Ihre eigene
+tf.data"-Pipeline schreiben können, wenn Sie wollen, haben wir zwei bequeme Methoden, um dies zu tun:
+
+- [`~TFPreTrainedModel.prepare_tf_dataset`]: Dies ist die Methode, die wir in den meisten Fällen empfehlen. Da es sich um eine Methode
+Ihres Modells ist, kann sie das Modell inspizieren, um automatisch herauszufinden, welche Spalten als Modelleingaben verwendet werden können, und
+verwirft die anderen, um einen einfacheren, leistungsfähigeren Datensatz zu erstellen.
+- [~datasets.Dataset.to_tf_dataset`]: Diese Methode ist eher auf niedriger Ebene angesiedelt und ist nützlich, wenn Sie genau kontrollieren wollen, wie
+Dataset erstellt wird, indem man genau angibt, welche `columns` und `label_cols` einbezogen werden sollen.
+
+Bevor Sie [~TFPreTrainedModel.prepare_tf_dataset`] verwenden können, müssen Sie die Tokenizer-Ausgaben als Spalten zu Ihrem Datensatz hinzufügen, wie in
+dem folgenden Codebeispiel:
+
+```py
+def tokenize_dataset(data):
+    # Keys of the returned dictionary will be added to the dataset as columns
+    return tokenizer(data["text"])
+
+
+dataset = dataset.map(tokenize_dataset)
+```
+
+Denken Sie daran, dass Hugging Face-Datensätze standardmäßig auf der Festplatte gespeichert werden, so dass dies nicht zu einem erhöhten Arbeitsspeicherbedarf führen wird! Sobald die
+Spalten hinzugefügt wurden, können Sie Batches aus dem Datensatz streamen und zu jedem Batch Auffüllungen hinzufügen, was die Anzahl der Auffüllungs-Token im Vergleich zum Auffüllen des gesamten Datensatzes reduziert.
+
+
+```py
+>>> tf_dataset = model.prepare_tf_dataset(dataset, batch_size=16, shuffle=True, tokenizer=tokenizer)
+```
+
+Beachten Sie, dass Sie im obigen Codebeispiel den Tokenizer an `prepare_tf_dataset` übergeben müssen, damit die Stapel beim Laden korrekt aufgefüllt werden können.
+Wenn alle Stichproben in Ihrem Datensatz die gleiche Länge haben und kein Auffüllen erforderlich ist, können Sie dieses Argument weglassen.
+Wenn Sie etwas Komplexeres als nur das Auffüllen von Stichproben benötigen (z. B. das Korrumpieren von Token für die maskierte Sprachmodellierung), können Sie das Argument
+Modellierung), können Sie stattdessen das Argument `collate_fn` verwenden, um eine Funktion zu übergeben, die aufgerufen wird, um die
+Liste von Stichproben in einen Stapel umwandelt und alle gewünschten Vorverarbeitungen vornimmt. Siehe unsere
+[examples](https://github.com/huggingface/transformers/tree/main/examples) oder
+[notebooks](https://huggingface.co/docs/transformers/notebooks), um diesen Ansatz in Aktion zu sehen.
+
+Sobald Sie einen `tf.data.Dataset` erstellt haben, können Sie das Modell wie zuvor kompilieren und anpassen:
+
+```py
+model.compile(optimizer=Adam(3e-5))
+
+model.fit(tf_dataset)
+```
+
+</tf>
+</frameworkcontent>
+
+<a id='pytorch_native'></a>
+
+## Trainieren in nativem PyTorch
+
+<frameworkcontent>
+<pt>
+<Youtube id="Dh9CL8fyG80"/>
+
+[`Trainer`] kümmert sich um die Trainingsschleife und ermöglicht die Feinabstimmung eines Modells in einer einzigen Codezeile. Für Benutzer, die es vorziehen, ihre eigene Trainingsschleife zu schreiben, können Sie auch eine Feinabstimmung eines 🤗 Transformers-Modells in nativem PyTorch vornehmen.
+
+An diesem Punkt müssen Sie möglicherweise Ihr Notebook neu starten oder den folgenden Code ausführen, um etwas Speicher freizugeben:
+
+```py
+del model
+del pytorch_model
+del trainer
+torch.cuda.empty_cache()
+```
+
+Als Nächstes müssen Sie den Datensatz `tokenized_dataset` manuell nachbearbeiten, um ihn für das Training vorzubereiten.
+
+1. Entfernen Sie die Spalte "Text", da das Modell keinen Rohtext als Eingabe akzeptiert:
+
+    ```py
+    >>> tokenized_datasets = tokenized_datasets.remove_columns(["text"])
+    ```
+
+2. Benennen Sie die Spalte "Label" in "Labels" um, da das Modell erwartet, dass das Argument "Labels" genannt wird:
+
+    ```py
+    >>> tokenized_datasets = tokenized_datasets.rename_column("label", "labels")
+    ```
+
+3. Stellen Sie das Format des Datensatzes so ein, dass PyTorch-Tensoren anstelle von Listen zurückgegeben werden:
+
+    ```py
+    >>> tokenized_datasets.set_format("torch")
+    ```
+
+Erstellen Sie dann eine kleinere Teilmenge des Datensatzes, wie zuvor gezeigt, um die Feinabstimmung zu beschleunigen:
+
+```py
+>>> small_train_dataset = tokenized_datasets["train"].shuffle(seed=42).select(range(1000))
+>>> small_eval_dataset = tokenized_datasets["test"].shuffle(seed=42).select(range(1000))
+```
+
+### DataLoader
+
+Erstellen Sie einen `DataLoader` für Ihre Trainings- und Testdatensätze, damit Sie über die Datenstapel iterieren können:
+
+```py
+>>> from torch.utils.data import DataLoader
+
+>>> train_dataloader = DataLoader(small_train_dataset, shuffle=True, batch_size=8)
+>>> eval_dataloader = DataLoader(small_eval_dataset, batch_size=8)
+```
+
+Laden Sie Ihr Modell mit der Anzahl der erwarteten Kennzeichnungen:
+
+```py
+>>> from transformers import AutoModelForSequenceClassification
+
+>>> model = AutoModelForSequenceClassification.from_pretrained("bert-base-cased", num_labels=5)
+```
+
+### Optimierer und Lernratensteuerung
+
+Erstellen Sie einen Optimierer und einen Scheduler für die Lernrate, um das Modell fein abzustimmen. Wir verwenden den Optimierer [`AdamW`](https://pytorch.org/docs/stable/generated/torch.optim.AdamW.html) aus PyTorch:
+
+```py
+>>> from torch.optim import AdamW
+
+>>> optimizer = AdamW(model.parameters(), lr=5e-5)
+```
+
+Erstellen Sie den Standard-Lernratenplaner aus [`Trainer`]:
+
+```py
+>>> from transformers import get_scheduler
+
+>>> num_epochs = 3
+>>> num_training_steps = num_epochs * len(train_dataloader)
+>>> lr_scheduler = get_scheduler(
+...     name="linear", optimizer=optimizer, num_warmup_steps=0, num_training_steps=num_training_steps
+... )
+```
+
+Geben Sie schließlich `device` an, um einen Grafikprozessor zu verwenden, wenn Sie Zugang zu einem solchen haben. Andernfalls kann das Training auf einer CPU mehrere Stunden statt ein paar Minuten dauern.
+
+```py
+>>> import torch
+
+>>> device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
+>>> model.to(device)
+```
+
+<Tip>
+
+Holen Sie sich mit einem gehosteten Notebook wie [Colaboratory](https://colab.research.google.com/) oder [SageMaker StudioLab](https://studiolab.sagemaker.aws/) kostenlosen Zugang zu einem Cloud-GPU, wenn Sie noch keinen haben.
+
+</Tip>
+
+Großartig, Sie sind bereit für das Training! 🥳 
+
+### Trainingsschleife
+
+Um Ihren Trainingsfortschritt zu verfolgen, verwenden Sie die [tqdm](https://tqdm.github.io/) Bibliothek, um einen Fortschrittsbalken über die Anzahl der Trainingsschritte hinzuzufügen:
+
+```py
+>>> from tqdm.auto import tqdm
+
+>>> progress_bar = tqdm(range(num_training_steps))
+
+>>> model.train()
+>>> for epoch in range(num_epochs):
+...     for batch in train_dataloader:
+...         batch = {k: v.to(device) for k, v in batch.items()}
+...         outputs = model(**batch)
+...         loss = outputs.loss
+...         loss.backward()
+
+...         optimizer.step()
+...         lr_scheduler.step()
+...         optimizer.zero_grad()
+...         progress_bar.update(1)
+```
+
+### Auswertung
+
+Genauso wie Sie eine Bewertungsfunktion zu [`Trainer`] hinzugefügt haben, müssen Sie dasselbe tun, wenn Sie Ihre eigene Trainingsschleife schreiben. Aber anstatt die Metrik am Ende jeder Epoche zu berechnen und zu melden, werden Sie dieses Mal alle Stapel mit [`~evaluate.add_batch`] akkumulieren und die Metrik ganz am Ende berechnen.
+
+```py
+>>> import evaluate
+
+>>> metric = evaluate.load("accuracy")
+>>> model.eval()
+>>> for batch in eval_dataloader:
+...     batch = {k: v.to(device) for k, v in batch.items()}
+...     with torch.no_grad():
+...         outputs = model(**batch)
+
+...     logits = outputs.logits
+...     predictions = torch.argmax(logits, dim=-1)
+...     metric.add_batch(predictions=predictions, references=batch["labels"])
+
+>>> metric.compute()
+```
+</pt>
+</frameworkcontent>
+
+<a id='additional-resources'></a>
+
+## Zusätzliche Ressourcen
+
+Weitere Beispiele für die Feinabstimmung finden Sie unter:
+
+- [🤗 Transformers Examples](https://github.com/huggingface/transformers/tree/main/examples) enthält Skripte
+  um gängige NLP-Aufgaben in PyTorch und TensorFlow zu trainieren.
+
+- [🤗 Transformers Notebooks](notebooks) enthält verschiedene Notebooks zur Feinabstimmung eines Modells für bestimmte Aufgaben in PyTorch und TensorFlow.

docs/source/en/_toctree.yml

@@ -21,44 +21,59 @@
     title: Share a model
   title: Tutorials
 - sections:
-  - local: fast_tokenizers
-    title: Use tokenizers from 🤗 Tokenizers
-  - local: create_a_model
-    title: Create a custom architecture
-  - local: custom_models
-    title: Sharing custom models
   - sections:
-    - local: tasks/sequence_classification
-      title: Text classification
-    - local: tasks/token_classification
-      title: Token classification
-    - local: tasks/question_answering
-      title: Question answering
-    - local: tasks/language_modeling
-      title: Language modeling
-    - local: tasks/translation
-      title: Translation
-    - local: tasks/summarization
-      title: Summarization
-    - local: tasks/multiple_choice
-      title: Multiple choice
+    - local: create_a_model
+      title: Create a custom architecture
+    - local: custom_models
+      title: Sharing custom models
+    - local: run_scripts
+      title: Train with a script
+    - local: sagemaker
+      title: Run training on Amazon SageMaker
+    - local: converting_tensorflow_models
+      title: Converting from TensorFlow checkpoints
+    - local: serialization
+      title: Export to ONNX
+    - local: torchscript
+      title: Export to TorchScript
+    - local: troubleshooting
+      title: Troubleshoot
+    title: General usage
+  - sections:
+    - local: fast_tokenizers
+      title: Use tokenizers from 🤗 Tokenizers
+    - local: multilingual
+      title: Inference for multilingual models
+    - sections:
+      - local: tasks/sequence_classification
+        title: Text classification
+      - local: tasks/token_classification
+        title: Token classification
+      - local: tasks/question_answering
+        title: Question answering
+      - local: tasks/language_modeling
+        title: Language modeling
+      - local: tasks/translation
+        title: Translation
+      - local: tasks/summarization
+        title: Summarization
+      - local: tasks/multiple_choice
+        title: Multiple choice
+      title: Task guides
+      isExpanded: false
+    title: Natural Language Processing
+  - sections:
     - local: tasks/audio_classification
       title: Audio classification
     - local: tasks/asr
       title: Automatic speech recognition
+    title: Audio
+  - sections:
     - local: tasks/image_classification
       title: Image classification
-    title: Fine-tune for downstream tasks
-  - local: run_scripts
-    title: Train with a script
-  - local: sagemaker
-    title: Run training on Amazon SageMaker
-  - local: multilingual
-    title: Inference for multilingual models
-  - local: converting_tensorflow_models
-    title: Converting TensorFlow Checkpoints
-  - local: serialization
-    title: Export 🤗 Transformers models
+    - local: tasks/semantic_segmentation
+      title: Semantic segmentation
+    title: Computer Vision
   - sections:
     - local: performance
       title: Overview
@@ -84,31 +99,35 @@
       title: Inference on Specialized Hardware
     - local: perf_hardware
       title: Custom hardware for training
+    - local: big_models
+      title: Instantiating a big model
+    - local: debugging
+      title: Debugging
+    - local: hpo_train
+      title: Hyperparameter Search using Trainer API
     title: Performance and scalability
-  - local: big_models
-    title: Instantiating a big model
+  - sections:
+    - local: contributing
+      title: How to contribute to transformers?
+    - local: add_new_model
+      title: How to add a model to 🤗 Transformers?
+    - local: add_tensorflow_model
+      title: How to convert a 🤗 Transformers model to TensorFlow?
+    - local: add_new_pipeline
+      title: How to add a pipeline to 🤗 Transformers?
+    - local: testing
+      title: Testing
+    - local: pr_checks
+      title: Checks on a Pull Request
+    title: Contribute
+  - local: notebooks
+    title: 🤗 Transformers Notebooks
+  - local: community
+    title: Community resources
   - local: benchmarks
     title: Benchmarks
   - local: migration
     title: Migrating from previous packages
-  - local: troubleshooting
-    title: Troubleshoot
-  - local: debugging
-    title: Debugging
-  - local: notebooks
-    title: 🤗 Transformers Notebooks
-  - local: community
-    title: Community
-  - local: contributing
-    title: How to contribute to transformers?
-  - local: add_new_model
-    title: How to add a model to 🤗 Transformers?
-  - local: add_new_pipeline
-    title: How to create a custom pipeline?
-  - local: testing
-    title: Testing
-  - local: pr_checks
-    title: Checks on a Pull Request
   title: How-to guides
 - sections:
   - local: philosophy
@@ -130,6 +149,8 @@
   title: Conceptual guides
 - sections:
   - sections:
+    - local: model_doc/auto
+      title: Auto Classes
     - local: main_classes/callback
       title: Callbacks
     - local: main_classes/configuration
@@ -164,8 +185,6 @@
       title: Feature Extractor
     title: Main Classes
   - sections:
-    - local: model_doc/auto
-      title: Auto Classes
     - isExpanded: false
       sections:
       - local: model_doc/albert
@@ -224,6 +243,10 @@
         title: ELECTRA
       - local: model_doc/encoder-decoder
         title: Encoder Decoder Models
+      - local: model_doc/ernie
+        title: ERNIE
+      - local: model_doc/esm
+        title: ESM
       - local: model_doc/flaubert
         title: FlauBERT
       - local: model_doc/fnet
@@ -238,6 +261,8 @@
         title: GPT Neo
       - local: model_doc/gpt_neox
         title: GPT NeoX
+      - local: model_doc/gpt_neox_japanese
+        title: GPT NeoX Japanese
       - local: model_doc/gptj
         title: GPT-J
       - local: model_doc/gpt2
@@ -260,6 +285,8 @@
         title: M2M100
       - local: model_doc/marian
         title: MarianMT
+      - local: model_doc/markuplm
+        title: MarkupLM
       - local: model_doc/mbart
         title: MBart and MBart-50
       - local: model_doc/megatron-bert
@@ -286,6 +313,8 @@
         title: OPT
       - local: model_doc/pegasus
         title: Pegasus
+      - local: model_doc/pegasus_x
+        title: PEGASUS-X
       - local: model_doc/phobert
         title: PhoBERT
       - local: model_doc/plbart
@@ -343,10 +372,14 @@
       sections:
       - local: model_doc/beit
         title: BEiT
+      - local: model_doc/conditional_detr
+        title: Conditional DETR
       - local: model_doc/convnext
         title: ConvNeXT
       - local: model_doc/cvt
         title: CvT
+      - local: model_doc/deformable_detr
+        title: Deformable DETR
       - local: model_doc/deit
         title: DeiT
       - local: model_doc/detr
@@ -365,8 +398,6 @@
         title: MaskFormer
       - local: model_doc/mobilevit
         title: MobileViT
-      - local: model_doc/owlvit
-        title: OWL-ViT
       - local: model_doc/poolformer
         title: PoolFormer
       - local: model_doc/regnet
@@ -381,10 +412,14 @@
         title: Swin Transformer V2
       - local: model_doc/van
         title: VAN
+      - local: model_doc/videomae
+        title: VideoMAE
       - local: model_doc/vit
         title: Vision Transformer (ViT)
       - local: model_doc/vit_mae
         title: ViTMAE
+      - local: model_doc/vit_msn
+        title: ViTMSN
       - local: model_doc/yolos
         title: YOLOS
       title: Vision models
@@ -414,6 +449,8 @@
         title: Wav2Vec2Phoneme
       - local: model_doc/wavlm
         title: WavLM
+      - local: model_doc/whisper
+        title: Whisper
       - local: model_doc/xls_r
         title: XLS-R
       - local: model_doc/xlsr_wav2vec2
@@ -425,6 +462,8 @@
         title: CLIP
       - local: model_doc/data2vec
         title: Data2Vec
+      - local: model_doc/donut
+        title: Donut
       - local: model_doc/flava
         title: FLAVA
       - local: model_doc/groupvit
@@ -437,6 +476,8 @@
         title: LayoutXLM
       - local: model_doc/lxmert
         title: LXMERT
+      - local: model_doc/owlvit
+        title: OWL-ViT
       - local: model_doc/perceiver
         title: Perceiver
       - local: model_doc/speech-encoder-decoder
@@ -451,6 +492,8 @@
         title: Vision Text Dual Encoder
       - local: model_doc/visual_bert
         title: VisualBERT
+      - local: model_doc/xclip
+        title: X-CLIP
       title: Multimodal models
     - isExpanded: false
       sections:
@@ -459,6 +502,11 @@
       - local: model_doc/trajectory_transformer
         title: Trajectory Transformer
       title: Reinforcement learning models
+    - isExpanded: false
+      sections:
+      - local: model_doc/time_series_transformer
+        title: Time Series Transformer
+      title: Time series models
     title: Models
   - sections:
     - local: internal/modeling_utils

docs/source/en/accelerate.mdx

@@ -12,7 +12,7 @@ specific language governing permissions and limitations under the License.
 
 # Distributed training with 🤗 Accelerate
 
-As models get bigger, parallelism has emerged as a strategy for training larger models on limited hardware and accelerating training speed by several orders of magnitude. At Hugging Face, we created the [🤗 Accelerate](https://huggingface.co/docs/accelerate/index.html) library to help users easily train a 🤗 Transformers model on any type of distributed setup, whether it is multiple GPU's on one machine or multiple GPU's across several machines. In this tutorial, learn how to customize your native PyTorch training loop to enable training in a distributed environment.
+As models get bigger, parallelism has emerged as a strategy for training larger models on limited hardware and accelerating training speed by several orders of magnitude. At Hugging Face, we created the [🤗 Accelerate](https://huggingface.co/docs/accelerate) library to help users easily train a 🤗 Transformers model on any type of distributed setup, whether it is multiple GPU's on one machine or multiple GPU's across several machines. In this tutorial, learn how to customize your native PyTorch training loop to enable training in a distributed environment.
 
 ## Setup
 
@@ -22,7 +22,7 @@ Get started by installing 🤗 Accelerate:
 pip install accelerate
 ```
 
-Then import and create an [`Accelerator`](https://huggingface.co/docs/accelerate/accelerator.html#accelerate.Accelerator) object. `Accelerator` will automatically detect your type of distributed setup and initialize all the necessary components for training. You don't need to explicitly place your model on a device.
+Then import and create an [`~accelerate.Accelerator`] object. The [`~accelerate.Accelerator`] will automatically detect your type of distributed setup and initialize all the necessary components for training. You don't need to explicitly place your model on a device.
 
 ```py
 >>> from accelerate import Accelerator
@@ -32,7 +32,7 @@ Then import and create an [`Accelerator`](https://huggingface.co/docs/accelerate
 
 ## Prepare to accelerate
 
-The next step is to pass all the relevant training objects to the [`prepare`](https://huggingface.co/docs/accelerate/accelerator.html#accelerate.Accelerator.prepare) method. This includes your training and evaluation DataLoaders, a model and an optimizer:
+The next step is to pass all the relevant training objects to the [`~accelerate.Accelerator.prepare`] method. This includes your training and evaluation DataLoaders, a model and an optimizer:
 
 ```py
 >>> train_dataloader, eval_dataloader, model, optimizer = accelerator.prepare(
@@ -42,7 +42,7 @@ The next step is to pass all the relevant training objects to the [`prepare`](ht
 
 ## Backward
 
-The last addition is to replace the typical `loss.backward()` in your training loop with 🤗 Accelerate's [`backward`](https://huggingface.co/docs/accelerate/accelerator.html#accelerate.Accelerator.backward) method:
+The last addition is to replace the typical `loss.backward()` in your training loop with 🤗 Accelerate's [`~accelerate.Accelerator.backward`]method:
 
 ```py
 >>> for epoch in range(num_epochs):
@@ -121,7 +121,7 @@ accelerate launch train.py
 
 ### Train with a notebook
 
-🤗 Accelerate can also run in a notebook if you're planning on using Colaboratory's TPUs. Wrap all the code responsible for training in a function, and pass it to `notebook_launcher`:
+🤗 Accelerate can also run in a notebook if you're planning on using Colaboratory's TPUs. Wrap all the code responsible for training in a function, and pass it to [`~accelerate.notebook_launcher`]:
 
 ```py
 >>> from accelerate import notebook_launcher
@@ -129,4 +129,4 @@ accelerate launch train.py
 >>> notebook_launcher(training_function)
 ```
 
-For more information about 🤗 Accelerate and it's rich features, refer to the [documentation](https://huggingface.co/docs/accelerate/index.html).
+For more information about 🤗 Accelerate and it's rich features, refer to the [documentation](https://huggingface.co/docs/accelerate).

docs/source/en/add_new_model.mdx

@@ -106,7 +106,7 @@ own regarding how code should be written :-)
    for a good example).
 2. The code should be fully understandable, even by a non-native English speaker. This means you should pick
    descriptive variable names and avoid abbreviations. As an example, `activation` is preferred to `act`.
-   One-letter variable names are strongly discouraged unless it's an index in a for loop. 
+   One-letter variable names are strongly discouraged unless it's an index in a for loop.
 3. More generally we prefer longer explicit code to short magical one.
 4. Avoid subclassing `nn.Sequential` in PyTorch but subclass `nn.Module` and write the forward pass, so that anyone
    using your code can quickly debug it by adding print statements or breaking points.
@@ -222,7 +222,7 @@ cd ..
 5. To port *brand_new_bert*, you will also need access to its original repository:
 
 ```bash
-git clone https://github.com/org_that_created_brand_new_bert_org/brand_new_bert.git 
+git clone https://github.com/org_that_created_brand_new_bert_org/brand_new_bert.git
 cd brand_new_bert
 pip install -e .
 ```
@@ -683,10 +683,11 @@ work left to be done should be a cakewalk 😊.
 At this point, you have successfully added a new model. However, it is very much possible that the model does not yet
 fully comply with the required design. To make sure, the implementation is fully compatible with 🤗 Transformers, all
 common tests should pass. The Cookiecutter should have automatically added a test file for your model, probably under
-the same `tests/test_modeling_brand_new_bert.py`. Run this test file to verify that all common tests pass:
+the same `tests/models/brand_new_bert/test_modeling_brand_new_bert.py`. Run this test file to verify that all common
+tests pass:
 
 ```bash
-pytest tests/test_modeling_brand_new_bert.py
+pytest tests/models/brand_new_bert/test_modeling_brand_new_bert.py
 ```
 
 Having fixed all common tests, it is now crucial to ensure that all the nice work you have done is well tested, so that
@@ -700,7 +701,7 @@ Cookiecutter, called `BrandNewBertModelIntegrationTests` and only has to be fill
 tests are passing, run
 
 ```bash
-RUN_SLOW=1 pytest -sv tests/test_modeling_brand_new_bert.py::BrandNewBertModelIntegrationTests
+RUN_SLOW=1 pytest -sv tests/models/brand_new_bert/test_modeling_brand_new_bert.py::BrandNewBertModelIntegrationTests
 ```
 
 <Tip>
@@ -758,7 +759,8 @@ contain a couple of hard-coded integration tests.
 **10. Run End-to-end integration tests**
 
 Having added the tokenizer, you should also add a couple of end-to-end integration tests using both the model and the
-tokenizer to `tests/test_modeling_brand_new_bert.py` in 🤗 Transformers. Such a test should show on a meaningful
+tokenizer to `tests/models/brand_new_bert/test_modeling_brand_new_bert.py` in 🤗 Transformers.
+Such a test should show on a meaningful
 text-to-text sample that the 🤗 Transformers implementation works as expected. A meaningful text-to-text sample can
 include *e.g.* a source-to-target-translation pair, an article-to-summary pair, a question-to-answer pair, etc… If none
 of the ported checkpoints has been fine-tuned on a downstream task it is enough to simply rely on the model tests. In a

docs/source/en/add_tensorflow_model.mdx

@@ -0,0 +1,346 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+-->
+
+# How to convert a 🤗 Transformers model to TensorFlow?
+
+Having multiple frameworks available to use with 🤗 Transformers gives you flexibility to play their strengths when
+designing your application, but it implies that compatibility must be added on a per-model basis. The good news is that
+adding TensorFlow compatibility to an existing model is simpler than [adding a new model from scratch](add_new_model)!
+Whether you wish to have a deeper understanding of large TensorFlow models, make a major open-source contribution, or
+enable TensorFlow for your model of choice, this guide is for you.
+
+This guide empowers you, a member of our community, to contribute TensorFlow model weights and/or
+architectures to be used in 🤗 Transformers, with minimal supervision from the Hugging Face team. Writing a new model
+is no small feat, but hopefully this guide will make it less of a rollercoaster 🎢 and more of a walk in the park 🚶.
+Harnessing our collective experiences is absolutely critical to make this process increasingly easier, and thus we
+highly encourage that you suggest improvements to this guide!
+
+Before you dive deeper, it is recommended that you check the following resources if you're new to 🤗 Transformers:
+- [General overview of 🤗 Transformers](add_new_model#general-overview-of-transformers)
+- [Hugging Face's TensorFlow Philosophy](https://huggingface.co/blog/tensorflow-philosophy)
+
+In the remainder of this guide, you will learn what's needed to add a new TensorFlow model architecture, the
+procedure to convert PyTorch into TensorFlow model weights, and how to efficiently debug mismatches across ML
+frameworks. Let's get started!
+
+<Tip>
+
+Are you unsure whether the model you wish to use already has a corresponding TensorFlow architecture?
+
+&nbsp;
+
+Check the `model_type` field of the `config.json` of your model of choice
+([example](https://huggingface.co/bert-base-uncased/blob/main/config.json#L14)). If the corresponding model folder in
+🤗 Transformers has a file whose name starts with "modeling_tf", it means that it has a corresponding TensorFlow
+architecture ([example](https://github.com/huggingface/transformers/tree/main/src/transformers/models/bert)).
+
+</Tip>
+
+
+## Step-by-step guide to add TensorFlow model architecture code
+
+There are many ways to design a large model architecture, and multiple ways of implementing said design. However,
+you might recall from our [general overview of 🤗 Transformers](add_new_model#general-overview-of-transformers)
+that we are an opinionated bunch - the ease of use of 🤗 Transformers relies on consistent design choices. From
+experience, we can tell you a few important things about adding TensorFlow models:
+
+- Don't reinvent the wheel! More often that not, there are at least two reference implementations you should check: the
+PyTorch equivalent of the model you are implementing and other TensorFlow models for the same class of problems.
+- Great model implementations survive the test of time. This doesn't happen because the code is pretty, but rather
+because the code is clear, easy to debug and build upon. If you make the life of the maintainers easy with your
+TensorFlow implementation, by replicating the same patterns as in other TensorFlow models and minimizing the mismatch
+to the PyTorch implementation, you ensure your contribution will be long lived.
+- Ask for help when you're stuck! The 🤗 Transformers team is here to help, and we've probably found solutions to the same
+problems you're facing.
+
+Here's an overview of the steps needed to add a TensorFlow model architecture:
+1. Select the model you wish to convert
+2. Prepare transformers dev environment
+3. (Optional) Understand theoretical aspects and the existing implementation
+4. Implement the model architecture
+5. Implement model tests
+6. Submit the pull request
+7. (Optional) Build demos and share with the world
+
+### 1.-3. Prepare your model contribution
+
+**1. Select the model you wish to convert**
+
+Let's start off with the basics: the first thing you need to know is the architecture you want to convert. If you
+don't have your eyes set on a specific architecture, asking the 🤗 Transformers team for suggestions is a great way to
+maximize your impact - we will guide you towards the most prominent architectures that are missing on the TensorFlow
+side. If the specific model you want to use with TensorFlow already has a TensorFlow architecture implementation in
+🤗 Transformers but is lacking weights, feel free to jump straight into the
+[weight conversion section](#adding-tensorflow-weights-to-hub)
+of this page.
+
+For simplicity, the remainder of this guide assumes you've decided to contribute with the TensorFlow version of
+*BrandNewBert* (the same example as in the [guide](add_new_model) to add a new model from scratch).
+
+<Tip>
+
+Before starting the work on a TensorFlow model architecture, double-check that there is no ongoing effort to do so.
+You can search for `BrandNewBert` on the
+[pull request GitHub page](https://github.com/huggingface/transformers/pulls?q=is%3Apr) to confirm that there is no
+TensorFlow-related pull request.
+
+</Tip>
+
+
+**2. Prepare transformers dev environment**
+
+Having selected the model architecture, open an draft PR to signal your intention to work on it. Follow the
+instructions below to set up your environment and open a draft PR.
+
+1. Fork the [repository](https://github.com/huggingface/transformers) by clicking on the 'Fork' button on the
+   repository's page. This creates a copy of the code under your GitHub user account.
+
+2. Clone your `transformers` fork to your local disk, and add the base repository as a remote:
+
+```bash
+git clone https://github.com/[your Github handle]/transformers.git
+cd transformers
+git remote add upstream https://github.com/huggingface/transformers.git
+```
+
+3. Set up a development environment, for instance by running the following command:
+
+```bash
+python -m venv .env
+source .env/bin/activate
+pip install -e ".[dev]"
+```
+
+**Note:** You don't need to have CUDA installed. Making the new model work on CPU is sufficient.
+
+4. Create a branch with a descriptive name from your main branch
+
+```bash
+git checkout -b add_tf_brand_new_bert
+```
+
+5. Fetch and rebase to current main
+
+```bash
+git fetch upstream
+git rebase upstream/main
+```
+
+6. Add an empty `.py` file in `transformers/src/models/brandnewbert/` named `modeling_tf_brandnewbert.py`. This will
+be your TensorFlow model file.
+
+7. Push the changes to your account using:
+
+```bash
+git add .
+git commit -m "initial commit"
+git push -u origin add_tf_brand_new_bert
+```
+
+8. Once you are satisfied, go to the webpage of your fork on GitHub. Click on “Pull request”. Make sure to add the
+   GitHub handle of some members of the Hugging Face team as reviewers, so that the Hugging Face team gets notified for
+   future changes.
+
+9. Change the PR into a draft by clicking on “Convert to draft” on the right of the GitHub pull request web page.
+
+
+Now you have set up a development environment to port *BrandNewBert* to TensorFlow in 🤗 Transformers.
+
+
+**3. (Optional) Understand theoretical aspects and the existing implementation**
+
+You should take some time to read *BrandNewBert's* paper, if such descriptive work exists. There might be large
+sections of the paper that are difficult to understand. If this is the case, this is fine - don't worry! The goal is
+not to get a deep theoretical understanding of the paper, but to extract the necessary information required to
+effectively re-implement the model in 🤗 Transformers using TensorFlow. That being said, you don't have to spend too
+much time on the theoretical aspects, but rather focus on the practical ones, namely the existing model documentation
+page (e.g. [model docs for BERT](model_doc/bert)).
+
+After you've grasped the basics of the models you are about to implement, it's important to understand the existing
+implementation. This is a great chance to confirm that a working implementation matches your expectations for the
+model, as well as to foresee technical challenges on the TensorFlow side.
+
+It's perfectly natural that you feel overwhelmed with the amount of information that you've just absorbed. It is
+definitely not a requirement that you understand all facets of the model at this stage. Nevertheless, we highly
+encourage you to clear any pressing questions in our [forum](https://discuss.huggingface.co/).
+
+
+### 4. Model implementation
+
+Now it's time to finally start coding. Our suggested starting point is the PyTorch file itself: copy the contents of
+`modeling_brand_new_bert.py` inside `src/transformers/models/brand_new_bert/` into
+`modeling_tf_brand_new_bert.py`. The goal of this section is to modify the file and update the import structure of
+🤗 Transformers such that you can import `TFBrandNewBert` and
+`TFBrandNewBert.from_pretrained(model_repo, from_pt=True)` sucessfully loads a working TensorFlow *BrandNewBert* model.
+
+Sadly, there is no prescription to convert a PyTorch model into TensorFlow. You can, however, follow our selection of
+tips to make the process as smooth as possible:
+- Prepend `TF` to the name of all classes (e.g. `BrandNewBert` becomes `TFBrandNewBert`).
+- Most PyTorch operations have a direct TensorFlow replacement. For example, `torch.nn.Linear` corresponds to
+  `tf.keras.layers.Dense`, `torch.nn.Dropout` corresponds to `tf.keras.layers.Dropout`, etc. If you're not sure
+  about a specific operation, you can use the [TensorFlow documentation](https://www.tensorflow.org/api_docs/python/tf)
+  or the [PyTorch documentation](https://pytorch.org/docs/stable/).
+- Look for patterns in the 🤗 Transformers codebase. If you come across a certain operation that doesn't have a direct
+   replacement, the odds are that someone else already had the same problem.
+- By default, keep the same variable names and structure as in PyTorch. This will make it easier to debug, track
+   issues, and add fixes down the line.
+- Some layers have different default values in each framework. A notable example is the batch normalization layer's
+   epsilon (`1e-5` in [PyTorch](https://pytorch.org/docs/stable/generated/torch.nn.BatchNorm2d.html#torch.nn.BatchNorm2d)
+   and `1e-3` in [TensorFlow](https://www.tensorflow.org/api_docs/python/tf/keras/layers/BatchNormalization)).
+   Double-check the documentation!
+- PyTorch's `nn.Parameter` variables typically need to be initialized within TF Layer's `build()`. See the following
+   example: [PyTorch](https://github.com/huggingface/transformers/blob/655f72a6896c0533b1bdee519ed65a059c2425ac/src/transformers/models/vit_mae/modeling_vit_mae.py#L212) /
+   [TensorFlow](https://github.com/huggingface/transformers/blob/655f72a6896c0533b1bdee519ed65a059c2425ac/src/transformers/models/vit_mae/modeling_tf_vit_mae.py#L220)
+- If the PyTorch model has a `#copied from ...` on top of a function, the odds are that your TensorFlow model can also
+   borrow that function from the architecture it was copied from, assuming it has a TensorFlow architecture.
+- Assigning the `name` attribute correctly in TensorFlow functions is critical to do the `from_pt=True` weight
+   cross-loading. `name` is almost always the name of the corresponding variable in the PyTorch code. If `name` is not
+   properly set, you will see it in the error message when loading the model weights.
+- The logic of the base model class, `BrandNewBertModel`, will actually reside in `TFBrandNewBertMainLayer`, a Keras
+   layer subclass ([example](https://github.com/huggingface/transformers/blob/4fd32a1f499e45f009c2c0dea4d81c321cba7e02/src/transformers/models/bert/modeling_tf_bert.py#L719)).
+   `TFBrandNewBertModel` will simply be a wrapper around this layer.
+- Keras models need to be built in order to load pretrained weights. For that reason, `TFBrandNewBertPreTrainedModel`
+   will need to hold an example of inputs to the model, the `dummy_inputs`
+   ([example](https://github.com/huggingface/transformers/blob/4fd32a1f499e45f009c2c0dea4d81c321cba7e02/src/transformers/models/bert/modeling_tf_bert.py#L916)).
+- If you get stuck, ask for help - we're here to help you! 🤗
+
+In addition to the model file itself, you will also need to add the pointers to the model classes and related
+documentation pages. You can complete this part entirely following the patterns in other PRs
+([example](https://github.com/huggingface/transformers/pull/18020/files)). Here's a list of the needed manual
+changes:
+- Include all public classes of *BrandNewBert* in `src/transformers/__init__.py`
+- Add *BrandNewBert* classes to the corresponing Auto classes in `src/transformers/models/auto/modeling_tf_auto.py`
+- Include the modeling file in the documentation test file list in `utils/documentation_tests.txt`
+- Add the lazy loading classes related to *BrandNewBert* in `src/transformers/utils/dummy_tf_objects.py`
+- Update the import structures for the public classes in `src/transformers/models/brand_new_bert/__init__.py`
+- Add the documentation pointers to the public methods of *BrandNewBert* in `docs/source/en/model_doc/brand_new_bert.mdx`
+- Add yourself to the list of contributors to *BrandNewBert* in `docs/source/en/model_doc/brand_new_bert.mdx`
+- Finally, add a green tick ✅ to the TensorFlow column of *BrandNewBert* in `docs/source/en/index.mdx`
+
+When you're happy with your implementation, run the following checklist to confirm that your model architecture is
+ready:
+1. All layers that behave differently at train time (e.g. Dropout) are called with a `training` argument, which is
+propagated all the way from the top-level classes
+2. You have used `#copied from ...` whenever possible
+3. `TFBrandNewBertMainLayer` and all classes that use it have their `call` function decorated with `@unpack_inputs`
+4. `TFBrandNewBertMainLayer` is decorated with `@keras_serializable`
+5. A TensorFlow model can be loaded from PyTorch weights using `TFBrandNewBert.from_pretrained(model_repo, from_pt=True)`
+6. You can call the TensorFlow model using the expected input format
+
+
+### 5. Add model tests
+
+Hurray, you've implemented a TensorFlow model! Now it's time to add tests to make sure that your model behaves as
+expected. As in the previous section, we suggest you start by copying the `test_modeling_brand_new_bert.py` file in
+`tests/models/brand_new_bert/` into `test_modeling_tf_brand_new_bert.py`, and continue by making the necessary
+TensorFlow replacements. For now, in all `.from_pretrained()` calls, you should use the `from_pt=True` flag to load
+the existing PyTorch weights.
+
+After you're done, it's time for the moment of truth: run the tests! 😬
+
+```bash
+NVIDIA_TF32_OVERRIDE=0 RUN_SLOW=1 RUN_PT_TF_CROSS_TESTS=1 \
+py.test -vv tests/models/brand_new_bert/test_modeling_tf_brand_new_bert.py
+```
+
+The most likely outcome is that you'll see a bunch of errors. Don't worry, this is expected! Debugging ML models is
+notoriously hard, and the key ingredient to success is patience (and `breakpoint()`). In our experience, the hardest
+problems arise from subtle mismatches between ML frameworks, for which we have a few pointers at the end of this guide.
+In other cases, a general test might not be directly applicable to your model, in which case we suggest an override
+at the model test class level. Regardless of the issue, don't hesitate to ask for help in your draft pull request if
+you're stuck.
+
+When all tests pass, congratulations, your model is nearly ready to be added to the 🤗 Transformers library! 🎉
+
+### 6.-7. Ensure everyone can use your model
+
+**6. Submit the pull request**
+
+Once you're done with the implementation and the tests, it's time to submit a pull request. Before pushing your code,
+run our code formatting utility, `make fixup` 🪄. This will automatically fix any formatting issues, which would cause
+our automatic checks to fail.
+
+It's now time to convert your draft pull request into a real pull request. To do so, click on the "Ready for
+review" button and add Joao (`@gante`) and Matt (`@Rocketknight1`) as reviewers. A model pull request will need
+at least 3 reviewers, but they will take care of finding appropriate additional reviewers for your model.
+
+After all reviewers are happy with the state of your PR, the final action point is to remove the `from_pt=True` flag in
+`.from_pretrained()` calls. Since there are no TensorFlow weights, you will have to add them! Check the section
+below for instructions on how to do it.
+
+Finally, when the TensorFlow weights get merged, you have at least 3 reviewer approvals, and all CI checks are
+green, double-check the tests locally one last time
+
+```bash
+NVIDIA_TF32_OVERRIDE=0 RUN_SLOW=1 RUN_PT_TF_CROSS_TESTS=1 \
+py.test -vv tests/models/brand_new_bert/test_modeling_tf_brand_new_bert.py
+```
+
+and we will merge your PR! Congratulations on the milestone 🎉
+
+**7. (Optional) Build demos and share with the world**
+
+One of the hardest parts about open-source is discovery. How can the other users learn about the existence of your
+fabulous TensorFlow contribution? With proper communication, of course! 📣
+
+There are two main ways to share your model with the community:
+- Build demos. These include Gradio demos, notebooks, and other fun ways to show off your model. We highly
+   encourage you to add a notebook to our [community-driven demos](https://huggingface.co/docs/transformers/community).
+- Share stories on social media like Twitter and LinkedIn. You should be proud of your work and share
+   your achievement with the community - your model can now be used by thousands of engineers and researchers around
+   the world 🌍! We will be happy to retweet your posts and help you share your work with the community.
+
+
+## Adding TensorFlow weights to 🤗 Hub
+
+Assuming that the TensorFlow model architecture is available in 🤗 Transformers, converting PyTorch weights into
+TensorFlow weights is a breeze!
+
+Here's how to do it:
+1. Make sure you are logged into your Hugging Face account in your terminal. You can log in using the command
+   `huggingface-cli login` (you can find your access tokens [here](https://huggingface.co/settings/tokens))
+2. Run `transformers-cli pt-to-tf --model-name foo/bar`, where `foo/bar` is the name of the model repository
+   containing the PyTorch weights you want to convert
+3. Tag `@joaogante` and `@Rocketknight1` in the 🤗 Hub PR the command above has just created
+
+That's it! 🎉
+
+
+## Debugging mismatches across ML frameworks 🐛
+
+At some point, when adding a new architecture or when creating TensorFlow weights for an existing architecture, you
+might come across errors compaining about mismatches between PyTorch and TensorFlow. You might even decide to open the
+model architecture code for the two frameworks, and find that they look identical. What's going on? 🤔
+
+First of all, let's talk about why understanding these mismatches matters. Many community members will use 🤗
+Transformers models out of the box, and trust that our models behave as expected. When there is a large mismatch
+between the two frameworks, it implies that the model is not following the reference implementation for at least one
+of the frameworks. This might lead to silent failures, in which the model runs but has poor performance. This is
+arguably worse than a model that fails to run at all! To that end, we aim at having a framework mismatch smaller than
+`1e-5` at all stages of the model.
+
+As in other numerical problems, the devil is in the details. And as in any detail-oriented craft, the secret
+ingredient here is patience. Here is our suggested workflow for when you come across this type of issues:
+1. Locate the source of mismatches. The model you're converting probably has near identical inner variables up to a
+   certain point. Place `breakpoint()` statements in the two frameworks' architectures, and compare the values of the
+   numerical variables in a top-down fashion until you find the source of the problems.
+2. Now that you've pinpointed the source of the issue, get in touch with the 🤗 Transformers team. It is possible
+   that we've seen a similar problem before and can promptly provide a solution. As a fallback, scan popular pages
+   like StackOverflow and GitHub issues.
+3. If there is no solution in sight, it means you'll have to go deeper. The good news is that you've located the
+   issue, so you can focus on the problematic instruction, abstracting away the rest of the model! The bad news is
+   that you'll have to venture into the source implementation of said instruction. In some cases, you might find an
+   issue with a reference implementation - don't abstain from opening an issue in the upstream repository.
+
+In some cases, in dicussion with the 🤗 Transformers team, we might find that the fixing the mismatch is infeasible.
+When the mismatch is very small in the output layers of the model (but potentially large in the hidden states), we
+might decide to ignore it in favor of distributing the model. The `pt-to-tf` CLI mentioned above has a `--max-error`
+flag to override the error message at weight conversion time.

docs/source/en/autoclass_tutorial.mdx

@@ -12,7 +12,7 @@ specific language governing permissions and limitations under the License.
 
 # Load pretrained instances with an AutoClass
 
-With so many different Transformer architectures, it can be challenging to create one for your checkpoint. As a part of 🤗 Transformers core philosophy to make the library easy, simple and flexible to use, an `AutoClass` automatically infer and load the correct architecture from a given checkpoint. The `from_pretrained` method lets you quickly load a pretrained model for any architecture so you don't have to devote time and resources to train a model from scratch. Producing this type of checkpoint-agnostic code means if your code works for one checkpoint, it will work with another checkpoint - as long as it was trained for a similar task - even if the architecture is different.
+With so many different Transformer architectures, it can be challenging to create one for your checkpoint. As a part of 🤗 Transformers core philosophy to make the library easy, simple and flexible to use, an `AutoClass` automatically infer and load the correct architecture from a given checkpoint. The `from_pretrained()` method lets you quickly load a pretrained model for any architecture so you don't have to devote time and resources to train a model from scratch. Producing this type of checkpoint-agnostic code means if your code works for one checkpoint, it will work with another checkpoint - as long as it was trained for a similar task - even if the architecture is different.
 
 <Tip>
 
@@ -95,6 +95,14 @@ Easily reuse the same checkpoint to load an architecture for a different task:
 >>> model = AutoModelForTokenClassification.from_pretrained("distilbert-base-uncased")
 ```
 
+<Tip warning={true}>
+
+For PyTorch models, the `from_pretrained()` method uses `torch.load()` which internally uses `pickle` and is known to be insecure. In general, never load a model that could have come from an untrusted source, or that could have been tampered with. This security risk is partially mitigated for public models hosted on the Hugging Face Hub, which are [scanned for malware](https://huggingface.co/docs/hub/security-malware) at each commit. See the [Hub documentation](https://huggingface.co/docs/hub/security) for best practices like [signed commit verification](https://huggingface.co/docs/hub/security-gpg#signing-commits-with-gpg) with GPG.
+
+TensorFlow and Flax checkpoints are not affected, and can be loaded within PyTorch architectures using the `from_tf` and `from_flax` kwargs for the `from_pretrained` method to circumvent this issue.
+
+</Tip>
+
 Generally, we recommend using the `AutoTokenizer` class and the `AutoModelFor` class to load pretrained instances of models. This will ensure you load the correct architecture every time. In the next [tutorial](preprocessing), learn how to use your newly loaded tokenizer, feature extractor and processor to preprocess a dataset for fine-tuning.
 </pt>
 <tf>

docs/source/en/converting_tensorflow_models.mdx

@@ -10,7 +10,7 @@ an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express o
 specific language governing permissions and limitations under the License.
 -->
 
-# Converting Tensorflow Checkpoints
+# Converting From Tensorflow Checkpoints
 
 A command-line interface is provided to convert original Bert/GPT/GPT-2/Transformer-XL/XLNet/XLM checkpoints to models
 that can be loaded using the `from_pretrained` methods of the library.

docs/source/en/debugging.mdx

@@ -77,7 +77,7 @@ to the normal command line arguments, or pass `debug="underflow_overflow"` when
 If you're using your own training loop or another Trainer you can accomplish the same with:
 
 ```python
-from .debug_utils import DebugUnderflowOverflow
+from transformers.debug_utils import DebugUnderflowOverflow
 
 debug_overflow = DebugUnderflowOverflow(model)
 ```
@@ -271,7 +271,7 @@ Additionally, if you're instantiating the debugger in your own code, you can adj
 its default, e.g.:
 
 ```python
-from .debug_utils import DebugUnderflowOverflow
+from transformers.debug_utils import DebugUnderflowOverflow
 
 debug_overflow = DebugUnderflowOverflow(model, max_frames_to_save=100)
 ```

docs/source/en/glossary.mdx

@@ -44,7 +44,7 @@ specific language governing permissions and limitations under the License.
 Every model is different yet bears similarities with the others. Therefore most models use the same inputs, which are
 detailed here alongside usage examples.
 
-<a id='input-ids'></a>
+
 
 ### Input IDs
 
@@ -113,7 +113,7 @@ we will see
 
 because this is the way a [`BertModel`] is going to expect its inputs.
 
-<a id='attention-mask'></a>
+
 
 ### Attention mask
 
@@ -171,7 +171,7 @@ in the dictionary returned by the tokenizer under the key "attention_mask":
 [[1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]]
 ```
 
-<a id='token-type-ids'></a>
+
 
 ### Token Type IDs
 
@@ -224,7 +224,7 @@ second sequence, corresponding to the "question", has all its tokens represented
 
 Some models, like [`XLNetModel`] use an additional token represented by a `2`.
 
-<a id='position-ids'></a>
+
 
 ### Position IDs
 
@@ -238,7 +238,7 @@ absolute positional embeddings.
 Absolute positional embeddings are selected in the range `[0, config.max_position_embeddings - 1]`. Some models use
 other types of positional embeddings, such as sinusoidal position embeddings or relative position embeddings.
 
-<a id='labels'></a>
+
 
 ### Labels
 
@@ -266,7 +266,7 @@ These labels are different according to the model head, for example:
 The base models (e.g., [`BertModel`]) do not accept labels, as these are the base transformer
 models, simply outputting features.
 
-<a id='decoder-input-ids'></a>
+
 
 ### Decoder input IDs
 
@@ -279,7 +279,6 @@ such models, passing the `labels` is the preferred way to handle training.
 
 Please check each model's docs to see how they handle these input IDs for sequence to sequence training.
 
-<a id='feed-forward-chunking'></a>
 
 ### Feed Forward Chunking
 

docs/source/en/hpo_train.mdx

@@ -0,0 +1,120 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+-->
+
+# Hyperparameter Search using Trainer API
+
+🤗 Transformers provides a [`Trainer`] class optimized for training 🤗 Transformers models, making it easier to start training without manually writing your own training loop. The [`Trainer`] provides API for hyperparameter search. This doc shows how to enable it in example. 
+
+## Hyperparameter Search backend
+
+[`Trainer`] supports four hyperparameter search backends currently:
+[optuna](https://optuna.org/), [sigopt](https://sigopt.com/), [raytune](https://docs.ray.io/en/latest/tune/index.html) and [wandb](https://wandb.ai/site/sweeps).
+
+you should install them before using them as the hyperparameter search backend
+```bash
+pip install optuna/sigopt/wandb/ray[tune] 
+```
+
+## How to enable Hyperparameter search in example
+
+Define the hyperparameter search space, different backends need different format.
+
+For sigopt, see sigopt [object_parameter](https://docs.sigopt.com/ai-module-api-references/api_reference/objects/object_parameter), it's like following:
+```py
+>>> def sigopt_hp_space(trial):
+...     return [
+...         {"bounds": {"min": 1e-6, "max": 1e-4}, "name": "learning_rate", "type": "double"},
+...         {
+...             "categorical_values": ["16", "32", "64", "128"],
+...             "name": "per_device_train_batch_size",
+...             "type": "categorical",
+...         },
+...     ]
+```
+
+For optuna, see optuna [object_parameter](https://optuna.readthedocs.io/en/stable/tutorial/10_key_features/002_configurations.html#sphx-glr-tutorial-10-key-features-002-configurations-py), it's like following:
+
+```py
+>>> def optuna_hp_space(trial):
+...     return {
+...         "learning_rate": trial.suggest_float("learning_rate", 1e-6, 1e-4, log=True),
+...         "per_device_train_batch_size": trial.suggest_categorical("per_device_train_batch_size", [16, 32, 64, 128]),
+...     }
+```
+
+For raytune, see raytune [object_parameter](https://docs.ray.io/en/latest/tune/api_docs/search_space.html), it's like following:
+
+```py
+>>> def ray_hp_space(trial):
+...     return {
+...         "learning_rate": tune.loguniform(1e-6, 1e-4),
+...         "per_device_train_batch_size": tune.choice([16, 32, 64, 128]),
+...     }
+```
+
+For wandb, see wandb [object_parameter](https://docs.wandb.ai/guides/sweeps/configuration), it's like following:
+
+```py
+>>> def wandb_hp_space(trial):
+...     return {
+...         "method": "random",
+...         "metric": {"name": "objective", "goal": "minimize"},
+...         "parameters": {
+...             "learning_rate": {"distribution": "uniform", "min": 1e-6, "max": 1e-4},
+...             "per_device_train_batch_size": {"values": [16, 32, 64, 128]},
+...         },
+...     }
+```
+
+Define a `model_init` function and pass it to the [`Trainer`], as an example:
+```py
+>>> def model_init(trial):
+...     return AutoModelForSequenceClassification.from_pretrained(
+...         model_args.model_name_or_path,
+...         from_tf=bool(".ckpt" in model_args.model_name_or_path),
+...         config=config,
+...         cache_dir=model_args.cache_dir,
+...         revision=model_args.model_revision,
+...         use_auth_token=True if model_args.use_auth_token else None,
+...     )
+```
+
+Create a [`Trainer`] with your `model_init` function, training arguments, training and test datasets, and evaluation function:
+
+```py
+>>> trainer = Trainer(
+...     model=None,
+...     args=training_args,
+...     train_dataset=small_train_dataset,
+...     eval_dataset=small_eval_dataset,
+...     compute_metrics=compute_metrics,
+...     tokenizer=tokenizer,
+...     model_init=model_init,
+...     data_collator=data_collator,
+... )
+```
+
+Call hyperparameter search, get the best trial parameters, backend could be `"optuna"`/`"sigopt"`/`"wandb"`/`"ray"`. direction can be`"minimize"` or `"maximize"`, which indicates whether to optimize greater or lower objective.
+
+You could define your own compute_objective function, if not defined, the default compute_objective will be called, and the sum of eval metric like f1 is returned as objective value.
+
+```py
+>>> best_trial = trainer.hyperparameter_search(
+...     direction="maximize",
+...     backend="optuna",
+...     hp_space=optuna_hp_space,
+...     n_trials=20,
+...     compute_objective=compute_objective,
+... )
+```
+
+## Hyperparameter search For DDP refinetune
+Currently, Hyperparameter search for DDP is enabled for optuna and sigopt. Only the rank-zero process will generate the search trial and pass the argument to other ranks.

docs/source/en/index.mdx

@@ -12,18 +12,18 @@ specific language governing permissions and limitations under the License.
 
 # 🤗 Transformers
 
-State-of-the-art Machine Learning for PyTorch, TensorFlow and JAX.
+State-of-the-art Machine Learning for [PyTorch](https://pytorch.org/), [TensorFlow](https://www.tensorflow.org/), and [JAX](https://jax.readthedocs.io/en/latest/).
 
-🤗 Transformers provides APIs to easily download and train state-of-the-art pretrained models. Using pretrained models can reduce your compute costs, carbon footprint, and save you time from training a model from scratch. The models can be used across different modalities such as:
+🤗 Transformers provides APIs and tools to easily download and train state-of-the-art pretrained models. Using pretrained models can reduce your compute costs, carbon footprint, and save you the time and resources required to train a model from scratch. These models support common tasks in different modalities, such as:
 
-* 📝 Text: text classification, information extraction, question answering, summarization, translation, and text generation in over 100 languages.
-* 🖼️ Images: image classification, object detection, and segmentation.
-* 🗣️ Audio: speech recognition and audio classification.
-* 🐙 Multimodal: table question answering, optical character recognition, information extraction from scanned documents, video classification, and visual question answering.
+📝 **Natural Language Processing**: text classification, named entity recognition, question answering, language modeling, summarization, translation, multiple choice, and text generation.<br>
+🖼️ **Computer Vision**: image classification, object detection, and segmentation.<br>
+🗣️ **Audio**: automatic speech recognition and audio classification.<br>
+🐙 **Multimodal**: table question answering, optical character recognition, information extraction from scanned documents, video classification, and visual question answering.
 
-Our library supports seamless integration between three of the most popular deep learning libraries: [PyTorch](https://pytorch.org/), [TensorFlow](https://www.tensorflow.org/) and [JAX](https://jax.readthedocs.io/en/latest/). Train your model in three lines of code in one framework, and load it for inference with another.
+🤗 Transformers support framework interoperability between PyTorch, TensorFlow, and JAX. This provides the flexibility to use a different framework at each stage of a model's life; train a model in three lines of code in one framework, and load it for inference in another. Models can also be exported to a format like ONNX and TorchScript for deployment in production environments.
 
-Each 🤗 Transformers architecture is defined in a standalone Python module so they can be easily customized for research and experiments.
+Join the growing community on the [Hub](https://huggingface.co/models), [forum](https://discuss.huggingface.co/), or [Discord](https://discord.com/invite/JfAtkvEtRb) today!
 
 ## If you are looking for custom support from the Hugging Face team
 
@@ -33,19 +33,17 @@ Each 🤗 Transformers architecture is defined in a standalone Python module so
 
 ## Contents
 
-The documentation is organized in five parts:
+The documentation is organized into five sections:
 
-- **GET STARTED** contains a quick tour and installation instructions to get up and running with 🤗 Transformers.
-- **TUTORIALS** are a great place to begin if you are new to our library. This section will help you gain the basic skills you need to start using 🤗 Transformers.
-- **HOW-TO GUIDES** will show you how to achieve a specific goal like fine-tuning a pretrained model for language modeling or how to create a custom model head.
-- **CONCEPTUAL GUIDES** provides more discussion and explanation of the underlying concepts and ideas behind models, tasks, and the design philosophy of 🤗 Transformers. 
-- **API** describes each class and function, grouped in:
+- **GET STARTED** provides a quick tour of the library and installation instructions to get up and running.
+- **TUTORIALS** are a great place to start if you're a beginner. This section will help you gain the basic skills you need to start using the library.
+- **HOW-TO GUIDES** show you how to achieve a specific goal, like finetuning a pretrained model for language modeling or how to write and share a custom model.
+- **CONCEPTUAL GUIDES** offers more discussion and explanation of the underlying concepts and ideas behind models, tasks, and the design philosophy of 🤗 Transformers.
+- **API** describes all classes and functions:
 
-  - **MAIN CLASSES** for the main classes exposing the important APIs of the library.
-  - **MODELS** for the classes and functions related to each model implemented in the library.
-  - **INTERNAL HELPERS** for the classes and functions we use internally.
-
-The library currently contains JAX, PyTorch and TensorFlow implementations, pretrained model weights, usage scripts and conversion utilities for the following models.
+  - **MAIN CLASSES** details the most important classes like configuration, model, tokenizer, and pipeline.
+  - **MODELS** details the classes and functions related to each model implemented in the library.
+  - **INTERNAL HELPERS** details utility classes and functions used internally.
 
 ### Supported models
 
@@ -70,6 +68,7 @@ The library currently contains JAX, PyTorch and TensorFlow implementations, pret
 1. **[CANINE](model_doc/canine)** (from Google Research) released with the paper [CANINE: Pre-training an Efficient Tokenization-Free Encoder for Language Representation](https://arxiv.org/abs/2103.06874) by Jonathan H. Clark, Dan Garrette, Iulia Turc, John Wieting.
 1. **[CLIP](model_doc/clip)** (from OpenAI) released with the paper [Learning Transferable Visual Models From Natural Language Supervision](https://arxiv.org/abs/2103.00020) by Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, Ilya Sutskever.
 1. **[CodeGen](model_doc/codegen)** (from Salesforce) released with the paper [A Conversational Paradigm for Program Synthesis](https://arxiv.org/abs/2203.13474) by Erik Nijkamp, Bo Pang, Hiroaki Hayashi, Lifu Tu, Huan Wang, Yingbo Zhou, Silvio Savarese, Caiming Xiong.
+1. **[Conditional DETR](model_doc/conditional_detr)** (from Microsoft Research Asia) released with the paper [Conditional DETR for Fast Training Convergence](https://arxiv.org/abs/2108.06152) by Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, Jingdong Wang.
 1. **[ConvBERT](model_doc/convbert)** (from YituTech) released with the paper [ConvBERT: Improving BERT with Span-based Dynamic Convolution](https://arxiv.org/abs/2008.02496) by Zihang Jiang, Weihao Yu, Daquan Zhou, Yunpeng Chen, Jiashi Feng, Shuicheng Yan.
 1. **[ConvNeXT](model_doc/convnext)** (from Facebook AI) released with the paper [A ConvNet for the 2020s](https://arxiv.org/abs/2201.03545) by Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie.
 1. **[CPM](model_doc/cpm)** (from Tsinghua University) released with the paper [CPM: A Large-scale Generative Chinese Pre-trained Language Model](https://arxiv.org/abs/2012.00413) by Zhengyan Zhang, Xu Han, Hao Zhou, Pei Ke, Yuxian Gu, Deming Ye, Yujia Qin, Yusheng Su, Haozhe Ji, Jian Guan, Fanchao Qi, Xiaozhi Wang, Yanan Zheng, Guoyang Zeng, Huanqi Cao, Shengqi Chen, Daixuan Li, Zhenbo Sun, Zhiyuan Liu, Minlie Huang, Wentao Han, Jie Tang, Juanzi Li, Xiaoyan Zhu, Maosong Sun.
@@ -79,15 +78,19 @@ The library currently contains JAX, PyTorch and TensorFlow implementations, pret
 1. **[DeBERTa](model_doc/deberta)** (from Microsoft) released with the paper [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://arxiv.org/abs/2006.03654) by Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen.
 1. **[DeBERTa-v2](model_doc/deberta-v2)** (from Microsoft) released with the paper [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://arxiv.org/abs/2006.03654) by Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen.
 1. **[Decision Transformer](model_doc/decision_transformer)** (from Berkeley/Facebook/Google) released with the paper [Decision Transformer: Reinforcement Learning via Sequence Modeling](https://arxiv.org/abs/2106.01345) by Lili Chen, Kevin Lu, Aravind Rajeswaran, Kimin Lee, Aditya Grover, Michael Laskin, Pieter Abbeel, Aravind Srinivas, Igor Mordatch.
+1. **[Deformable DETR](model_doc/deformable_detr)** (from SenseTime Research) released with the paper [Deformable DETR: Deformable Transformers for End-to-End Object Detection](https://arxiv.org/abs/2010.04159) by Xizhou Zhu, Weijie Su, Lewei Lu, Bin Li, Xiaogang Wang, Jifeng Dai.
 1. **[DeiT](model_doc/deit)** (from Facebook) released with the paper [Training data-efficient image transformers & distillation through attention](https://arxiv.org/abs/2012.12877) by Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, Hervé Jégou.
 1. **[DETR](model_doc/detr)** (from Facebook) released with the paper [End-to-End Object Detection with Transformers](https://arxiv.org/abs/2005.12872) by Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, Sergey Zagoruyko.
 1. **[DialoGPT](model_doc/dialogpt)** (from Microsoft Research) released with the paper [DialoGPT: Large-Scale Generative Pre-training for Conversational Response Generation](https://arxiv.org/abs/1911.00536) by Yizhe Zhang, Siqi Sun, Michel Galley, Yen-Chun Chen, Chris Brockett, Xiang Gao, Jianfeng Gao, Jingjing Liu, Bill Dolan.
 1. **[DistilBERT](model_doc/distilbert)** (from HuggingFace), released together with the paper [DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter](https://arxiv.org/abs/1910.01108) by Victor Sanh, Lysandre Debut and Thomas Wolf. The same method has been applied to compress GPT2 into [DistilGPT2](https://github.com/huggingface/transformers/tree/main/examples/research_projects/distillation), RoBERTa into [DistilRoBERTa](https://github.com/huggingface/transformers/tree/main/examples/research_projects/distillation), Multilingual BERT into [DistilmBERT](https://github.com/huggingface/transformers/tree/main/examples/research_projects/distillation) and a German version of DistilBERT.
 1. **[DiT](model_doc/dit)** (from Microsoft Research) released with the paper [DiT: Self-supervised Pre-training for Document Image Transformer](https://arxiv.org/abs/2203.02378) by Junlong Li, Yiheng Xu, Tengchao Lv, Lei Cui, Cha Zhang, Furu Wei.
+1. **[Donut](model_doc/donut)** (from NAVER), released together with the paper [OCR-free Document Understanding Transformer](https://arxiv.org/abs/2111.15664) by Geewook Kim, Teakgyu Hong, Moonbin Yim, Jeongyeon Nam, Jinyoung Park, Jinyeong Yim, Wonseok Hwang, Sangdoo Yun, Dongyoon Han, Seunghyun Park.
 1. **[DPR](model_doc/dpr)** (from Facebook) released with the paper [Dense Passage Retrieval for Open-Domain Question Answering](https://arxiv.org/abs/2004.04906) by Vladimir Karpukhin, Barlas Oğuz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih.
 1. **[DPT](master/model_doc/dpt)** (from Intel Labs) released with the paper [Vision Transformers for Dense Prediction](https://arxiv.org/abs/2103.13413) by René Ranftl, Alexey Bochkovskiy, Vladlen Koltun.
 1. **[ELECTRA](model_doc/electra)** (from Google Research/Stanford University) released with the paper [ELECTRA: Pre-training text encoders as discriminators rather than generators](https://arxiv.org/abs/2003.10555) by Kevin Clark, Minh-Thang Luong, Quoc V. Le, Christopher D. Manning.
 1. **[EncoderDecoder](model_doc/encoder-decoder)** (from Google Research) released with the paper [Leveraging Pre-trained Checkpoints for Sequence Generation Tasks](https://arxiv.org/abs/1907.12461) by Sascha Rothe, Shashi Narayan, Aliaksei Severyn.
+1. **[ERNIE](model_doc/ernie)** (from Baidu) released with the paper [ERNIE: Enhanced Representation through Knowledge Integration](https://arxiv.org/abs/1904.09223) by Yu Sun, Shuohuan Wang, Yukun Li, Shikun Feng, Xuyi Chen, Han Zhang, Xin Tian, Danxiang Zhu, Hao Tian, Hua Wu.
+1. **[ESM](model_doc/esm)** (from Meta AI) are transformer protein language models.  **ESM-1b** was released with the paper [Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences](https://www.pnas.org/content/118/15/e2016239118) by Alexander Rives, Joshua Meier, Tom Sercu, Siddharth Goyal, Zeming Lin, Jason Liu, Demi Guo, Myle Ott, C. Lawrence Zitnick, Jerry Ma, and Rob Fergus. **ESM-1v** was released with the paper [Language models enable zero-shot prediction of the effects of mutations on protein function](https://doi.org/10.1101/2021.07.09.450648) by Joshua Meier, Roshan Rao, Robert Verkuil, Jason Liu, Tom Sercu and Alexander Rives. **ESM-2** was released with the paper [Language models of protein sequences at the scale of evolution enable accurate structure prediction](https://doi.org/10.1101/2022.07.20.500902) by Zeming Lin, Halil Akin, Roshan Rao, Brian Hie, Zhongkai Zhu, Wenting Lu, Allan dos Santos Costa, Maryam Fazel-Zarandi, Tom Sercu, Sal Candido, Alexander Rives.
 1. **[FlauBERT](model_doc/flaubert)** (from CNRS) released with the paper [FlauBERT: Unsupervised Language Model Pre-training for French](https://arxiv.org/abs/1912.05372) by Hang Le, Loïc Vial, Jibril Frej, Vincent Segonne, Maximin Coavoux, Benjamin Lecouteux, Alexandre Allauzen, Benoît Crabbé, Laurent Besacier, Didier Schwab.
 1. **[FLAVA](model_doc/flava)** (from Facebook AI) released with the paper [FLAVA: A Foundational Language And Vision Alignment Model](https://arxiv.org/abs/2112.04482) by Amanpreet Singh, Ronghang Hu, Vedanuj Goswami, Guillaume Couairon, Wojciech Galuba, Marcus Rohrbach, and Douwe Kiela.
 1. **[FNet](model_doc/fnet)** (from Google Research) released with the paper [FNet: Mixing Tokens with Fourier Transforms](https://arxiv.org/abs/2105.03824) by James Lee-Thorp, Joshua Ainslie, Ilya Eckstein, Santiago Ontanon.
@@ -96,6 +99,7 @@ The library currently contains JAX, PyTorch and TensorFlow implementations, pret
 1. **[GPT](model_doc/openai-gpt)** (from OpenAI) released with the paper [Improving Language Understanding by Generative Pre-Training](https://blog.openai.com/language-unsupervised/) by Alec Radford, Karthik Narasimhan, Tim Salimans and Ilya Sutskever.
 1. **[GPT Neo](model_doc/gpt_neo)** (from EleutherAI) released in the repository [EleutherAI/gpt-neo](https://github.com/EleutherAI/gpt-neo) by Sid Black, Stella Biderman, Leo Gao, Phil Wang and Connor Leahy.
 1. **[GPT NeoX](model_doc/gpt_neox)** (from EleutherAI) released with the paper [GPT-NeoX-20B: An Open-Source Autoregressive Language Model](https://arxiv.org/abs/2204.06745) by Sid Black, Stella Biderman, Eric Hallahan, Quentin Anthony, Leo Gao, Laurence Golding, Horace He, Connor Leahy, Kyle McDonell, Jason Phang, Michael Pieler, USVSN Sai Prashanth, Shivanshu Purohit, Laria Reynolds, Jonathan Tow, Ben Wang, Samuel Weinbach
+1. **[GPT NeoX Japanese](model_doc/gpt_neox_japanese)** (from ABEJA) released by Shinya Otani, Takayoshi Makabe, Anuj Arora, and Kyo Hattori.
 1. **[GPT-2](model_doc/gpt2)** (from OpenAI) released with the paper [Language Models are Unsupervised Multitask Learners](https://blog.openai.com/better-language-models/) by Alec Radford*, Jeffrey Wu*, Rewon Child, David Luan, Dario Amodei** and Ilya Sutskever**.
 1. **[GPT-J](model_doc/gptj)** (from EleutherAI) released in the repository [kingoflolz/mesh-transformer-jax](https://github.com/kingoflolz/mesh-transformer-jax/) by Ben Wang and Aran Komatsuzaki.
 1. **[GroupViT](model_doc/groupvit)** (from UCSD, NVIDIA) released with the paper [GroupViT: Semantic Segmentation Emerges from Text Supervision](https://arxiv.org/abs/2202.11094) by Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, Xiaolong Wang.
@@ -105,7 +109,7 @@ The library currently contains JAX, PyTorch and TensorFlow implementations, pret
 1. **[LayoutLM](model_doc/layoutlm)** (from Microsoft Research Asia) released with the paper [LayoutLM: Pre-training of Text and Layout for Document Image Understanding](https://arxiv.org/abs/1912.13318) by Yiheng Xu, Minghao Li, Lei Cui, Shaohan Huang, Furu Wei, Ming Zhou.
 1. **[LayoutLMv2](model_doc/layoutlmv2)** (from Microsoft Research Asia) released with the paper [LayoutLMv2: Multi-modal Pre-training for Visually-Rich Document Understanding](https://arxiv.org/abs/2012.14740) by Yang Xu, Yiheng Xu, Tengchao Lv, Lei Cui, Furu Wei, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Wanxiang Che, Min Zhang, Lidong Zhou.
 1. **[LayoutLMv3](model_doc/layoutlmv3)** (from Microsoft Research Asia) released with the paper [LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking](https://arxiv.org/abs/2204.08387) by Yupan Huang, Tengchao Lv, Lei Cui, Yutong Lu, Furu Wei.
-1. **[LayoutXLM](model_doc/layoutlmv2)** (from Microsoft Research Asia) released with the paper [LayoutXLM: Multimodal Pre-training for Multilingual Visually-rich Document Understanding](https://arxiv.org/abs/2104.08836) by Yiheng Xu, Tengchao Lv, Lei Cui, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Furu Wei.
+1. **[LayoutXLM](model_doc/layoutxlm)** (from Microsoft Research Asia) released with the paper [LayoutXLM: Multimodal Pre-training for Multilingual Visually-rich Document Understanding](https://arxiv.org/abs/2104.08836) by Yiheng Xu, Tengchao Lv, Lei Cui, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Furu Wei.
 1. **[LED](model_doc/led)** (from AllenAI) released with the paper [Longformer: The Long-Document Transformer](https://arxiv.org/abs/2004.05150) by Iz Beltagy, Matthew E. Peters, Arman Cohan.
 1. **[LeViT](model_doc/levit)** (from Meta AI) released with the paper [LeViT: A Vision Transformer in ConvNet's Clothing for Faster Inference](https://arxiv.org/abs/2104.01136) by Ben Graham, Alaaeldin El-Nouby, Hugo Touvron, Pierre Stock, Armand Joulin, Hervé Jégou, Matthijs Douze.
 1. **[Longformer](model_doc/longformer)** (from AllenAI) released with the paper [Longformer: The Long-Document Transformer](https://arxiv.org/abs/2004.05150) by Iz Beltagy, Matthew E. Peters, Arman Cohan.
@@ -115,6 +119,7 @@ The library currently contains JAX, PyTorch and TensorFlow implementations, pret
 1. **[M-CTC-T](model_doc/mctct)** (from Facebook) released with the paper [Pseudo-Labeling For Massively Multilingual Speech Recognition](https://arxiv.org/abs/2111.00161) by Loren Lugosch, Tatiana Likhomanenko, Gabriel Synnaeve, and Ronan Collobert.
 1. **[M2M100](model_doc/m2m_100)** (from Facebook) released with the paper [Beyond English-Centric Multilingual Machine Translation](https://arxiv.org/abs/2010.11125) by Angela Fan, Shruti Bhosale, Holger Schwenk, Zhiyi Ma, Ahmed El-Kishky, Siddharth Goyal, Mandeep Baines, Onur Celebi, Guillaume Wenzek, Vishrav Chaudhary, Naman Goyal, Tom Birch, Vitaliy Liptchinsky, Sergey Edunov, Edouard Grave, Michael Auli, Armand Joulin.
 1. **[MarianMT](model_doc/marian)** Machine translation models trained using [OPUS](http://opus.nlpl.eu/) data by Jörg Tiedemann. The [Marian Framework](https://marian-nmt.github.io/) is being developed by the Microsoft Translator Team.
+1. **[MarkupLM](model_doc/markuplm)** (from Microsoft Research Asia) released with the paper [MarkupLM: Pre-training of Text and Markup Language for Visually-rich Document Understanding](https://arxiv.org/abs/2110.08518) by Junlong Li, Yiheng Xu, Lei Cui, Furu Wei.
 1. **[MaskFormer](model_doc/maskformer)** (from Meta and UIUC) released with the paper [Per-Pixel Classification is Not All You Need for Semantic Segmentation](https://arxiv.org/abs/2107.06278) by Bowen Cheng, Alexander G. Schwing, Alexander Kirillov.
 1. **[mBART](model_doc/mbart)** (from Facebook) released with the paper [Multilingual Denoising Pre-training for Neural Machine Translation](https://arxiv.org/abs/2001.08210) by Yinhan Liu, Jiatao Gu, Naman Goyal, Xian Li, Sergey Edunov, Marjan Ghazvininejad, Mike Lewis, Luke Zettlemoyer.
 1. **[mBART-50](model_doc/mbart)** (from Facebook) released with the paper [Multilingual Translation with Extensible Multilingual Pretraining and Finetuning](https://arxiv.org/abs/2008.00401) by Yuqing Tang, Chau Tran, Xian Li, Peng-Jen Chen, Naman Goyal, Vishrav Chaudhary, Jiatao Gu, Angela Fan.
@@ -132,6 +137,7 @@ The library currently contains JAX, PyTorch and TensorFlow implementations, pret
 1. **[OPT](master/model_doc/opt)** (from Meta AI) released with the paper [OPT: Open Pre-trained Transformer Language Models](https://arxiv.org/abs/2205.01068) by Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen et al.
 1. **[OWL-ViT](model_doc/owlvit)** (from Google AI) released with the paper [Simple Open-Vocabulary Object Detection with Vision Transformers](https://arxiv.org/abs/2205.06230) by Matthias Minderer, Alexey Gritsenko, Austin Stone, Maxim Neumann, Dirk Weissenborn, Alexey Dosovitskiy, Aravindh Mahendran, Anurag Arnab, Mostafa Dehghani, Zhuoran Shen, Xiao Wang, Xiaohua Zhai, Thomas Kipf, and Neil Houlsby.
 1. **[Pegasus](model_doc/pegasus)** (from Google) released with the paper [PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization](https://arxiv.org/abs/1912.08777) by Jingqing Zhang, Yao Zhao, Mohammad Saleh and Peter J. Liu.
+1. **[PEGASUS-X](model_doc/pegasus_x)** (from Google) released with the paper [Investigating Efficiently Extending Transformers for Long Input Summarization](https://arxiv.org/abs/2208.04347) by Jason Phang, Yao Zhao, and Peter J. Liu.
 1. **[Perceiver IO](model_doc/perceiver)** (from Deepmind) released with the paper [Perceiver IO: A General Architecture for Structured Inputs & Outputs](https://arxiv.org/abs/2107.14795) by Andrew Jaegle, Sebastian Borgeaud, Jean-Baptiste Alayrac, Carl Doersch, Catalin Ionescu, David Ding, Skanda Koppula, Daniel Zoran, Andrew Brock, Evan Shelhamer, Olivier Hénaff, Matthew M. Botvinick, Andrew Zisserman, Oriol Vinyals, João Carreira.
 1. **[PhoBERT](model_doc/phobert)** (from VinAI Research) released with the paper [PhoBERT: Pre-trained language models for Vietnamese](https://www.aclweb.org/anthology/2020.findings-emnlp.92/) by Dat Quoc Nguyen and Anh Tuan Nguyen.
 1. **[PLBart](model_doc/plbart)** (from UCLA NLP) released with the paper [Unified Pre-training for Program Understanding and Generation](https://arxiv.org/abs/2103.06333) by Wasi Uddin Ahmad, Saikat Chakraborty, Baishakhi Ray, Kai-Wei Chang.
@@ -159,6 +165,7 @@ The library currently contains JAX, PyTorch and TensorFlow implementations, pret
 1. **[T5v1.1](model_doc/t5v1.1)** (from Google AI) released in the repository [google-research/text-to-text-transfer-transformer](https://github.com/google-research/text-to-text-transfer-transformer/blob/main/released_checkpoints.md#t511) by Colin Raffel and Noam Shazeer and Adam Roberts and Katherine Lee and Sharan Narang and Michael Matena and Yanqi Zhou and Wei Li and Peter J. Liu.
 1. **[TAPAS](model_doc/tapas)** (from Google AI) released with the paper [TAPAS: Weakly Supervised Table Parsing via Pre-training](https://arxiv.org/abs/2004.02349) by Jonathan Herzig, Paweł Krzysztof Nowak, Thomas Müller, Francesco Piccinno and Julian Martin Eisenschlos.
 1. **[TAPEX](model_doc/tapex)** (from Microsoft Research) released with the paper [TAPEX: Table Pre-training via Learning a Neural SQL Executor](https://arxiv.org/abs/2107.07653) by Qian Liu, Bei Chen, Jiaqi Guo, Morteza Ziyadi, Zeqi Lin, Weizhu Chen, Jian-Guang Lou.
+1. **[Time Series Transformer](model_doc/time_series_transformer)**  (from HuggingFace).
 1. **[Trajectory Transformer](model_doc/trajectory_transformers)** (from the University of California at Berkeley) released with the paper [Offline Reinforcement Learning as One Big Sequence Modeling Problem](https://arxiv.org/abs/2106.02039) by Michael Janner, Qiyang Li, Sergey Levine
 1. **[Transformer-XL](model_doc/transfo-xl)** (from Google/CMU) released with the paper [Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context](https://arxiv.org/abs/1901.02860) by Zihang Dai*, Zhilin Yang*, Yiming Yang, Jaime Carbonell, Quoc V. Le, Ruslan Salakhutdinov.
 1. **[TrOCR](model_doc/trocr)** (from Microsoft), released together with the paper [TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models](https://arxiv.org/abs/2109.10282) by Minghao Li, Tengchao Lv, Lei Cui, Yijuan Lu, Dinei Florencio, Cha Zhang, Zhoujun Li, Furu Wei.
@@ -166,14 +173,18 @@ The library currently contains JAX, PyTorch and TensorFlow implementations, pret
 1. **[UniSpeech](model_doc/unispeech)** (from Microsoft Research) released with the paper [UniSpeech: Unified Speech Representation Learning with Labeled and Unlabeled Data](https://arxiv.org/abs/2101.07597) by Chengyi Wang, Yu Wu, Yao Qian, Kenichi Kumatani, Shujie Liu, Furu Wei, Michael Zeng, Xuedong Huang.
 1. **[UniSpeechSat](model_doc/unispeech-sat)** (from Microsoft Research) released with the paper [UNISPEECH-SAT: UNIVERSAL SPEECH REPRESENTATION LEARNING WITH SPEAKER AWARE PRE-TRAINING](https://arxiv.org/abs/2110.05752) by Sanyuan Chen, Yu Wu, Chengyi Wang, Zhengyang Chen, Zhuo Chen, Shujie Liu, Jian Wu, Yao Qian, Furu Wei, Jinyu Li, Xiangzhan Yu.
 1. **[VAN](model_doc/van)** (from Tsinghua University and Nankai University) released with the paper [Visual Attention Network](https://arxiv.org/abs/2202.09741) by Meng-Hao Guo, Cheng-Ze Lu, Zheng-Ning Liu, Ming-Ming Cheng, Shi-Min Hu.
+1. **[VideoMAE](model_doc/videomae)** (from Multimedia Computing Group, Nanjing University) released with the paper [VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training](https://arxiv.org/abs/2203.12602) by Zhan Tong, Yibing Song, Jue Wang, Limin Wang.
 1. **[ViLT](model_doc/vilt)** (from NAVER AI Lab/Kakao Enterprise/Kakao Brain) released with the paper [ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision](https://arxiv.org/abs/2102.03334) by Wonjae Kim, Bokyung Son, Ildoo Kim.
 1. **[Vision Transformer (ViT)](model_doc/vit)** (from Google AI) released with the paper [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, Neil Houlsby.
 1. **[VisualBERT](model_doc/visual_bert)** (from UCLA NLP) released with the paper [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) by Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang.
 1. **[ViTMAE](model_doc/vit_mae)** (from Meta AI) released with the paper [Masked Autoencoders Are Scalable Vision Learners](https://arxiv.org/abs/2111.06377) by Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick.
+1. **[ViTMSN](model_doc/vit_msn)** (from Meta AI) released with the paper [Masked Siamese Networks for Label-Efficient Learning](https://arxiv.org/abs/2204.07141) by Mahmoud Assran, Mathilde Caron, Ishan Misra, Piotr Bojanowski, Florian Bordes, Pascal Vincent, Armand Joulin, Michael Rabbat, Nicolas Ballas.
 1. **[Wav2Vec2](model_doc/wav2vec2)** (from Facebook AI) released with the paper [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://arxiv.org/abs/2006.11477) by Alexei Baevski, Henry Zhou, Abdelrahman Mohamed, Michael Auli.
 1. **[Wav2Vec2-Conformer](model_doc/wav2vec2-conformer)** (from Facebook AI) released with the paper [FAIRSEQ S2T: Fast Speech-to-Text Modeling with FAIRSEQ](https://arxiv.org/abs/2010.05171) by Changhan Wang, Yun Tang, Xutai Ma, Anne Wu, Sravya Popuri, Dmytro Okhonko, Juan Pino.
 1. **[Wav2Vec2Phoneme](model_doc/wav2vec2_phoneme)** (from Facebook AI) released with the paper [Simple and Effective Zero-shot Cross-lingual Phoneme Recognition](https://arxiv.org/abs/2109.11680) by Qiantong Xu, Alexei Baevski, Michael Auli.
 1. **[WavLM](model_doc/wavlm)** (from Microsoft Research) released with the paper [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://arxiv.org/abs/2110.13900) by Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng, Furu Wei.
+1. **[Whisper](model_doc/whisper)** (from OpenAI) released with the paper [Robust Speech Recognition via Large-Scale Weak Supervision](https://cdn.openai.com/papers/whisper.pdf) by Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, Ilya Sutskever.
+1. **[X-CLIP](model_doc/xclip)** (from Microsoft Research) released with the paper [Expanding Language-Image Pretrained Models for General Video Recognition](https://arxiv.org/abs/2208.02816) by Bolin Ni, Houwen Peng, Minghao Chen, Songyang Zhang, Gaofeng Meng, Jianlong Fu, Shiming Xiang, Haibin Ling.
 1. **[XGLM](model_doc/xglm)** (From Facebook AI) released with the paper [Few-shot Learning with Multilingual Language Models](https://arxiv.org/abs/2112.10668) by Xi Victoria Lin, Todor Mihaylov, Mikel Artetxe, Tianlu Wang, Shuohui Chen, Daniel Simig, Myle Ott, Naman Goyal, Shruti Bhosale, Jingfei Du, Ramakanth Pasunuru, Sam Shleifer, Punit Singh Koura, Vishrav Chaudhary, Brian O'Horo, Jeff Wang, Luke Zettlemoyer, Zornitsa Kozareva, Mona Diab, Veselin Stoyanov, Xian Li.
 1. **[XLM](model_doc/xlm)** (from Facebook) released together with the paper [Cross-lingual Language Model Pretraining](https://arxiv.org/abs/1901.07291) by Guillaume Lample and Alexis Conneau.
 1. **[XLM-ProphetNet](model_doc/xlm-prophetnet)** (from Microsoft Research) released with the paper [ProphetNet: Predicting Future N-gram for Sequence-to-Sequence Pre-training](https://arxiv.org/abs/2001.04063) by Yu Yan, Weizhen Qi, Yeyun Gong, Dayiheng Liu, Nan Duan, Jiusheng Chen, Ruofei Zhang and Ming Zhou.
@@ -210,6 +221,7 @@ Flax), PyTorch, and/or TensorFlow.
 |           CANINE            |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
 |            CLIP             |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
 |           CodeGen           |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
+|      Conditional DETR       |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |          ConvBERT           |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
 |          ConvNeXT           |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
 |            CTRL             |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
@@ -220,13 +232,17 @@ Flax), PyTorch, and/or TensorFlow.
 |           DeBERTa           |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
 |         DeBERTa-v2          |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
 |    Decision Transformer     |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|       Deformable DETR       |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |            DeiT             |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
 |            DETR             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |         DistilBERT          |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|          DonutSwin          |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |             DPR             |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
 |             DPT             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |           ELECTRA           |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
 |       Encoder decoder       |       ❌       |       ❌       |       ✅        |         ✅         |      ✅      |
+|            ERNIE            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|             ESM             |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
 | FairSeq Machine-Translation |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
 |          FlauBERT           |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
 |            FLAVA            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
@@ -235,14 +251,15 @@ Flax), PyTorch, and/or TensorFlow.
 |            GLPN             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |           GPT Neo           |       ❌       |       ❌       |       ✅        |         ❌         |      ✅      |
 |          GPT NeoX           |       ❌       |       ✅       |       ✅        |         ❌         |      ❌      |
+|      GPT NeoX Japanese      |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
 |            GPT-J            |       ❌       |       ❌       |       ✅        |         ✅         |      ✅      |
-|          GroupViT           |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|          GroupViT           |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
 |           Hubert            |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
 |           I-BERT            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |          ImageGPT           |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |          LayoutLM           |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
 |         LayoutLMv2          |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
-|         LayoutLMv3          |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
+|         LayoutLMv3          |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
 |             LED             |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
 |            LeViT            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |         Longformer          |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
@@ -252,11 +269,12 @@ Flax), PyTorch, and/or TensorFlow.
 |           M-CTC-T           |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |           M2M100            |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
 |           Marian            |       ✅       |       ❌       |       ✅        |         ✅         |      ✅      |
+|          MarkupLM           |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
 |         MaskFormer          |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |            mBART            |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
 |        Megatron-BERT        |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |         MobileBERT          |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
-|          MobileViT          |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|          MobileViT          |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
 |            MPNet            |       ✅       |       ✅       |       ✅        |         ✅         |      ❌      |
 |             MT5             |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
 |             MVP             |       ✅       |       ✅       |       ✅        |         ❌         |      ❌      |
@@ -267,6 +285,7 @@ Flax), PyTorch, and/or TensorFlow.
 |             OPT             |       ❌       |       ❌       |       ✅        |         ✅         |      ✅      |
 |           OWL-ViT           |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |           Pegasus           |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
+|          PEGASUS-X          |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |          Perceiver          |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
 |           PLBart            |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
 |         PoolFormer          |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
@@ -293,22 +312,27 @@ Flax), PyTorch, and/or TensorFlow.
 |     Swin Transformer V2     |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |             T5              |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
 |            TAPAS            |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
+|   Time Series Transformer   |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |   Trajectory Transformer    |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |       Transformer-XL        |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
 |            TrOCR            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |          UniSpeech          |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |        UniSpeechSat         |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |             VAN             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|          VideoMAE           |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |            ViLT             |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |   Vision Encoder decoder    |       ❌       |       ❌       |       ✅        |         ✅         |      ✅      |
 |    VisionTextDualEncoder    |       ❌       |       ❌       |       ✅        |         ❌         |      ✅      |
 |         VisualBERT          |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |             ViT             |       ❌       |       ❌       |       ✅        |         ✅         |      ✅      |
 |           ViTMAE            |       ❌       |       ❌       |       ✅        |         ✅         |      ❌      |
+|           ViTMSN            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |          Wav2Vec2           |       ✅       |       ❌       |       ✅        |         ✅         |      ✅      |
 |     Wav2Vec2-Conformer      |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
 |            WavLM            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
-|            XGLM             |       ✅       |       ✅       |       ✅        |         ❌         |      ✅      |
+|           Whisper           |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
+|           X-CLIP            |       ❌       |       ❌       |       ✅        |         ❌         |      ❌      |
+|            XGLM             |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |
 |             XLM             |       ✅       |       ❌       |       ✅        |         ✅         |      ❌      |
 |       XLM-ProphetNet        |       ✅       |       ❌       |       ✅        |         ❌         |      ❌      |
 |         XLM-RoBERTa         |       ✅       |       ✅       |       ✅        |         ✅         |      ✅      |

docs/source/en/installation.mdx

@@ -139,11 +139,11 @@ conda install -c huggingface transformers
 
 ## Cache setup
 
-Pretrained models are downloaded and locally cached at: `~/.cache/huggingface/transformers/`. This is the default directory given by the shell environment variable `TRANSFORMERS_CACHE`. On Windows, the default directory is given by `C:\Users\username\.cache\huggingface\transformers`. You can change the shell environment variables shown below - in order of priority - to specify a different cache directory:
+Pretrained models are downloaded and locally cached at: `~/.cache/huggingface/hub`. This is the default directory given by the shell environment variable `TRANSFORMERS_CACHE`. On Windows, the default directory is given by `C:\Users\username\.cache\huggingface\hub`. You can change the shell environment variables shown below - in order of priority - to specify a different cache directory:
 
-1. Shell environment variable (default): `TRANSFORMERS_CACHE`.
-2. Shell environment variable: `HF_HOME` + `transformers/`.
-3. Shell environment variable: `XDG_CACHE_HOME` + `/huggingface/transformers`.
+1. Shell environment variable (default): `HUGGINGFACE_HUB_CACHE` or `TRANSFORMERS_CACHE`.
+2. Shell environment variable: `HF_HOME`.
+3. Shell environment variable: `XDG_CACHE_HOME` + `/huggingface`.
 
 <Tip>
 

docs/source/en/main_classes/callback.mdx

@@ -32,6 +32,7 @@ By default a [`Trainer`] will use the following callbacks:
 - [`~integrations.WandbCallback`] if [wandb](https://www.wandb.com/) is installed.
 - [`~integrations.CometCallback`] if [comet_ml](https://www.comet.ml/site/) is installed.
 - [`~integrations.MLflowCallback`] if [mlflow](https://www.mlflow.org/) is installed.
+- [`~integrations.NeptuneCallback`] if [neptune](https://neptune.ai/) is installed.
 - [`~integrations.AzureMLCallback`] if [azureml-sdk](https://pypi.org/project/azureml-sdk/) is
   installed.
 - [`~integrations.CodeCarbonCallback`] if [codecarbon](https://pypi.org/project/codecarbon/) is
@@ -70,6 +71,8 @@ Here is the list of the available [`TrainerCallback`] in the library:
 
 [[autodoc]] integrations.CodeCarbonCallback
 
+[[autodoc]] integrations.NeptuneCallback
+
 ## TrainerCallback
 
 [[autodoc]] TrainerCallback

docs/source/en/main_classes/deepspeed.mdx

@@ -37,7 +37,7 @@ won't be possible on a single GPU.
 2. If you don't use [`Trainer`] and want to use your own Trainer where you integrated DeepSpeed
    yourself, core functionality functions like `from_pretrained` and `from_config` include integration of essential
    parts of DeepSpeed like `zero.Init` for ZeRO stage 3 and higher. To tap into this feature read the docs on
-   [deepspeed-non-trainer-integration](#deepspeed-non-trainer-integration).
+   [non-Trainer DeepSpeed Integration](#nontrainer-deepspeed-integration).
 
 What is integrated:
 
@@ -49,7 +49,7 @@ Inference:
 
 1. DeepSpeed ZeRO Inference supports ZeRO stage 3 with ZeRO-Infinity. It uses the same ZeRO protocol as training, but
    it doesn't use an optimizer and a lr scheduler and only stage 3 is relevant. For more details see:
-   [deepspeed-zero-inference](#deepspeed-zero-inference).
+   [zero-inference](#zero-inference).
 
 There is also DeepSpeed Inference - this is a totally different technology which uses Tensor Parallelism instead of
 ZeRO (coming soon).
@@ -81,7 +81,7 @@ pip install transformers[deepspeed]
 or find more details on [the DeepSpeed's GitHub page](https://github.com/microsoft/deepspeed#installation) and
 [advanced install](https://www.deepspeed.ai/tutorials/advanced-install/).
 
-If you're still struggling with the build, first make sure to read [zero-install-notes](#zero-install-notes).
+If you're still struggling with the build, first make sure to read [CUDA Extension Installation Notes](trainer#cuda-extension-installation-notes).
 
 If you don't prebuild the extensions and rely on them to be built at run time and you tried all of the above solutions
 to no avail, the next thing to try is to pre-build the modules before installing them.
@@ -1849,7 +1849,6 @@ In this case you usually need to raise the value of `initial_scale_power`. Setti
 
 
 
-<a id='deepspeed-non-trainer-integration'></a>
 
 ## Non-Trainer Deepspeed Integration
 

docs/source/en/main_classes/model.mdx

@@ -105,7 +105,7 @@ You can also write your own device map following the same format (a dictionary l
 device_map = {"shared": 0, "encoder": 0, "decoder": 1, "lm_head": 1}
 ```
 
-Another way to minimize the memory impact of your model is to instantiate it at a lower precision dtype (like `torch.float16`).
+Another way to minimize the memory impact of your model is to instantiate it at a lower precision dtype (like `torch.float16`) or use direct quantization techniques as described below.
 
 ### Model Instantiation dtype
 
@@ -134,7 +134,6 @@ model = AutoModel.from_config(config)
 Due to Pytorch design, this functionality is only available for floating dtypes.
 
 
-
 ## ModuleUtilsMixin
 
 [[autodoc]] modeling_utils.ModuleUtilsMixin

docs/source/en/main_classes/output.mdx

@@ -16,7 +16,7 @@ All models have outputs that are instances of subclasses of [`~utils.ModelOutput
 data structures containing all the information returned by the model, but that can also be used as tuples or
 dictionaries.
 
-Let's see of this looks on an example:
+Let's see how this looks in an example:
 
 ```python
 from transformers import BertTokenizer, BertForSequenceClassification

docs/source/en/main_classes/pipelines.mdx

@@ -25,10 +25,12 @@ There are two categories of pipeline abstractions to be aware about:
   - [`AudioClassificationPipeline`]
   - [`AutomaticSpeechRecognitionPipeline`]
   - [`ConversationalPipeline`]
+  - [`DocumentQuestionAnsweringPipeline`]
   - [`FeatureExtractionPipeline`]
   - [`FillMaskPipeline`]
   - [`ImageClassificationPipeline`]
   - [`ImageSegmentationPipeline`]
+  - [`ImageToTextPipeline`]
   - [`ObjectDetectionPipeline`]
   - [`QuestionAnsweringPipeline`]
   - [`SummarizationPipeline`]
@@ -41,6 +43,7 @@ There are two categories of pipeline abstractions to be aware about:
   - [`VisualQuestionAnsweringPipeline`]
   - [`ZeroShotClassificationPipeline`]
   - [`ZeroShotImageClassificationPipeline`]
+  - [`ZeroShotObjectDetectionPipeline`]
 
 ## The pipeline abstraction
 
@@ -341,6 +344,12 @@ That should enable you to do all the custom code you want.
     - __call__
     - all
 
+### DocumentQuestionAnsweringPipeline
+
+[[autodoc]] DocumentQuestionAnsweringPipeline
+    - __call__
+    - all
+
 ### FeatureExtractionPipeline
 
 [[autodoc]] FeatureExtractionPipeline
@@ -365,6 +374,12 @@ That should enable you to do all the custom code you want.
     - __call__
     - all
 
+### ImageToTextPipeline
+
+[[autodoc]] ImageToTextPipeline
+    - __call__
+    - all
+
 ### NerPipeline
 
 [[autodoc]] NerPipeline
@@ -442,6 +457,12 @@ See [`TokenClassificationPipeline`] for all details.
     - __call__
     - all
 
+### ZeroShotObjectDetectionPipeline
+
+[[autodoc]] ZeroShotObjectDetectionPipeline
+    - __call__
+    - all
+
 ## Parent class: `Pipeline`
 
 [[autodoc]] Pipeline

docs/source/en/main_classes/trainer.mdx

@@ -567,14 +567,22 @@ as the model saving with FSDP activated is only available with recent fixes.
   For this, add `--fsdp full_shard` to the command line arguments. 
   - SHARD_GRAD_OP : Shards optimizer states + gradients across data parallel workers/GPUs.
     For this, add `--fsdp shard_grad_op` to the command line arguments.
+  - NO_SHARD : No sharding. For this, add `--fsdp no_shard` to the command line arguments.
 - To offload the parameters and gradients to the CPU, 
 add `--fsdp "full_shard offload"` or `--fsdp "shard_grad_op offload"` to the command line arguments.
 -  To automatically recursively wrap layers with FSDP using `default_auto_wrap_policy`, 
 add `--fsdp "full_shard auto_wrap"` or `--fsdp "shard_grad_op auto_wrap"` to the command line arguments.
 - To enable both CPU offloading and auto wrapping, 
 add `--fsdp "full_shard offload auto_wrap"` or `--fsdp "shard_grad_op offload auto_wrap"` to the command line arguments.
-- If auto wrapping is enabled, please add `--fsdp_min_num_params <number>` to command line arguments.
-It specifies FSDP's minimum number of parameters for Default Auto Wrapping.
+- If auto wrapping is enabled, you can either use transformer based auto wrap policy or size based auto wrap policy.
+  - For transformer based auto wrap policy, please add `--fsdp_transformer_layer_cls_to_wrap <value>` to command line arguments.
+  This specifies the transformer layer class name (case-sensitive) to wrap ,e.g, `BertLayer`, `GPTJBlock`, `T5Block` ....
+  This is important because submodules that share weights (e.g., embedding layer) should not end up in different FSDP wrapped units. 
+  Using this policy, wrapping happens for each block containing Multi-Head Attention followed by couple of MLP layers. 
+  Remaining layers including the shared embeddings are conviniently wrapped in same outermost FSDP unit.
+  Therefore, use this for transformer based models.
+  - For size based auto wrap policy, please add `--fsdp_min_num_params <number>` to command line arguments.
+  It specifies FSDP's minimum number of parameters for auto wrapping.
 
 **Few caveats to be aware of**
 - Mixed precision is currently not supported with FSDP as we wait for PyTorch to fix support for it.
@@ -583,6 +591,66 @@ More details in this [issues](https://github.com/pytorch/pytorch/issues/75676).
 More details mentioned in this [issue](https://github.com/pytorch/pytorch/issues/76501)
 (`The original model parameters' .grads are not set, meaning that they cannot be optimized separately (which is why we cannot support multiple parameter groups)`).
 
+### Using Trainer for accelerated PyTorch Training on Mac 
+
+With PyTorch v1.12 release, developers and researchers can take advantage of Apple silicon GPUs for significantly faster model training. 
+This unlocks the ability to perform machine learning workflows like prototyping and fine-tuning locally, right on Mac.
+Apple's Metal Performance Shaders (MPS) as a backend for PyTorch enables this and can be used via the new `"mps"` device. 
+This will map computational graphs and primitives on the MPS Graph framework and tuned kernels provided by MPS.
+For more information please refer official documents [Introducing Accelerated PyTorch Training on Mac](https://pytorch.org/blog/introducing-accelerated-pytorch-training-on-mac/)
+and [MPS BACKEND](https://pytorch.org/docs/stable/notes/mps.html). 
+
+<Tip warning={false}>
+
+We strongly recommend to install PyTorch >= 1.13 (nightly version at the time of writing) on your MacOS machine. 
+It has major fixes related to model correctness and performance improvements for transformer based models.
+Please refer to https://github.com/pytorch/pytorch/issues/82707 for more details.
+
+</Tip>
+
+**Benefits of Training and Inference using Apple Silicon Chips**
+
+1. Enables users to train larger networks or batch sizes locally
+2. Reduces data retrieval latency and provides the GPU with direct access to the full memory store due to unified memory architecture. 
+Therefore, improving end-to-end performance.
+3. Reduces costs associated with cloud-based development or the need for additional local GPUs.
+
+**Pre-requisites**: To install torch with mps support, 
+please follow this nice medium article [GPU-Acceleration Comes to PyTorch on M1 Macs](https://medium.com/towards-data-science/gpu-acceleration-comes-to-pytorch-on-m1-macs-195c399efcc1).
+
+**Usage**:
+User has to just pass `--use_mps_device` argument. 
+For example, you can run the offical Glue text classififcation task (from the root folder) using Apple Silicon GPU with below command:
+
+```bash
+export TASK_NAME=mrpc
+
+python examples/pytorch/text-classification/run_glue.py \
+  --model_name_or_path bert-base-cased \
+  --task_name $TASK_NAME \
+  --do_train \
+  --do_eval \
+  --max_seq_length 128 \
+  --per_device_train_batch_size 32 \
+  --learning_rate 2e-5 \
+  --num_train_epochs 3 \
+  --output_dir /tmp/$TASK_NAME/ \
+  --use_mps_device \
+  --overwrite_output_dir
+```
+
+**A few caveats to be aware of**
+
+1. Some PyTorch operations have not been implemented in mps and will throw an error. 
+One way to get around that is to set the environment variable `PYTORCH_ENABLE_MPS_FALLBACK=1`, 
+which will fallback to CPU for these operations. It still throws a UserWarning however.
+2. Distributed setups `gloo` and `nccl` are not working with `mps` device. 
+This means that currently only single GPU of `mps` device type can be used.
+
+Finally, please, remember that, 🤗 `Trainer` only integrates MPS backend, therefore if you
+have any problems or questions with regards to MPS backend usage, please, 
+file an issue with [PyTorch GitHub](https://github.com/pytorch/pytorch/issues).
+
 Sections that were moved:
 
 [ <a href="./deepspeed#deepspeed-trainer-integration">DeepSpeed</a><a id="deepspeed"></a>

docs/source/en/model_doc/auto.mdx

@@ -114,10 +114,18 @@ Likewise, if your `NewModel` is a subclass of [`PreTrainedModel`], make sure its
 
 [[autodoc]] AutoModelForTableQuestionAnswering
 
+## AutoModelForDocumentQuestionAnswering
+
+[[autodoc]] AutoModelForDocumentQuestionAnswering
+
 ## AutoModelForImageClassification
 
 [[autodoc]] AutoModelForImageClassification
 
+## AutoModelForVideoClassification
+
+[[autodoc]] AutoModelForVideoClassification
+
 ## AutoModelForVision2Seq
 
 [[autodoc]] AutoModelForVision2Seq
@@ -166,6 +174,10 @@ Likewise, if your `NewModel` is a subclass of [`PreTrainedModel`], make sure its
 
 [[autodoc]] AutoModelForInstanceSegmentation
 
+## AutoModelForZeroShotObjectDetection
+
+[[autodoc]] AutoModelForZeroShotObjectDetection
+
 ## TFAutoModel
 
 [[autodoc]] TFAutoModel
@@ -182,6 +194,10 @@ Likewise, if your `NewModel` is a subclass of [`PreTrainedModel`], make sure its
 
 [[autodoc]] TFAutoModelForImageClassification
 
+## TFAutoModelForSemanticSegmentation
+
+[[autodoc]] TFAutoModelForSemanticSegmentation
+
 ## TFAutoModelForMaskedLM
 
 [[autodoc]] TFAutoModelForMaskedLM
@@ -206,6 +222,10 @@ Likewise, if your `NewModel` is a subclass of [`PreTrainedModel`], make sure its
 
 [[autodoc]] TFAutoModelForTableQuestionAnswering
 
+## TFAutoModelForDocumentQuestionAnswering
+
+[[autodoc]] TFAutoModelForDocumentQuestionAnswering
+
 ## TFAutoModelForTokenClassification
 
 [[autodoc]] TFAutoModelForTokenClassification

docs/source/en/model_doc/beit.mdx

@@ -59,6 +59,11 @@ Tips:
   `use_relative_position_bias` attribute of [`BeitConfig`] to `True` in order to add
   position embeddings.
 
+<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/beit_architecture.jpg"
+alt="drawing" width="600"/> 
+
+<small> BEiT pre-training. Taken from the <a href="https://arxiv.org/abs/2106.08254">original paper.</a> </small>
+
 This model was contributed by [nielsr](https://huggingface.co/nielsr). The JAX/FLAX version of this model was
 contributed by [kamalkraj](https://huggingface.co/kamalkraj). The original code can be found [here](https://github.com/microsoft/unilm/tree/master/beit).
 
@@ -77,6 +82,7 @@ contributed by [kamalkraj](https://huggingface.co/kamalkraj). The original code
 
 [[autodoc]] BeitFeatureExtractor
     - __call__
+    - post_process_semantic_segmentation
 
 ## BeitModel
 

docs/source/en/model_doc/bloom.mdx

@@ -18,11 +18,11 @@ The BLOOM model has been proposed with its various versions through the [BigScie
 The architecture of BLOOM is essentially similar to GPT3 (auto-regressive model for next token prediction), but has been trained on 46 different languages and 13 programming languages.
 Several smaller versions of the models have been trained on the same dataset. BLOOM is available in the following versions:
 
-- [bloom-350m](https://huggingface.co/bigscience/bloom-350m)
-- [bloom-760m](https://huggingface.co/bigscience/bloom-760m)
-- [bloom-1b3](https://huggingface.co/bigscience/bloom-1b3)
-- [bloom-2b5](https://huggingface.co/bigscience/bloom-2b5)
-- [bloom-6b3](https://huggingface.co/bigscience/bloom-6b3)
+- [bloom-560m](https://huggingface.co/bigscience/bloom-560m)
+- [bloom-1b1](https://huggingface.co/bigscience/bloom-1b1)
+- [bloom-1b7](https://huggingface.co/bigscience/bloom-1b7)
+- [bloom-3b](https://huggingface.co/bigscience/bloom-3b)
+- [bloom-7b1](https://huggingface.co/bigscience/bloom-7b1)
 - [bloom](https://huggingface.co/bigscience/bloom) (176B parameters)
 
 
@@ -55,3 +55,8 @@ Several smaller versions of the models have been trained on the same dataset. BL
 
 [[autodoc]] BloomForTokenClassification
     - forward
+
+## BloomForQuestionAnswering
+
+[[autodoc]] BloomForQuestionAnswering
+    - forward

docs/source/en/model_doc/conditional_detr.mdx

@@ -0,0 +1,57 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# Conditional DETR
+
+## Overview
+
+The Conditional DETR model was proposed in [Conditional DETR for Fast Training Convergence](https://arxiv.org/abs/2108.06152) by Depu Meng, Xiaokang Chen, Zejia Fan, Gang Zeng, Houqiang Li, Yuhui Yuan, Lei Sun, Jingdong Wang. Conditional DETR presents a conditional cross-attention mechanism for fast DETR training. Conditional DETR converges 6.7× to 10× faster than DETR.
+
+The abstract from the paper is the following:
+
+*The recently-developed DETR approach applies the transformer encoder and decoder architecture to object detection and achieves promising performance. In this paper, we handle the critical issue, slow training convergence, and present a conditional cross-attention mechanism for fast DETR training. Our approach is motivated by that the cross-attention in DETR relies highly on the content embeddings for localizing the four extremities and predicting the box, which increases the need for high-quality content embeddings and thus the training difficulty. Our approach, named conditional DETR, learns a conditional spatial query from the decoder embedding for decoder multi-head cross-attention. The benefit is that through the conditional spatial query, each cross-attention head is able to attend to a band containing a distinct region, e.g., one object extremity or a region inside the object box. This narrows down the spatial range for localizing the distinct regions for object classification and box regression, thus relaxing the dependence on the content embeddings and easing the training. Empirical results show that conditional DETR converges 6.7× faster for the backbones R50 and R101 and 10× faster for stronger backbones DC5-R50 and DC5-R101. Code is available at https://github.com/Atten4Vis/ConditionalDETR.*
+
+<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/conditional_detr_curve.jpg"
+alt="drawing" width="600"/> 
+
+<small> Conditional DETR shows much faster convergence compared to the original DETR. Taken from the <a href="https://arxiv.org/abs/2108.06152">original paper</a>.</small>
+
+This model was contributed by [DepuMeng](https://huggingface.co/DepuMeng). The original code can be found [here](https://github.com/Atten4Vis/ConditionalDETR).
+
+
+## ConditionalDetrConfig
+
+[[autodoc]] ConditionalDetrConfig
+
+## ConditionalDetrFeatureExtractor
+
+[[autodoc]] ConditionalDetrFeatureExtractor
+    - __call__
+    - pad_and_create_pixel_mask
+    - post_process
+    - post_process_segmentation
+    - post_process_panoptic
+
+## ConditionalDetrModel
+
+[[autodoc]] ConditionalDetrModel
+    - forward
+
+## ConditionalDetrForObjectDetection
+
+[[autodoc]] ConditionalDetrForObjectDetection
+    - forward
+
+## ConditionalDetrForSegmentation
+
+[[autodoc]] ConditionalDetrForSegmentation
+    - forward

docs/source/en/model_doc/deformable_detr.mdx

@@ -0,0 +1,60 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# Deformable DETR
+
+## Overview
+
+The Deformable DETR model was proposed in [Deformable DETR: Deformable Transformers for End-to-End Object Detection](https://arxiv.org/abs/2010.04159) by Xizhou Zhu, Weijie Su, Lewei Lu, Bin Li, Xiaogang Wang, Jifeng Dai.
+Deformable DETR mitigates the slow convergence issues and limited feature spatial resolution of the original [DETR](detr) by leveraging a new deformable attention module which only attends to a small set of key sampling points around a reference.
+
+The abstract from the paper is the following:
+
+*DETR has been recently proposed to eliminate the need for many hand-designed components in object detection while demonstrating good performance. However, it suffers from slow convergence and limited feature spatial resolution, due to the limitation of Transformer attention modules in processing image feature maps. To mitigate these issues, we proposed Deformable DETR, whose attention modules only attend to a small set of key sampling points around a reference. Deformable DETR can achieve better performance than DETR (especially on small objects) with 10 times less training epochs. Extensive experiments on the COCO benchmark demonstrate the effectiveness of our approach.*
+
+Tips:
+
+- One can use the [`AutoFeatureExtractor`] API to prepare images (and optional targets) for the model. This will instantiate a [`DetrFeatureExtractor`] behind the scenes.
+- Training Deformable DETR is equivalent to training the original [DETR](detr) model. Demo notebooks can be found [here](https://github.com/NielsRogge/Transformers-Tutorials/tree/master/DETR).
+
+<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/deformable_detr_architecture.png"
+alt="drawing" width="600"/> 
+
+<small> Deformable DETR architecture. Taken from the <a href="https://arxiv.org/abs/2010.04159">original paper</a>.</small>
+
+This model was contributed by [nielsr](https://huggingface.co/nielsr). The original code can be found [here](https://github.com/fundamentalvision/Deformable-DETR).
+
+## DeformableDetrFeatureExtractor
+
+[[autodoc]] DeformableDetrFeatureExtractor
+    - __call__
+    - pad_and_create_pixel_mask
+    - post_process
+    - post_process_segmentation
+    - post_process_panoptic
+
+
+## DeformableDetrConfig
+
+[[autodoc]] DeformableDetrConfig
+
+
+## DeformableDetrModel
+
+[[autodoc]] DeformableDetrModel
+    - forward
+
+
+## DeformableDetrForObjectDetection
+
+[[autodoc]] DeformableDetrForObjectDetection
+    - forward

docs/source/en/model_doc/detr.mdx

@@ -171,9 +171,10 @@ mean Average Precision (mAP) and Panoptic Quality (PQ). The latter objects are i
 [[autodoc]] DetrFeatureExtractor
     - __call__
     - pad_and_create_pixel_mask
-    - post_process
-    - post_process_segmentation
-    - post_process_panoptic
+    - post_process_object_detection
+    - post_process_semantic_segmentation
+    - post_process_instance_segmentation
+    - post_process_panoptic_segmentation
 
 ## DetrModel
 

docs/source/en/model_doc/donut.mdx

@@ -0,0 +1,214 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the
+License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an
+"AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License. -->
+
+# Donut
+
+## Overview
+
+The Donut model was proposed in [OCR-free Document Understanding Transformer](https://arxiv.org/abs/2111.15664) by
+Geewook Kim, Teakgyu Hong, Moonbin Yim, Jeongyeon Nam, Jinyoung Park, Jinyeong Yim, Wonseok Hwang, Sangdoo Yun, Dongyoon Han, Seunghyun Park.
+Donut consists of an image Transformer encoder and an autoregressive text Transformer decoder to perform document understanding
+tasks such as document image classification, form understanding and visual question answering.
+
+The abstract from the paper is the following:
+
+*Understanding document images (e.g., invoices) is a core but challenging task since it requires complex functions such as reading text and a holistic understanding of the document. Current Visual Document Understanding (VDU) methods outsource the task of reading text to off-the-shelf Optical Character Recognition (OCR) engines and focus on the understanding task with the OCR outputs. Although such OCR-based approaches have shown promising performance, they suffer from 1) high computational costs for using OCR; 2) inflexibility of OCR models on languages or types of document; 3) OCR error propagation to the subsequent process. To address these issues, in this paper, we introduce a novel OCR-free VDU model named Donut, which stands for Document understanding transformer. As the first step in OCR-free VDU research, we propose a simple architecture (i.e., Transformer) with a pre-training objective (i.e., cross-entropy loss). Donut is conceptually simple yet effective. Through extensive experiments and analyses, we show a simple OCR-free VDU model, Donut, achieves state-of-the-art performances on various VDU tasks in terms of both speed and accuracy. In addition, we offer a synthetic data generator that helps the model pre-training to be flexible in various languages and domains.*
+
+<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/donut_architecture.jpg"
+alt="drawing" width="600"/> 
+
+<small> Donut high-level overview. Taken from the <a href="https://arxiv.org/abs/2111.15664">original paper</a>. </small>
+
+This model was contributed by [nielsr](https://huggingface.co/nielsr). The original code can be found
+[here](https://github.com/clovaai/donut).
+
+Tips:
+
+- The quickest way to get started with Donut is by checking the [tutorial
+  notebooks](https://github.com/NielsRogge/Transformers-Tutorials/tree/master/Donut), which show how to use the model
+  at inference time as well as fine-tuning on custom data.
+- Donut is always used within the [VisionEncoderDecoder](vision-encoder-decoder) framework.
+
+## Inference
+
+Donut's [`VisionEncoderDecoder`] model accepts images as input and makes use of
+[`~generation_utils.GenerationMixin.generate`] to autoregressively generate text given the input image.
+
+The [`DonutFeatureExtractor`] class is responsible for preprocessing the input image and
+[`XLMRobertaTokenizer`/`XLMRobertaTokenizerFast`] decodes the generated target tokens to the target string. The
+[`DonutProcessor`] wraps [`DonutFeatureExtractor`] and [`XLMRobertaTokenizer`/`XLMRobertaTokenizerFast`]
+into a single instance to both extract the input features and decode the predicted token ids.
+
+- Step-by-step Document Image Classification
+
+```py
+>>> import re
+
+>>> from transformers import DonutProcessor, VisionEncoderDecoderModel
+>>> from datasets import load_dataset
+>>> import torch
+
+>>> processor = DonutProcessor.from_pretrained("naver-clova-ix/donut-base-finetuned-rvlcdip")
+>>> model = VisionEncoderDecoderModel.from_pretrained("naver-clova-ix/donut-base-finetuned-rvlcdip")
+
+>>> device = "cuda" if torch.cuda.is_available() else "cpu"
+>>> model.to(device)  # doctest: +IGNORE_RESULT
+
+>>> # load document image
+>>> dataset = load_dataset("hf-internal-testing/example-documents", split="test")
+>>> image = dataset[1]["image"]
+
+>>> # prepare decoder inputs
+>>> task_prompt = "<s_rvlcdip>"
+>>> decoder_input_ids = processor.tokenizer(task_prompt, add_special_tokens=False, return_tensors="pt").input_ids
+
+>>> pixel_values = processor(image, return_tensors="pt").pixel_values
+
+>>> outputs = model.generate(
+...     pixel_values.to(device),
+...     decoder_input_ids=decoder_input_ids.to(device),
+...     max_length=model.decoder.config.max_position_embeddings,
+...     early_stopping=True,
+...     pad_token_id=processor.tokenizer.pad_token_id,
+...     eos_token_id=processor.tokenizer.eos_token_id,
+...     use_cache=True,
+...     num_beams=1,
+...     bad_words_ids=[[processor.tokenizer.unk_token_id]],
+...     return_dict_in_generate=True,
+... )
+
+>>> sequence = processor.batch_decode(outputs.sequences)[0]
+>>> sequence = sequence.replace(processor.tokenizer.eos_token, "").replace(processor.tokenizer.pad_token, "")
+>>> sequence = re.sub(r"<.*?>", "", sequence, count=1).strip()  # remove first task start token
+>>> print(processor.token2json(sequence))
+{'class': 'advertisement'}
+```
+
+- Step-by-step Document Parsing
+
+```py
+>>> import re
+
+>>> from transformers import DonutProcessor, VisionEncoderDecoderModel
+>>> from datasets import load_dataset
+>>> import torch
+
+>>> processor = DonutProcessor.from_pretrained("naver-clova-ix/donut-base-finetuned-cord-v2")
+>>> model = VisionEncoderDecoderModel.from_pretrained("naver-clova-ix/donut-base-finetuned-cord-v2")
+
+>>> device = "cuda" if torch.cuda.is_available() else "cpu"
+>>> model.to(device)  # doctest: +IGNORE_RESULT
+
+>>> # load document image
+>>> dataset = load_dataset("hf-internal-testing/example-documents", split="test")
+>>> image = dataset[2]["image"]
+
+>>> # prepare decoder inputs
+>>> task_prompt = "<s_cord-v2>"
+>>> decoder_input_ids = processor.tokenizer(task_prompt, add_special_tokens=False, return_tensors="pt").input_ids
+
+>>> pixel_values = processor(image, return_tensors="pt").pixel_values
+
+>>> outputs = model.generate(
+...     pixel_values.to(device),
+...     decoder_input_ids=decoder_input_ids.to(device),
+...     max_length=model.decoder.config.max_position_embeddings,
+...     early_stopping=True,
+...     pad_token_id=processor.tokenizer.pad_token_id,
+...     eos_token_id=processor.tokenizer.eos_token_id,
+...     use_cache=True,
+...     num_beams=1,
+...     bad_words_ids=[[processor.tokenizer.unk_token_id]],
+...     return_dict_in_generate=True,
+... )
+
+>>> sequence = processor.batch_decode(outputs.sequences)[0]
+>>> sequence = sequence.replace(processor.tokenizer.eos_token, "").replace(processor.tokenizer.pad_token, "")
+>>> sequence = re.sub(r"<.*?>", "", sequence, count=1).strip()  # remove first task start token
+>>> print(processor.token2json(sequence))
+{'menu': {'nm': 'CINNAMON SUGAR', 'unitprice': '17,000', 'cnt': '1 x', 'price': '17,000'}, 'sub_total': {'subtotal_price': '17,000'}, 'total': {'total_price': '17,000', 'cashprice': '20,000', 'changeprice': '3,000'}}
+```
+
+- Step-by-step Document Visual Question Answering (DocVQA)
+
+```py
+>>> import re
+
+>>> from transformers import DonutProcessor, VisionEncoderDecoderModel
+>>> from datasets import load_dataset
+>>> import torch
+
+>>> processor = DonutProcessor.from_pretrained("naver-clova-ix/donut-base-finetuned-docvqa")
+>>> model = VisionEncoderDecoderModel.from_pretrained("naver-clova-ix/donut-base-finetuned-docvqa")
+
+>>> device = "cuda" if torch.cuda.is_available() else "cpu"
+>>> model.to(device)  # doctest: +IGNORE_RESULT
+
+>>> # load document image from the DocVQA dataset
+>>> dataset = load_dataset("hf-internal-testing/example-documents", split="test")
+>>> image = dataset[0]["image"]
+
+>>> # prepare decoder inputs
+>>> task_prompt = "<s_docvqa><s_question>{user_input}</s_question><s_answer>"
+>>> question = "When is the coffee break?"
+>>> prompt = task_prompt.replace("{user_input}", question)
+>>> decoder_input_ids = processor.tokenizer(prompt, add_special_tokens=False, return_tensors="pt").input_ids
+
+>>> pixel_values = processor(image, return_tensors="pt").pixel_values
+
+>>> outputs = model.generate(
+...     pixel_values.to(device),
+...     decoder_input_ids=decoder_input_ids.to(device),
+...     max_length=model.decoder.config.max_position_embeddings,
+...     early_stopping=True,
+...     pad_token_id=processor.tokenizer.pad_token_id,
+...     eos_token_id=processor.tokenizer.eos_token_id,
+...     use_cache=True,
+...     num_beams=1,
+...     bad_words_ids=[[processor.tokenizer.unk_token_id]],
+...     return_dict_in_generate=True,
+... )
+
+>>> sequence = processor.batch_decode(outputs.sequences)[0]
+>>> sequence = sequence.replace(processor.tokenizer.eos_token, "").replace(processor.tokenizer.pad_token, "")
+>>> sequence = re.sub(r"<.*?>", "", sequence, count=1).strip()  # remove first task start token
+>>> print(processor.token2json(sequence))
+{'question': 'When is the coffee break?', 'answer': '11-14 to 11:39 a.m.'}
+```
+
+See the [model hub](https://huggingface.co/models?filter=donut) to look for Donut checkpoints.
+
+## Training
+
+We refer to the [tutorial notebooks](https://github.com/NielsRogge/Transformers-Tutorials/tree/master/Donut).
+
+## DonutSwinConfig
+
+[[autodoc]] DonutSwinConfig
+
+## DonutFeatureExtractor
+
+[[autodoc]] DonutFeatureExtractor
+    - __call__
+
+## DonutProcessor
+
+[[autodoc]] DonutProcessor
+    - __call__
+    - from_pretrained
+    - save_pretrained
+    - batch_decode
+    - decode
+
+## DonutSwinModel
+
+[[autodoc]] DonutSwinModel
+    - forward

docs/source/en/model_doc/dpt.mdx

@@ -37,6 +37,7 @@ This model was contributed by [nielsr](https://huggingface.co/nielsr). The origi
 
 [[autodoc]] DPTFeatureExtractor
     - __call__
+    - post_process_semantic_segmentation
 
 
 ## DPTModel

docs/source/en/model_doc/ernie.mdx

@@ -0,0 +1,102 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# ERNIE
+
+## Overview
+ERNIE is a series of powerful models proposed by baidu, especially in Chinese tasks,
+including [ERNIE1.0](https://arxiv.org/abs/1904.09223), [ERNIE2.0](https://ojs.aaai.org/index.php/AAAI/article/view/6428),
+[ERNIE3.0](https://arxiv.org/abs/2107.02137), [ERNIE-Gram](https://arxiv.org/abs/2010.12148), [ERNIE-health](https://arxiv.org/abs/2110.07244), etc.
+
+These models are contributed by [nghuyong](https://huggingface.co/nghuyong) and the official code can be found in [PaddleNLP](https://github.com/PaddlePaddle/PaddleNLP) (in PaddlePaddle).
+
+### How to use
+Take `ernie-1.0-base-zh` as an example:
+
+```Python
+from transformers import AutoTokenizer, AutoModel
+tokenizer = AutoTokenizer.from_pretrained("nghuyong/ernie-1.0-base-zh")
+model = AutoModel.from_pretrained("nghuyong/ernie-1.0-base-zh")
+```
+
+### Supported Models
+
+|     Model Name      | Language |           Description           |
+|:-------------------:|:--------:|:-------------------------------:|
+|  ernie-1.0-base-zh  | Chinese  | Layer:12, Heads:12, Hidden:768  |
+|  ernie-2.0-base-en  | English  | Layer:12, Heads:12, Hidden:768  |
+| ernie-2.0-large-en  | English  | Layer:24, Heads:16, Hidden:1024 |
+|  ernie-3.0-base-zh  | Chinese  | Layer:12, Heads:12, Hidden:768  |
+| ernie-3.0-medium-zh | Chinese  |  Layer:6, Heads:12, Hidden:768  |
+|  ernie-3.0-mini-zh  | Chinese  |  Layer:6, Heads:12, Hidden:384  |
+| ernie-3.0-micro-zh  | Chinese  |  Layer:4, Heads:12, Hidden:384  |
+|  ernie-3.0-nano-zh  | Chinese  |  Layer:4, Heads:12, Hidden:312  |
+|   ernie-health-zh   | Chinese  | Layer:12, Heads:12, Hidden:768  |
+|    ernie-gram-zh    | Chinese  | Layer:12, Heads:12, Hidden:768  |
+
+You can find all the supported models from huggingface's model hub: [huggingface.co/nghuyong](https://huggingface.co/nghuyong), and model details from paddle's official
+repo: [PaddleNLP](https://paddlenlp.readthedocs.io/zh/latest/model_zoo/transformers/ERNIE/contents.html)
+and [ERNIE](https://github.com/PaddlePaddle/ERNIE/blob/repro).
+
+## ErnieConfig
+
+[[autodoc]] ErnieConfig
+    - all
+
+## Ernie specific outputs
+
+[[autodoc]] models.ernie.modeling_ernie.ErnieForPreTrainingOutput
+
+## ErnieModel
+
+[[autodoc]] ErnieModel
+    - forward
+
+## ErnieForPreTraining
+
+[[autodoc]] ErnieForPreTraining
+    - forward
+
+## ErnieForCausalLM
+
+[[autodoc]] ErnieForCausalLM
+    - forward
+
+## ErnieForMaskedLM
+
+[[autodoc]] ErnieForMaskedLM
+    - forward
+
+## ErnieForNextSentencePrediction
+
+[[autodoc]] ErnieForNextSentencePrediction
+    - forward
+
+## ErnieForSequenceClassification
+
+[[autodoc]] ErnieForSequenceClassification
+    - forward
+
+## ErnieForMultipleChoice
+
+[[autodoc]] ErnieForMultipleChoice
+    - forward
+
+## ErnieForTokenClassification
+
+[[autodoc]] ErnieForTokenClassification
+    - forward
+
+## ErnieForQuestionAnswering
+
+[[autodoc]] ErnieForQuestionAnswering
+    - forward

docs/source/en/model_doc/esm.mdx

@@ -0,0 +1,109 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# ESM
+
+## Overview
+This page provides code and pre-trained weights for Transformer protein language models from Meta AI's Fundamental 
+AI Research Team, providing the state-of-the-art ESM-2, and the previously released ESM-1b and ESM-1v. Transformer 
+protein language models were introduced in the paper [Biological structure and function emerge from scaling 
+unsupervised learning to 250 million protein sequences](https://www.pnas.org/content/118/15/e2016239118) by 
+Alexander Rives, Joshua Meier, Tom Sercu, Siddharth Goyal, Zeming Lin, Jason Liu, Demi Guo, Myle Ott, 
+C. Lawrence Zitnick, Jerry Ma, and Rob Fergus.
+The first version of this paper was [preprinted in 2019](https://www.biorxiv.org/content/10.1101/622803v1?versioned=true).
+
+ESM-2 outperforms all tested single-sequence protein language models across a range of structure prediction tasks,
+and enables atomic resolution structure prediction.
+It was released with the paper [Language models of protein sequences at the scale of evolution enable accurate
+structure prediction](https://doi.org/10.1101/2022.07.20.500902) by Zeming Lin, Halil Akin, Roshan Rao, Brian Hie,
+Zhongkai Zhu, Wenting Lu, Allan dos Santos Costa, Maryam Fazel-Zarandi, Tom Sercu, Sal Candido and Alexander Rives.
+
+
+The abstract from 
+"Biological structure and function emerge from scaling unsupervised learning to 250 
+million protein sequences" is
+
+
+*In the field of artificial intelligence, a combination of scale in data and model capacity enabled by unsupervised
+learning has led to major advances in representation learning and statistical generation. In the life sciences, the
+anticipated growth of sequencing promises unprecedented data on natural sequence diversity. Protein language modeling
+at the scale of evolution is a logical step toward predictive and generative artificial intelligence for biology. To
+this end, we use unsupervised learning to train a deep contextual language model on 86 billion amino acids across 250
+million protein sequences spanning evolutionary diversity. The resulting model contains information about biological
+properties in its representations. The representations are learned from sequence data alone. The learned representation
+space has a multiscale organization reflecting structure from the level of biochemical properties of amino acids to
+remote homology of proteins. Information about secondary and tertiary structure is encoded in the representations and
+can be identified by linear projections. Representation learning produces features that generalize across a range of
+applications, enabling state-of-the-art supervised prediction of mutational effect and secondary structure and
+improving state-of-the-art features for long-range contact prediction.*
+
+
+The abstract from
+"Language models of protein sequences at the scale of evolution enable accurate structure prediction" is
+
+*Large language models have recently been shown to develop emergent capabilities with scale, going beyond
+simple pattern matching to perform higher level reasoning and generate lifelike images and text. While
+language models trained on protein sequences have been studied at a smaller scale, little is known about
+what they learn about biology as they are scaled up. In this work we train models up to 15 billion parameters,
+the largest language models of proteins to be evaluated to date. We find that as models are scaled they learn
+information enabling the prediction of the three-dimensional structure of a protein at the resolution of
+individual atoms. We present ESMFold for high accuracy end-to-end atomic level structure prediction directly
+from the individual sequence of a protein. ESMFold has similar accuracy to AlphaFold2 and RoseTTAFold for
+sequences with low perplexity that are well understood by the language model. ESMFold inference is an
+order of magnitude faster than AlphaFold2, enabling exploration of the structural space of metagenomic
+proteins in practical timescales.*
+
+
+
+
+Tips:
+
+- ESM models are trained with a masked language modeling (MLM) objective.
+
+The original code can be found [here](https://github.com/facebookresearch/esm) and was
+was developed by the Fundamental AI Research team at Meta AI.
+This model was contributed to huggingface by [jasonliu](https://huggingface.co/jasonliu) 
+and [Matt](https://huggingface.co/Rocketknight1).
+
+## EsmConfig
+
+[[autodoc]] EsmConfig
+    - all
+
+## EsmTokenizer
+
+[[autodoc]] EsmTokenizer
+    - build_inputs_with_special_tokens
+    - get_special_tokens_mask
+    - create_token_type_ids_from_sequences
+    - save_vocabulary
+
+
+## EsmModel
+
+[[autodoc]] EsmModel
+    - forward
+
+## EsmForMaskedLM
+
+[[autodoc]] EsmForMaskedLM
+    - forward
+
+## EsmForSequenceClassification
+
+[[autodoc]] EsmForSequenceClassification
+    - forward
+
+## EsmForTokenClassification
+
+[[autodoc]] EsmForTokenClassification
+    - forward

docs/source/en/model_doc/gpt_neox_japanese.mdx

@@ -0,0 +1,66 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# GPT-NeoX-Japanese
+
+## Overview
+
+We introduce GPT-NeoX-Japanese, which is an autoregressive language model for Japanese, trained on top of [https://github.com/EleutherAI/gpt-neox](https://github.com/EleutherAI/gpt-neox).
+Japanese is a unique language with its large vocabulary and a combination of hiragana, katakana, and kanji writing scripts.
+To address this distinct structure of the Japanese language, we use a [special sub-word tokenizer](https://github.com/tanreinama/Japanese-BPEEncoder_V2). We are very grateful to *tanreinama* for open-sourcing this incredibly helpful tokenizer.
+Following the recommendations from Google's research on [PaLM](https://ai.googleblog.com/2022/04/pathways-language-model-palm-scaling-to.html), we have removed bias parameters from transformer blocks, achieving better model performance. Please refer [this article](https://medium.com/ml-abeja/training-a-better-gpt-2-93b157662ae4) in detail.
+
+Development of the model was led by [Shinya Otani](https://github.com/SO0529), [Takayoshi Makabe](https://github.com/spider-man-tm), [Anuj Arora](https://github.com/Anuj040), and [Kyo Hattori](https://github.com/go5paopao) from [ABEJA, Inc.](https://www.abejainc.com/). For more information on this model-building activity, please refer [here (ja)](https://tech-blog.abeja.asia/entry/abeja-gpt-project-202207).
+
+### Generation
+
+The `generate()` method can be used to generate text using GPT NeoX Japanese model.
+
+```python
+>>> from transformers import GPTNeoXJapaneseForCausalLM, GPTNeoXJapaneseTokenizer
+
+>>> model = GPTNeoXJapaneseForCausalLM.from_pretrained("abeja/gpt-neox-japanese-2.7b")
+>>> tokenizer = GPTNeoXJapaneseTokenizer.from_pretrained("abeja/gpt-neox-japanese-2.7b")
+
+>>> prompt = "人とAIが協調するためには、"
+
+>>> input_ids = tokenizer(prompt, return_tensors="pt").input_ids
+
+>>> gen_tokens = model.generate(
+...     input_ids,
+...     do_sample=True,
+...     temperature=0.9,
+...     max_length=100,
+... )
+>>> gen_text = tokenizer.batch_decode(gen_tokens, skip_special_tokens=True)[0]
+
+>>> print(gen_text)
+人とAIが協調するためには、AIと人が共存し、AIを正しく理解する必要があります。
+```
+
+## GPTNeoXJapaneseConfig
+
+[[autodoc]] GPTNeoXJapaneseConfig
+
+## GPTNeoXJapaneseTokenizer
+
+[[autodoc]] GPTNeoXJapaneseTokenizer
+
+## GPTNeoXJapaneseModel
+
+[[autodoc]] GPTNeoXJapaneseModel
+    - forward
+
+## GPTNeoXJapaneseForCausalLM
+
+[[autodoc]] GPTNeoXJapaneseForCausalLM
+    - forward

docs/source/en/model_doc/groupvit.mdx

@@ -26,7 +26,7 @@ Tips:
 - You may specify `output_segmentation=True` in the forward of `GroupViTModel` to get the segmentation logits of input texts. 
 - The quickest way to get started with GroupViT is by checking the [example notebooks](https://github.com/xvjiarui/GroupViT/blob/main/demo/GroupViT_hf_inference_notebook.ipynb) (which showcase zero-shot segmentation inference). One can also check out the [HuggingFace Spaces demo](https://huggingface.co/spaces/xvjiarui/GroupViT) to play with GroupViT. 
 
-This model was contributed by [xvjiarui](https://huggingface.co/xvjiarui).
+This model was contributed by [xvjiarui](https://huggingface.co/xvjiarui). The TensorFlow version was contributed by [ariG23498](https://huggingface.co/ariG23498) with the help of [Yih-Dar SHIEH](https://huggingface.co/ydshieh), [Amy Roberts](https://huggingface.co/amyeroberts), and [Joao Gante](https://huggingface.co/joaogante).
 The original code can be found [here](https://github.com/NVlabs/GroupViT).
 
 
@@ -59,3 +59,20 @@ The original code can be found [here](https://github.com/NVlabs/GroupViT).
 
 [[autodoc]] GroupViTVisionModel
     - forward
+
+## TFGroupViTModel
+
+[[autodoc]] TFGroupViTModel
+    - call
+    - get_text_features
+    - get_image_features
+
+## TFGroupViTTextModel
+
+[[autodoc]] TFGroupViTTextModel
+    - call
+
+## TFGroupViTVisionModel
+
+[[autodoc]] TFGroupViTVisionModel
+    - call

docs/source/en/model_doc/layoutlm.mdx

@@ -67,7 +67,8 @@ occurs. Those can be obtained using the Python Image Library (PIL) library for e
 ```python
 from PIL import Image
 
-image = Image.open("name_of_your_document - can be a png file, pdf, etc.")
+# Document can be a png, jpg, etc. PDFs must be converted to images.
+image = Image.open(name_of_your_document).convert("RGB")
 
 width, height = image.size
 ```
@@ -107,6 +108,10 @@ This model was contributed by [liminghao1630](https://huggingface.co/liminghao16
 
 [[autodoc]] LayoutLMForTokenClassification
 
+## LayoutLMForQuestionAnswering
+
+[[autodoc]] LayoutLMForQuestionAnswering
+
 ## TFLayoutLMModel
 
 [[autodoc]] TFLayoutLMModel
@@ -122,3 +127,7 @@ This model was contributed by [liminghao1630](https://huggingface.co/liminghao16
 ## TFLayoutLMForTokenClassification
 
 [[autodoc]] TFLayoutLMForTokenClassification
+
+## TFLayoutLMForQuestionAnswering
+
+[[autodoc]] TFLayoutLMForQuestionAnswering

docs/source/en/model_doc/layoutlmv3.mdx

@@ -26,18 +26,18 @@ Tips:
 
 - In terms of data processing, LayoutLMv3 is identical to its predecessor [LayoutLMv2](layoutlmv2), except that:
     - images need to be resized and normalized with channels in regular RGB format. LayoutLMv2 on the other hand normalizes the images internally and expects the channels in BGR format.
-    - text is tokenized using byte-pair encoding (BPE), as opposed to WordPiece. 
+    - text is tokenized using byte-pair encoding (BPE), as opposed to WordPiece.
   Due to these differences in data preprocessing, one can use [`LayoutLMv3Processor`] which internally combines a [`LayoutLMv3FeatureExtractor`] (for the image modality) and a [`LayoutLMv3Tokenizer`]/[`LayoutLMv3TokenizerFast`] (for the text modality) to prepare all data for the model.
-- Regarding usage of [`LayoutLMv3Processor`], we refer to the [usage guide](layoutlmv2#usage-layoutlmv2processor) of its predecessor. 
+- Regarding usage of [`LayoutLMv3Processor`], we refer to the [usage guide](layoutlmv2#usage-layoutlmv2processor) of its predecessor.
 - Demo notebooks for LayoutLMv3 can be found [here](https://github.com/NielsRogge/Transformers-Tutorials/tree/master/LayoutLMv3).
 - Demo scripts can be found [here](https://github.com/huggingface/transformers/tree/main/examples/research_projects/layoutlmv3).
 
 <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/layoutlmv3_architecture.png"
-alt="drawing" width="600"/> 
+alt="drawing" width="600"/>
 
 <small> LayoutLMv3 architecture. Taken from the <a href="https://arxiv.org/abs/2204.08387">original paper</a>. </small>
 
-This model was contributed by [nielsr](https://huggingface.co/nielsr). The original code can be found [here](https://github.com/microsoft/unilm/tree/master/layoutlmv3).
+This model was contributed by [nielsr](https://huggingface.co/nielsr). The TensorFlow version of this model was added by [chriskoo](https://huggingface.co/chriskoo), [tokec](https://huggingface.co/tokec), and [lre](https://huggingface.co/lre). The original code can be found [here](https://github.com/microsoft/unilm/tree/master/layoutlmv3).
 
 
 ## LayoutLMv3Config
@@ -84,3 +84,23 @@ This model was contributed by [nielsr](https://huggingface.co/nielsr). The origi
 
 [[autodoc]] LayoutLMv3ForQuestionAnswering
     - forward
+
+## TFLayoutLMv3Model
+
+[[autodoc]] TFLayoutLMv3Model
+    - call
+
+## TFLayoutLMv3ForSequenceClassification
+
+[[autodoc]] TFLayoutLMv3ForSequenceClassification
+    - call
+
+## TFLayoutLMv3ForTokenClassification
+
+[[autodoc]] TFLayoutLMv3ForTokenClassification
+    - call
+
+## TFLayoutLMv3ForQuestionAnswering
+
+[[autodoc]] TFLayoutLMv3ForQuestionAnswering
+    - call

docs/source/en/model_doc/longt5.mdx

@@ -37,7 +37,7 @@ Tips:
 - [`LongT5ForConditionalGeneration`] is an extension of [`T5ForConditionalGeneration`] exchanging the traditional
 encoder *self-attention* layer with efficient either *local* attention or *transient-global* (*tglobal*) attention.
 - Unlike the T5 model, LongT5 does not use a task prefix. Furthermore, it uses a different pre-training objective
-inspired by the pre-training of `[PegasusForConditionalGeneration]`.
+inspired by the pre-training of [`PegasusForConditionalGeneration`].
 - LongT5 model is designed to work efficiently and very well on long-range *sequence-to-sequence* tasks where the
 input sequence exceeds commonly used 512 tokens. It is capable of handling input sequences of a length up to 16,384 tokens.
 - For *Local Attention*, the sparse sliding-window local attention operation allows a given token to attend only `r`

docs/source/en/model_doc/m2m_100.mdx

@@ -57,7 +57,7 @@ tgt_text = "La vie est comme une boîte de chocolat."
 
 model_inputs = tokenizer(src_text, text_target=tgt_text, return_tensors="pt")
 
-loss = model(**model_inputs, labels=labels)  # forward pass
+loss = model(**model_inputs).loss  # forward pass
 ```
 
 - Generation

docs/source/en/model_doc/markuplm.mdx

@@ -0,0 +1,246 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# MarkupLM
+
+## Overview
+
+The MarkupLM model was proposed in [MarkupLM: Pre-training of Text and Markup Language for Visually-rich Document
+Understanding](https://arxiv.org/abs/2110.08518) by Junlong Li, Yiheng Xu, Lei Cui, Furu Wei. MarkupLM is BERT, but
+applied to HTML pages instead of raw text documents. The model incorporates additional embedding layers to improve
+performance, similar to [LayoutLM](layoutlm).
+
+The model can be used for tasks like question answering on web pages or information extraction from web pages. It obtains
+state-of-the-art results on 2 important benchmarks:
+- [WebSRC](https://x-lance.github.io/WebSRC/), a dataset for Web-Based Structual Reading Comprehension (a bit like SQuAD but for web pages)
+- [SWDE](https://www.researchgate.net/publication/221299838_From_one_tree_to_a_forest_a_unified_solution_for_structured_web_data_extraction), a dataset
+for information extraction from web pages (basically named-entity recogntion on web pages)
+
+The abstract from the paper is the following:
+
+*Multimodal pre-training with text, layout, and image has made significant progress for Visually-rich Document
+Understanding (VrDU), especially the fixed-layout documents such as scanned document images. While, there are still a
+large number of digital documents where the layout information is not fixed and needs to be interactively and
+dynamically rendered for visualization, making existing layout-based pre-training approaches not easy to apply. In this
+paper, we propose MarkupLM for document understanding tasks with markup languages as the backbone such as
+HTML/XML-based documents, where text and markup information is jointly pre-trained. Experiment results show that the
+pre-trained MarkupLM significantly outperforms the existing strong baseline models on several document understanding
+tasks. The pre-trained model and code will be publicly available.*
+
+Tips:
+- In addition to `input_ids`, [`~MarkupLMModel.forward`] expects 2 additional inputs, namely `xpath_tags_seq` and `xpath_subs_seq`.
+These are the XPATH tags and subscripts respectively for each token in the input sequence.
+- One can use [`MarkupLMProcessor`] to prepare all data for the model. Refer to the [usage guide](#usage-markuplmprocessor) for more info.
+- Demo notebooks can be found [here](https://github.com/NielsRogge/Transformers-Tutorials/tree/master/MarkupLM).
+
+<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/markuplm_architecture.jpg"
+alt="drawing" width="600"/> 
+
+<small> MarkupLM architecture. Taken from the <a href="https://arxiv.org/abs/2110.08518">original paper.</a> </small>
+
+This model was contributed by [nielsr](https://huggingface.co/nielsr). The original code can be found [here](https://github.com/microsoft/unilm/tree/master/markuplm).
+
+## Usage: MarkupLMProcessor
+
+The easiest way to prepare data for the model is to use [`MarkupLMProcessor`], which internally combines a feature extractor
+([`MarkupLMFeatureExtractor`]) and a tokenizer ([`MarkupLMTokenizer`] or [`MarkupLMTokenizerFast`]). The feature extractor is
+used to extract all nodes and xpaths from the HTML strings, which are then provided to the tokenizer, which turns them into the
+token-level inputs of the model (`input_ids` etc.). Note that you can still use the feature extractor and tokenizer separately,
+if you only want to handle one of the two tasks.
+
+```python
+from transformers import MarkupLMFeatureExtractor, MarkupLMTokenizerFast, MarkupLMProcessor
+
+feature_extractor = MarkupLMFeatureExtractor()
+tokenizer = MarkupLMTokenizerFast.from_pretrained("microsoft/markuplm-base")
+processor = MarkupLMProcessor(feature_extractor, tokenizer)
+```
+
+In short, one can provide HTML strings (and possibly additional data) to [`MarkupLMProcessor`],
+and it will create the inputs expected by the model. Internally, the processor first uses
+[`MarkupLMFeatureExtractor`] to get a list of nodes and corresponding xpaths. The nodes and
+xpaths are then provided to [`MarkupLMTokenizer`] or [`MarkupLMTokenizerFast`], which converts them
+to token-level `input_ids`, `attention_mask`, `token_type_ids`, `xpath_subs_seq`, `xpath_tags_seq`.
+Optionally, one can provide node labels to the processor, which are turned into token-level `labels`.
+
+[`MarkupLMFeatureExtractor`] uses [Beautiful Soup](https://www.crummy.com/software/BeautifulSoup/bs4/doc/), a Python library for
+pulling data out of HTML and XML files, under the hood. Note that you can still use your own parsing solution of
+choice, and provide the nodes and xpaths yourself to [`MarkupLMTokenizer`] or [`MarkupLMTokenizerFast`].
+
+In total, there are 5 use cases that are supported by the processor. Below, we list them all. Note that each of these
+use cases work for both batched and non-batched inputs (we illustrate them for non-batched inputs).
+
+**Use case 1: web page classification (training, inference) + token classification (inference), parse_html = True**
+
+This is the simplest case, in which the processor will use the feature extractor to get all nodes and xpaths from the HTML.
+
+```python
+>>> from transformers import MarkupLMProcessor
+
+>>> processor = MarkupLMProcessor.from_pretrained("microsoft/markuplm-base")
+
+>>> html_string = """
+...  <!DOCTYPE html>
+...  <html>
+...  <head>
+...  <title>Hello world</title>
+...  </head>
+...  <body>
+
+...  <h1>Welcome</h1>
+...  <p>Here is my website.</p>
+
+...  </body>
+...  </html>"""
+
+>>> # note that you can also add provide all tokenizer parameters here such as padding, truncation
+>>> encoding = processor(html_string, return_tensors="pt")
+>>> print(encoding.keys())
+dict_keys(['input_ids', 'token_type_ids', 'attention_mask', 'xpath_tags_seq', 'xpath_subs_seq'])
+```
+
+**Use case 2: web page classification (training, inference) + token classification (inference), parse_html=False**
+
+In case one already has obtained all nodes and xpaths, one doesn't need the feature extractor. In that case, one should
+provide the nodes and corresponding xpaths themselves to the processor, and make sure to set `parse_html` to `False`.
+
+```python
+>>> from transformers import MarkupLMProcessor
+
+>>> processor = MarkupLMProcessor.from_pretrained("microsoft/markuplm-base")
+>>> processor.parse_html = False
+
+>>> nodes = ["hello", "world", "how", "are"]
+>>> xpaths = ["/html/body/div/li[1]/div/span", "/html/body/div/li[1]/div/span", "html/body", "html/body/div"]
+>>> encoding = processor(nodes=nodes, xpaths=xpaths, return_tensors="pt")
+>>> print(encoding.keys())
+dict_keys(['input_ids', 'token_type_ids', 'attention_mask', 'xpath_tags_seq', 'xpath_subs_seq'])
+```
+
+**Use case 3: token classification (training), parse_html=False**
+
+For token classification tasks (such as [SWDE](https://paperswithcode.com/dataset/swde)), one can also provide the
+corresponding node labels in order to train a model. The processor will then convert these into token-level `labels`.
+By default, it will only label the first wordpiece of a word, and label the remaining wordpieces with -100, which is the
+`ignore_index` of PyTorch's CrossEntropyLoss. In case you want all wordpieces of a word to be labeled, you can
+initialize the tokenizer with `only_label_first_subword` set to `False`.
+
+```python
+>>> from transformers import MarkupLMProcessor
+
+>>> processor = MarkupLMProcessor.from_pretrained("microsoft/markuplm-base")
+>>> processor.parse_html = False
+
+>>> nodes = ["hello", "world", "how", "are"]
+>>> xpaths = ["/html/body/div/li[1]/div/span", "/html/body/div/li[1]/div/span", "html/body", "html/body/div"]
+>>> node_labels = [1, 2, 2, 1]
+>>> encoding = processor(nodes=nodes, xpaths=xpaths, node_labels=node_labels, return_tensors="pt")
+>>> print(encoding.keys())
+dict_keys(['input_ids', 'token_type_ids', 'attention_mask', 'xpath_tags_seq', 'xpath_subs_seq', 'labels'])
+```
+
+**Use case 4: web page question answering (inference), parse_html=True**
+
+For question answering tasks on web pages, you can provide a question to the processor. By default, the
+processor will use the feature extractor to get all nodes and xpaths, and create [CLS] question tokens [SEP] word tokens [SEP].
+
+```python
+>>> from transformers import MarkupLMProcessor
+
+>>> processor = MarkupLMProcessor.from_pretrained("microsoft/markuplm-base")
+
+>>> html_string = """
+...  <!DOCTYPE html>
+...  <html>
+...  <head>
+...  <title>Hello world</title>
+...  </head>
+...  <body>
+
+...  <h1>Welcome</h1>
+...  <p>My name is Niels.</p>
+
+...  </body>
+...  </html>"""
+
+>>> question = "What's his name?"
+>>> encoding = processor(html_string, questions=question, return_tensors="pt")
+>>> print(encoding.keys())
+dict_keys(['input_ids', 'token_type_ids', 'attention_mask', 'xpath_tags_seq', 'xpath_subs_seq'])
+```
+
+**Use case 5: web page question answering (inference), apply_ocr=False**
+
+For question answering tasks (such as WebSRC), you can provide a question to the processor. If you have extracted
+all nodes and xpaths yourself, you can provide them directly to the processor. Make sure to set `parse_html` to `False`.
+
+```python
+>>> from transformers import MarkupLMProcessor
+
+>>> processor = MarkupLMProcessor.from_pretrained("microsoft/markuplm-base")
+>>> processor.parse_html = False
+
+>>> nodes = ["hello", "world", "how", "are"]
+>>> xpaths = ["/html/body/div/li[1]/div/span", "/html/body/div/li[1]/div/span", "html/body", "html/body/div"]
+>>> question = "What's his name?"
+>>> encoding = processor(nodes=nodes, xpaths=xpaths, questions=question, return_tensors="pt")
+>>> print(encoding.keys())
+dict_keys(['input_ids', 'token_type_ids', 'attention_mask', 'xpath_tags_seq', 'xpath_subs_seq'])
+```
+
+## MarkupLMConfig
+
+[[autodoc]] MarkupLMConfig
+    - all
+
+## MarkupLMFeatureExtractor
+
+[[autodoc]] MarkupLMFeatureExtractor
+    - __call__
+
+## MarkupLMTokenizer
+
+[[autodoc]] MarkupLMTokenizer
+    - build_inputs_with_special_tokens
+    - get_special_tokens_mask
+    - create_token_type_ids_from_sequences
+    - save_vocabulary
+
+## MarkupLMTokenizerFast
+
+[[autodoc]] MarkupLMTokenizerFast
+    - all
+
+## MarkupLMProcessor
+
+[[autodoc]] MarkupLMProcessor
+    - __call__
+
+## MarkupLMModel
+
+[[autodoc]] MarkupLMModel
+    - forward
+
+## MarkupLMForSequenceClassification
+
+[[autodoc]] MarkupLMForSequenceClassification
+    - forward
+
+## MarkupLMForTokenClassification
+
+[[autodoc]] MarkupLMForTokenClassification
+    - forward
+
+## MarkupLMForQuestionAnswering
+
+[[autodoc]] MarkupLMForQuestionAnswering
+    - forward

docs/source/en/model_doc/maskformer.mdx

@@ -58,6 +58,7 @@ This model was contributed by [francesco](https://huggingface.co/francesco). The
     - encode_inputs
     - post_process_segmentation
     - post_process_semantic_segmentation
+    - post_process_instance_segmentation
     - post_process_panoptic_segmentation
 
 ## MaskFormerModel

docs/source/en/model_doc/mobilevit.mdx

@@ -22,12 +22,40 @@ The abstract from the paper is the following:
 
 Tips:
 
-- MobileViT is more like a CNN than a Transformer model. It does not work on sequence data but on batches of images. Unlike ViT, there are no embeddings. The backbone model outputs a feature map.
+- MobileViT is more like a CNN than a Transformer model. It does not work on sequence data but on batches of images. Unlike ViT, there are no embeddings. The backbone model outputs a feature map. You can follow [this tutorial](https://keras.io/examples/vision/mobilevit) for a lightweight introduction.
 - One can use [`MobileViTFeatureExtractor`] to prepare images for the model. Note that if you do your own preprocessing, the pretrained checkpoints expect images to be in BGR pixel order (not RGB).
 - The available image classification checkpoints are pre-trained on [ImageNet-1k](https://huggingface.co/datasets/imagenet-1k) (also referred to as ILSVRC 2012, a collection of 1.3 million images and 1,000 classes).
 - The segmentation model uses a [DeepLabV3](https://arxiv.org/abs/1706.05587) head. The available semantic segmentation checkpoints are pre-trained on [PASCAL VOC](http://host.robots.ox.ac.uk/pascal/VOC/). 
+- As the name suggests MobileViT was desgined to be performant and efficient on mobile phones. The TensorFlow versions of the MobileViT models are fully compatible with [TensorFlow Lite](https://www.tensorflow.org/lite). 
 
-This model was contributed by [matthijs](https://huggingface.co/Matthijs). The original code and weights can be found [here](https://github.com/apple/ml-cvnets).
+  You can use the following code to convert a MobileViT checkpoint (be it image classification or semantic segmentation) to generate a 
+  TensorFlow Lite model:
+
+```py
+from transformers import TFMobileViTForImageClassification
+import tensorflow as tf
+
+
+model_ckpt = "apple/mobilevit-xx-small"
+model = TFMobileViTForImageClassification.from_pretrained(model_ckpt)
+
+converter = tf.lite.TFLiteConverter.from_keras_model(model)
+converter.optimizations = [tf.lite.Optimize.DEFAULT]
+converter.target_spec.supported_ops = [
+    tf.lite.OpsSet.TFLITE_BUILTINS,
+    tf.lite.OpsSet.SELECT_TF_OPS,
+]
+tflite_model = converter.convert()
+tflite_filename = model_ckpt.split("/")[-1] + ".tflite"
+with open(tflite_filename, "wb") as f:
+    f.write(tflite_model)
+```
+
+  The resulting model will be just **about an MB** making it a good fit for mobile applications where resources and network
+  bandwidth can be constrained. 
+
+
+This model was contributed by [matthijs](https://huggingface.co/Matthijs). The TensorFlow version of the model was contributed by [sayakpaul](https://huggingface.co/sayakpaul). The original code and weights can be found [here](https://github.com/apple/ml-cvnets).
 
 
 ## MobileViTConfig
@@ -38,6 +66,7 @@ This model was contributed by [matthijs](https://huggingface.co/Matthijs). The o
 
 [[autodoc]] MobileViTFeatureExtractor
     - __call__
+    - post_process_semantic_segmentation
 
 ## MobileViTModel
 
@@ -53,3 +82,18 @@ This model was contributed by [matthijs](https://huggingface.co/Matthijs). The o
 
 [[autodoc]] MobileViTForSemanticSegmentation
     - forward
+
+## TFMobileViTModel
+
+[[autodoc]] TFMobileViTModel
+    - call 
+
+## TFMobileViTForImageClassification
+
+[[autodoc]] TFMobileViTForImageClassification
+    - call 
+
+## TFMobileViTForSemanticSegmentation
+
+[[autodoc]] TFMobileViTForSemanticSegmentation
+    - call 

docs/source/en/model_doc/opt.mdx

@@ -59,6 +59,11 @@ The original code can be found [here](https://github.com/facebookresearch/metase
 [[autodoc]] OPTForSequenceClassification
     - forward
 
+## OPTForQuestionAnswering
+
+[[autodoc]] OPTForQuestionAnswering
+    - forward
+
 ## FlaxOPTModel
 
 [[autodoc]] FlaxOPTModel

docs/source/en/model_doc/owlvit.mdx

@@ -57,8 +57,8 @@ OWL-ViT is a zero-shot text-conditioned object detection model. OWL-ViT uses [CL
 ...     box = [round(i, 2) for i in box.tolist()]
 ...     if score >= score_threshold:
 ...         print(f"Detected {text[label]} with confidence {round(score.item(), 3)} at location {box}")
-Detected a photo of a cat with confidence 0.243 at location [1.42, 50.69, 308.58, 370.48]
-Detected a photo of a cat with confidence 0.298 at location [348.06, 20.56, 642.33, 372.61]
+Detected a photo of a cat with confidence 0.707 at location [324.97, 20.44, 640.58, 373.29]
+Detected a photo of a cat with confidence 0.717 at location [1.46, 55.26, 315.55, 472.17]
 ```
 
 This model was contributed by [adirik](https://huggingface.co/adirik). The original code can be found [here](https://github.com/google-research/scenic/tree/main/scenic/projects/owl_vit).

docs/source/en/model_doc/pegasus_x.mdx

@@ -0,0 +1,45 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# PEGASUS-X
+
+## Overview
+
+The PEGASUS-X model was proposed in [Investigating Efficiently Extending Transformers for Long Input Summarization](https://arxiv.org/abs/2208.04347)  by Jason Phang, Yao Zhao and Peter J. Liu.
+
+PEGASUS-X (PEGASUS eXtended) extends the PEGASUS models for long input summarization through additional long input pretraining and using staggered block-local attention with global tokens in the encoder.
+
+The abstract from the paper is the following:
+
+*While large pretrained Transformer models have proven highly capable at tackling natural language tasks, handling long sequence inputs continues to be a significant challenge. One such task is long input summarization, where inputs are longer than the maximum input context of most pretrained models. Through an extensive set of experiments, we investigate what model architectural changes and pretraining paradigms can most efficiently adapt a pretrained Transformer for long input summarization. We find that a staggered, block-local Transformer with global encoder tokens strikes a good balance of performance and efficiency, and that an additional pretraining phase on long sequences meaningfully improves downstream summarization performance. Based on our findings, we introduce PEGASUS-X, an extension of the PEGASUS model with additional long input pretraining to handle inputs of up to 16K tokens. PEGASUS-X achieves strong performance on long input summarization tasks comparable with much larger models while adding few additional parameters and not requiring model parallelism to train.*
+
+Tips:
+
+* PEGASUS-X uses the same tokenizer as PEGASUS.
+
+This model was contributed by [zphang](<https://huggingface.co/zphang). The original code can be found [here](https://github.com/google-research/pegasus).
+
+## PegasusXConfig
+
+[[autodoc]] PegasusXConfig
+
+
+## PegasusXModel
+
+[[autodoc]] PegasusXModel
+    - forward
+
+
+## PegasusXForConditionalGeneration
+
+[[autodoc]] PegasusXForConditionalGeneration
+    - forward

docs/source/en/model_doc/segformer.mdx

@@ -93,6 +93,7 @@ SegFormer's results on the segmentation datasets like ADE20k, refer to the [pape
 
 [[autodoc]] SegformerFeatureExtractor
     - __call__
+    - post_process_semantic_segmentation
 
 ## SegformerModel
 

docs/source/en/model_doc/t5.mdx

@@ -187,12 +187,15 @@ ignored. The code example below illustrates all of this.
 
 >>> # encode the targets
 >>> target_encoding = tokenizer(
-...     [output_sequence_1, output_sequence_2], padding="longest", max_length=max_target_length, truncation=True
+...     [output_sequence_1, output_sequence_2],
+...     padding="longest",
+...     max_length=max_target_length,
+...     truncation=True,
+...     return_tensors="pt",
 ... )
 >>> labels = target_encoding.input_ids
 
 >>> # replace padding token id's of the labels by -100 so it's ignored by the loss
->>> labels = torch.tensor(labels)
 >>> labels[labels == tokenizer.pad_token_id] = -100
 
 >>> # forward pass

docs/source/en/model_doc/time_series_transformer.mdx

@@ -0,0 +1,73 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# Time Series Transformer
+
+<Tip>
+
+This is a recently introduced model so the API hasn't been tested extensively. There may be some bugs or slight
+breaking changes to fix it in the future. If you see something strange, file a [Github Issue](https://github.com/huggingface/transformers/issues/new?assignees=&labels=&template=bug-report.md&title).
+
+</Tip>
+
+## Overview
+
+The Time Series Transformer model is a vanilla encoder-decoder Transformer for time series forecasting.
+
+Tips:
+
+- Similar to other models in the library, [`TimeSeriesTransformerModel`] is the raw Transformer without any head on top, and [`TimeSeriesTransformerForPrediction`]
+adds a distribution head on top of the former, which can be used for time-series forecasting. Note that this is a so-called probabilistic forecasting model, not a
+point forecasting model. This means that the model learns a distribution, from which one can sample. The model doesn't directly output values.
+- [`TimeSeriesTransformerForPrediction`] consists of 2 blocks: an encoder, which takes a `context_length` of time series values as input (called `past_values`),
+and a decoder, which predicts a `prediction_length` of time series values into the future (called `future_values`). During training, one needs to provide
+pairs of (`past_values` and `future_values`) to the model.
+- In addition to the raw (`past_values` and `future_values`), one typically provides additional features to the model. These can be the following:
+    - `past_time_features`: temporal features which the model will add to `past_values`. These serve as "positional encodings" for the Transformer encoder.
+    Examples are "day of the month", "month of the year", etc. as scalar values (and then stacked together as a vector).
+    e.g. if a given time-series value was obtained on the 11th of August, then one could have [11, 8] as time feature vector (11 being "day of the month", 8 being "month of the year").
+    - `future_time_features`: temporal features which the model will add to `future_values`. These serve as "positional encodings" for the Transformer decoder.
+    Examples are "day of the month", "month of the year", etc. as scalar values (and then stacked together as a vector).
+    e.g. if a given time-series value was obtained on the 11th of August, then one could have [11, 8] as time feature vector (11 being "day of the month", 8 being "month of the year").
+    - `static_categorical_features`: categorical features which are static over time (i.e., have the same value for all `past_values` and `future_values`).
+    An example here is the store ID or region ID that identifies a given time-series.
+    Note that these features need to be known for ALL data points (also those in the future).
+    - `static_real_features`: real-valued features which are static over time (i.e., have the same value for all `past_values` and `future_values`).
+    An example here is the image representation of the product for which you have the time-series values (like the [ResNet](resnet) embedding of a "shoe" picture,
+    if your time-series is about the sales of shoes).
+    Note that these features need to be known for ALL data points (also those in the future).
+- The model is trained using "teacher-forcing", similar to how a Transformer is trained for machine translation. This means that, during training, one shifts the
+`future_values` one position to the right as input to the decoder, prepended by the last value of `past_values`. At each time step, the model needs to predict the
+next target. So the set-up of training is similar to a GPT model for language, except that there's no notion of `decoder_start_token_id` (we just use the last value
+of the context as initial input for the decoder).
+- At inference time, we give the final value of the `past_values` as input to the decoder. Next, we can sample from the model to make a prediction at the next time step,
+which is then fed to the decoder in order to make the next prediction (also called autoregressive generation).
+
+
+This model was contributed by [kashif](<https://huggingface.co/kashif).
+
+
+## TimeSeriesTransformerConfig
+
+[[autodoc]] TimeSeriesTransformerConfig
+
+
+## TimeSeriesTransformerModel
+
+[[autodoc]] TimeSeriesTransformerModel
+    - forward
+
+
+## TimeSeriesTransformerForPrediction
+
+[[autodoc]] TimeSeriesTransformerForPrediction
+    - forward

docs/source/en/model_doc/videomae.mdx

@@ -0,0 +1,60 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# VideoMAE
+
+## Overview
+
+The VideoMAE model was proposed in [VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training](https://arxiv.org/abs/2203.12602) by Zhan Tong, Yibing Song, Jue Wang, Limin Wang.
+VideoMAE extends masked auto encoders ([MAE](vit_mae)) to video, claiming state-of-the-art performance on several video classification benchmarks.
+
+The abstract from the paper is the following:
+
+*Pre-training video transformers on extra large-scale datasets is generally required to achieve premier performance on relatively small datasets. In this paper, we show that video masked autoencoders (VideoMAE) are data-efficient learners for self-supervised video pre-training (SSVP). We are inspired by the recent ImageMAE and propose customized video tube masking and reconstruction. These simple designs turn out to be effective for overcoming information leakage caused by the temporal correlation during video reconstruction. We obtain three important findings on SSVP: (1) An extremely high proportion of masking ratio (i.e., 90% to 95%) still yields favorable performance of VideoMAE. The temporally redundant video content enables higher masking ratio than that of images. (2) VideoMAE achieves impressive results on very small datasets (i.e., around 3k-4k videos) without using any extra data. This is partially ascribed to the challenging task of video reconstruction to enforce high-level structure learning. (3) VideoMAE shows that data quality is more important than data quantity for SSVP. Domain shift between pre-training and target datasets are important issues in SSVP. Notably, our VideoMAE with the vanilla ViT backbone can achieve 83.9% on Kinects-400, 75.3% on Something-Something V2, 90.8% on UCF101, and 61.1% on HMDB51 without using any extra data.*
+
+Tips:
+
+- One can use [`VideoMAEFeatureExtractor`] to prepare videos for the model. It will resize + normalize all frames of a video for you.
+- [`VideoMAEForPreTraining`] includes the decoder on top for self-supervised pre-training.
+
+<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/videomae_architecture.jpeg"
+alt="drawing" width="600"/> 
+
+<small> VideoMAE pre-training. Taken from the <a href="https://arxiv.org/abs/2203.12602">original paper</a>. </small>
+
+This model was contributed by [nielsr](https://huggingface.co/nielsr).
+The original code can be found [here](https://github.com/MCG-NJU/VideoMAE).
+
+
+## VideoMAEConfig
+
+[[autodoc]] VideoMAEConfig
+
+## VideoMAEFeatureExtractor
+
+[[autodoc]] VideoMAEFeatureExtractor
+    - __call__
+
+## VideoMAEModel
+
+[[autodoc]] VideoMAEModel
+    - forward
+
+## VideoMAEForPreTraining
+
+[[autodoc]] transformers.VideoMAEForPreTraining
+    - forward
+
+## VideoMAEForVideoClassification
+
+[[autodoc]] transformers.VideoMAEForVideoClassification
+    - forward

docs/source/en/model_doc/vit.mdx

@@ -12,13 +12,6 @@ specific language governing permissions and limitations under the License.
 
 # Vision Transformer (ViT)
 
-<Tip>
-
-This is a recently introduced model so the API hasn't been tested extensively. There may be some bugs or slight
-breaking changes to fix it in the future. If you see something strange, file a [Github Issue](https://github.com/huggingface/transformers/issues/new?assignees=&labels=&template=bug-report.md&title).
-
-</Tip>
-
 ## Overview
 
 The Vision Transformer (ViT) model was proposed in [An Image is Worth 16x16 Words: Transformers for Image Recognition
@@ -63,6 +56,11 @@ Tips:
   language modeling). With this approach, the smaller ViT-B/16 model achieves 79.9% accuracy on ImageNet, a significant
   improvement of 2% to training from scratch, but still 4% behind supervised pre-training.
 
+<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/vit_architecture.jpg"
+alt="drawing" width="600"/> 
+
+<small> ViT architecture. Taken from the <a href="https://arxiv.org/abs/2010.11929">original paper.</a> </small>
+
 Following the original Vision Transformer, some follow-up works have been made:
 
 - [DeiT](deit) (Data-efficient Image Transformers) by Facebook AI. DeiT models are distilled vision transformers.

docs/source/en/model_doc/vit_msn.mdx

@@ -0,0 +1,64 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# ViTMSN
+
+## Overview
+
+The ViTMSN model was proposed in [Masked Siamese Networks for Label-Efficient Learning](https://arxiv.org/abs/2204.07141) by Mahmoud Assran, Mathilde Caron, Ishan Misra, Piotr Bojanowski, Florian Bordes,
+Pascal Vincent, Armand Joulin, Michael Rabbat, Nicolas Ballas. The paper presents a joint-embedding architecture to match the prototypes
+of masked patches with that of the unmasked patches. With this setup, their method yields excellent performance in the low-shot and extreme low-shot
+regimes.
+
+The abstract from the paper is the following:
+
+*We propose Masked Siamese Networks (MSN), a self-supervised learning framework for learning image representations. Our
+approach matches the representation of an image view containing randomly masked patches to the representation of the original
+unmasked image. This self-supervised pre-training strategy is particularly scalable when applied to Vision Transformers since only the
+unmasked patches are processed by the network. As a result, MSNs improve the scalability of joint-embedding architectures,
+while producing representations of a high semantic level that perform competitively on low-shot image classification. For instance,
+on ImageNet-1K, with only 5,000 annotated images, our base MSN model achieves 72.4% top-1 accuracy,
+and with 1% of ImageNet-1K labels, we achieve 75.7% top-1 accuracy, setting a new state-of-the-art for self-supervised learning on this benchmark.*
+
+Tips:
+
+- MSN (masked siamese networks) is a method for self-supervised pre-training of Vision Transformers (ViTs). The pre-training
+objective is to match the prototypes assigned to the unmasked views of the images to that of the masked views of the same images.
+- The authors have only released pre-trained weights of the backbone (ImageNet-1k pre-training). So, to use that on your own image classification dataset,
+use the [`ViTMSNForImageClassification`] class which is initialized from [`ViTMSNModel`]. Follow
+[this notebook](https://github.com/huggingface/notebooks/blob/main/examples/image_classification.ipynb) for a detailed tutorial on fine-tuning.
+- MSN is particularly useful in the low-shot and extreme low-shot regimes. Notably, it achieves 75.7% top-1 accuracy with only 1% of ImageNet-1K
+labels when fine-tuned.
+
+
+<img src="https://i.ibb.co/W6PQMdC/Screenshot-2022-09-13-at-9-08-40-AM.png" alt="drawing" width="600"/> 
+
+<small> MSN architecture. Taken from the <a href="https://arxiv.org/abs/2204.07141">original paper.</a> </small>
+
+This model was contributed by [sayakpaul](https://huggingface.co/sayakpaul). The original code can be found [here](https://github.com/facebookresearch/msn). 
+
+
+## ViTMSNConfig
+
+[[autodoc]] ViTMSNConfig
+
+
+## ViTMSNModel
+
+[[autodoc]] ViTMSNModel
+    - forward
+
+
+## ViTMSNForImageClassification
+
+[[autodoc]] ViTMSNForImageClassification
+    - forward

docs/source/en/model_doc/whisper.mdx

@@ -0,0 +1,79 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# Whisper
+
+## Overview
+
+The Whisper model was proposed in [Robust Speech Recognition via Large-Scale Weak Supervision](https://cdn.openai.com/papers/whisper.pdf) by Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, Ilya Sutskever.
+
+The abstract from the paper is the following:
+
+*We study the capabilities of speech processing systems trained simply to predict large amounts of transcripts of audio on the internet. When scaled to 680,000 hours of multilingual and multitask supervision, the resulting models generalize well to standard benchmarks and are often competitive with prior fully supervised results but in a zeroshot transfer setting without the need for any finetuning. When compared to humans, the models approach their accuracy and robustness. We are releasing models and inference code to serve as a foundation for further work on robust speech processing.*
+
+
+Tips:
+
+- The model usually performs well without requiring any finetuning.
+- The architecture follows a classic encoder-decoder architecture, which means that it relies on the [`~generation_utils.GenerationMixin.generate`] function for inference.
+- One can use [`WhisperProcessor`] to prepare audio for the model, and decode the predicted ID's back into text.
+
+This model was contributed by [Arthur Zucker](https://huggingface.co/ArthurZ). The Tensorflow version of this model was contributed by [amyeroberts](https://huggingface.co/amyeroberts).
+The original code can be found [here](https://github.com/openai/whisper).
+
+
+## WhisperConfig
+
+[[autodoc]] WhisperConfig
+
+## WhisperTokenizer
+
+[[autodoc]] WhisperTokenizer
+    - build_inputs_with_special_tokens
+    - get_special_tokens_mask
+    - create_token_type_ids_from_sequences
+    - save_vocabulary
+
+## WhisperFeatureExtractor
+
+[[autodoc]] WhisperFeatureExtractor
+    - __call__
+
+## WhisperProcessor
+
+[[autodoc]] WhisperProcessor
+    - __call__
+    - from_pretrained
+    - save_pretrained
+    - batch_decode
+    - decode
+
+## WhisperModel
+
+[[autodoc]] WhisperModel
+    - forward
+
+## WhisperForConditionalGeneration
+
+[[autodoc]] WhisperForConditionalGeneration
+    - forward
+
+
+## TFWhisperModel
+
+[[autodoc]] TFWhisperModel
+    - call
+
+## TFWhisperForConditionalGeneration
+
+[[autodoc]] TFWhisperForConditionalGeneration
+    - call

docs/source/en/model_doc/xclip.mdx

@@ -0,0 +1,70 @@
+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# X-CLIP
+
+## Overview
+
+The X-CLIP model was proposed in [Expanding Language-Image Pretrained Models for General Video Recognition](https://arxiv.org/abs/2208.02816) by Bolin Ni, Houwen Peng, Minghao Chen, Songyang Zhang, Gaofeng Meng, Jianlong Fu, Shiming Xiang, Haibin Ling.
+X-CLIP is a minimal extension of [CLIP](clip) for video. The model consists of a text encoder, a cross-frame vision encoder, a multi-frame integration Transformer, and a video-specific prompt generator.
+
+The abstract from the paper is the following:
+
+*Contrastive language-image pretraining has shown great success in learning visual-textual joint representation from web-scale data, demonstrating remarkable "zero-shot" generalization ability for various image tasks. However, how to effectively expand such new language-image pretraining methods to video domains is still an open problem. In this work, we present a simple yet effective approach that adapts the pretrained language-image models to video recognition directly, instead of pretraining a new model from scratch. More concretely, to capture the long-range dependencies of frames along the temporal dimension, we propose a cross-frame attention mechanism that explicitly exchanges information across frames. Such module is lightweight and can be plugged into pretrained language-image models seamlessly. Moreover, we propose a video-specific prompting scheme, which leverages video content information for generating discriminative textual prompts. Extensive experiments demonstrate that our approach is effective and can be generalized to different video recognition scenarios. In particular, under fully-supervised settings, our approach achieves a top-1 accuracy of 87.1% on Kinectics-400, while using 12 times fewer FLOPs compared with Swin-L and ViViT-H. In zero-shot experiments, our approach surpasses the current state-of-the-art methods by +7.6% and +14.9% in terms of top-1 accuracy under two popular protocols. In few-shot scenarios, our approach outperforms previous best methods by +32.1% and +23.1% when the labeled data is extremely limited.*
+
+Tips:
+
+- Usage of X-CLIP is identical to [CLIP](clip).
+- Demo notebooks for X-CLIP can be found [here](https://github.com/NielsRogge/Transformers-Tutorials/tree/master/X-CLIP).
+
+<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/xclip_architecture.png"
+alt="drawing" width="600"/> 
+
+<small> X-CLIP architecture. Taken from the <a href="https://arxiv.org/abs/2208.02816">original paper.</a> </small>
+
+This model was contributed by [nielsr](https://huggingface.co/nielsr).
+The original code can be found [here](https://github.com/microsoft/VideoX/tree/master/X-CLIP).
+
+
+## XCLIPProcessor
+
+[[autodoc]] XCLIPProcessor
+
+## XCLIPConfig
+
+[[autodoc]] XCLIPConfig
+    - from_text_vision_configs
+
+## XCLIPTextConfig
+
+[[autodoc]] XCLIPTextConfig
+
+## XCLIPVisionConfig
+
+[[autodoc]] XCLIPVisionConfig
+
+## XCLIPModel
+
+[[autodoc]] XCLIPModel
+    - forward
+    - get_text_features
+    - get_video_features
+
+## XCLIPTextModel
+
+[[autodoc]] XCLIPTextModel
+    - forward
+
+## XCLIPVisionModel
+
+[[autodoc]] XCLIPVisionModel
+    - forward

docs/source/en/model_doc/xglm.mdx

@@ -64,6 +64,16 @@ This model was contributed by [Suraj](https://huggingface.co/valhalla). The orig
 [[autodoc]] XGLMForCausalLM
     - forward
 
+## TFXGLMModel
+
+[[autodoc]] TFXGLMModel
+    - call
+
+## TFXGLMForCausalLM
+
+[[autodoc]] TFXGLMForCausalLM
+    - call
+
 ## FlaxXGLMModel
 
 [[autodoc]] FlaxXGLMModel

docs/source/en/model_sharing.mdx

@@ -225,4 +225,4 @@ To make sure users understand your model's capabilities, limitations, potential
 * Manually creating and uploading a `README.md` file.
 * Clicking on the **Edit model card** button in your model repository.
 
-Take a look at the DistilBert [model card](https://huggingface.co/distilbert-base-uncased) for a good example of the type of information a model card should include. For more details about other options you can control in the `README.md` file such as a model's carbon footprint or widget examples, refer to the documentation [here](https://huggingface.co/docs/hub/model-repos).
+Take a look at the DistilBert [model card](https://huggingface.co/distilbert-base-uncased) for a good example of the type of information a model card should include. For more details about other options you can control in the `README.md` file such as a model's carbon footprint or widget examples, refer to the documentation [here](https://huggingface.co/docs/hub/models-cards).

docs/source/en/perf_infer_gpu_many.mdx

@@ -11,4 +11,9 @@ an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express o
 
 # Efficient Inference on a Multiple GPUs
 
-This document will be completed soon with information on how to infer on a multiple GPUs. In the meantime you can check out [the guide for training on a single GPU](perf_train_gpu_one) and [the guide for inference on CPUs](perf_infer_cpu).
+This document contains information on how to efficiently infer on a multiple GPUs. 
+<Tip>
+
+Note: A multi GPU setup can use the majority of the strategies described in the [single GPU section](./perf_infer_gpu_one). You must be aware of simple techniques, though, that can be used for a better usage.
+
+</Tip>

docs/source/en/perf_infer_gpu_one.mdx

@@ -11,4 +11,66 @@ an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express o
 
 # Efficient Inference on a Single GPU
 
-This document will be completed soon with information on how to infer on a single GPU. In the meantime you can check out [the guide for training on a single GPU](perf_train_gpu_one) and [the guide for inference on CPUs](perf_infer_cpu).
+This document will be completed soon with information on how to infer on a single GPU. In the meantime you can check out [the guide for training on a single GPU](perf_train_gpu_one) and [the guide for inference on CPUs](perf_infer_cpu).
+
+## `bitsandbytes` integration for Int8 mixed-precision matrix decomposition
+
+Note that this feature is also totally applicable in a multi GPU setup as well.
+
+From the paper [`LLM.int8() : 8-bit Matrix Multiplication for Transformers at Scale`](https://arxiv.org/abs/2208.07339), we support HuggingFace integration for all models in the Hub with a few lines of code. 
+The method reduce `nn.Linear` size by 2 for `float16` and `bfloat16` weights and by 4 for `float32` weights, with close to no impact to the quality by operating on the outliers in half-precision.
+
+![HFxbitsandbytes.png](https://s3.amazonaws.com/moonup/production/uploads/1659861207959-62441d1d9fdefb55a0b7d12c.png)
+
+Int8 mixed-precision matrix decomposition works by separating a matrix multiplication into two streams: (1) a systematic feature outlier stream matrix multiplied in fp16 (0.01%), (2) a regular stream of int8 matrix multiplication (99.9%). With this method, int8 inference with no predictive degradation is possible for very large models.
+For more details regarding the method, check out the [paper](https://arxiv.org/abs/2208.07339) or our [blogpost about the integration](https://huggingface.co/blog/hf-bitsandbytes-integration).
+
+![MixedInt8.gif](https://s3.amazonaws.com/moonup/production/uploads/1660567469965-62441d1d9fdefb55a0b7d12c.gif)
+
+Note, that you would require a GPU to run mixed-8bit models as the kernels have been compiled for GPUs only. Make sure that you have enough GPU memory to store the quarter (or half if your model weights are in half precision) of the model before using this feature. 
+Below are some notes to help you use this module, or follow the demos on [Google colab](#colab-demos).
+
+### Requirements
+
+- Make sure you run that on NVIDIA GPUs that support 8-bit tensor cores (Turing, Ampere or newer architectures - e.g. T4, RTX20s RTX30s, A40-A100).
+- Install the correct version of `bitsandbytes` by running:
+`pip install bitsandbytes>=0.31.5`
+- Install `accelerate`
+`pip install accelerate>=0.12.0`
+
+### Running mixed-int8 models - single GPU setup
+
+After installing the required libraries, the way to load your mixed 8-bit model is as follows:
+```py
+model_name = "bigscience/bloom-2b5"
+model_8bit = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto", load_in_8bit=True)
+```
+
+### Running mixed-int8 models - multi GPU setup
+
+The way to load your mixed 8-bit model in multiple GPUs is as follows (same command as single GPU setup):
+```py
+model_name = "bigscience/bloom-2b5"
+model_8bit = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto", load_in_8bit=True)
+```
+But you can control the GPU RAM you want to allocate on each GPU using `accelerate`. Use the `max_memory` argument as follows:
+
+```py
+max_memory_mapping = {0: "1GB", 1: "2GB"}
+model_name = "bigscience/bloom-3b"
+model_8bit = AutoModelForCausalLM.from_pretrained(
+    model_name, device_map="auto", load_in_8bit=True, max_memory=max_memory_mapping
+)
+```
+In this example, the first GPU will use 1GB of memory and the second 2GB.
+
+### Colab demos
+
+With this method you can infer on models that were not possible to infer on a Google Colab before. 
+Check out the demo for running T5-11b (42GB in fp32)! Using 8-bit quantization on Google Colab:
+
+[![Open In Colab: T5-11b demo](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1YORPWx4okIHXnjW7MSAidXN29mPVNT7F?usp=sharing)
+
+Or this demo for BLOOM-3B:
+
+[![Open In Colab: BLOOM-3b demo](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1qOjXfQIAULfKvZqwCen8-MoWKGdSatZ4?usp=sharing)

docs/source/en/perf_train_cpu_many.mdx

@@ -36,8 +36,36 @@ pip install oneccl_bind_pt=={pytorch_version} -f https://software.intel.com/ipex
 ```
 where `{pytorch_version}` should be your PyTorch version, for instance 1.12.0.
 Check more approaches for [oneccl_bind_pt installation](https://github.com/intel/torch-ccl).
+Versions of oneCCL and PyTorch must match.
 
-### Usage in Trainer
+<Tip warning={true}>
+
+oneccl_bindings_for_pytorch 1.12.0 prebuilt wheel does not work with PyTorch 1.12.1 (it is for PyTorch 1.12.0)
+
+</Tip>
+
+## Intel® MPI library
+Use this standards-based MPI implementation to deliver flexible, efficient, scalable cluster messaging on Intel® architecture. This component is part of the Intel® oneAPI HPC Toolkit.
+
+oneccl_bindings_for_pytorch is installed along with the MPI tool set. Need to source the environment before using it.
+
+for Intel® oneCCL 1.12.0
+```
+oneccl_bindings_for_pytorch_path=$(python -c "from oneccl_bindings_for_pytorch import cwd; print(cwd)")
+source $oneccl_bindings_for_pytorch_path/env/setvars.sh
+```
+
+for Intel® oneCCL whose version < 1.12.0
+```
+torch_ccl_path=$(python -c "import torch; import torch_ccl; import os;  print(os.path.abspath(os.path.dirname(torch_ccl.__file__)))")
+source $torch_ccl_path/env/setvars.sh
+```
+
+
+The following "Usage in Trainer" takes mpirun in Intel® MPI library as an example.
+
+
+## Usage in Trainer
 To enable multi CPU distributed training in the Trainer with the ccl backend, users should add **`--xpu_backend ccl`** in the command arguments.
 
 Let's see an example with the [question-answering example](https://github.com/huggingface/transformers/tree/main/examples/pytorch/question-answering)

docs/source/en/perf_train_gpu_one.mdx

@@ -288,7 +288,7 @@ Even when we set the batch size to 1 and use gradient accumulation we can still
 
 Gradient checkpointing strikes a compromise between the two approaches and saves strategically selected activations throughout the computational graph so only a fraction of the activations need to be re-computed for the gradients. See [this great article](https://medium.com/tensorflow/fitting-larger-networks-into-memory-583e3c758ff9) explaining the ideas behind gradient checkpointing.
 
-To enable gradient checkpointing in the [`Trainer`] we only need ot pass it as a flag to the [`TrainingArguments`]. Everything else is handled under the hood:
+To enable gradient checkpointing in the [`Trainer`] we only need to pass it as a flag to the [`TrainingArguments`]. Everything else is handled under the hood:
 
 ```py
 training_args = TrainingArguments(
@@ -425,7 +425,7 @@ $ python examples/pytorch/translation/run_translation.py -h | grep "\-optim"
 
 For example, if you have [NVIDIA/apex](https://github.com/NVIDIA/apex) installed `--optim adamw_apex_fused` will give you the fastest training experience among all supported AdamW optimizers.
 
-On the other hand [8bit BNB optimizer](https://github.com/facebookresearch/bitsandbytes) can save 3/4 of memory normally used by a typical AdamW optimizer if it is configured to quantize all optimizer states, but in some situations only some optimizer states are quintized and then more memory is used. XXX: update once  https://github.com/huggingface/transformers/pull/15622 is merged.
+On the other hand [8bit BNB optimizer](https://github.com/facebookresearch/bitsandbytes) can save 3/4 of memory normally used by a typical AdamW optimizer if it is configured to quantize all optimizer states, but in some situations only some optimizer states are quintized and then more memory is used.
 
 Let's get a feel for the numbers and use for example use a 3B-parameter model, like `t5-3b`. Note that since a Gigabyte correpsonds to a billion bytes we can simply multiply the parameters (in billions) with the number of necessary bytes per parameter to get Gigabytes of GPU memory usage:
 
@@ -609,7 +609,7 @@ for step, batch in enumerate(dataloader, start=1):
         optimizer.zero_grad()
 ```
 
-First we wrap the dataset in a [`DataLoader`](https://pytorch.org/docs/stable/data.html#torch.utils.data.DataLoader). Then we can enable gradient checkpointing by calling the model's [`~PreTrainedModel.gradient_checkpointing_enable`] method. When we initialize the [`Accelerator`](https://huggingface.co/docs/accelerate/accelerator.html#accelerate.Accelerator) we can specifiy if we want to use mixed precision training and it will take care of it for us in the [`prepare`] call. During the [`prepare`](https://huggingface.co/docs/accelerate/accelerator.html#accelerate.Accelerator.prepare) call the dataloader will also be distributed across workers should we use multiple GPUs. We use the same 8-bit optimizer from the earlier experiments.
+First we wrap the dataset in a [`DataLoader`](https://pytorch.org/docs/stable/data.html#torch.utils.data.DataLoader). Then we can enable gradient checkpointing by calling the model's [`~PreTrainedModel.gradient_checkpointing_enable`] method. When we initialize the [`Accelerator`](https://huggingface.co/docs/accelerate/package_reference/accelerator#accelerate.Accelerator) we can specifiy if we want to use mixed precision training and it will take care of it for us in the [`prepare`] call. During the [`prepare`](https://huggingface.co/docs/accelerate/package_reference/accelerator#accelerate.Accelerator.prepare) call the dataloader will also be distributed across workers should we use multiple GPUs. We use the same 8-bit optimizer from the earlier experiments.
 
 Finally, we can write the main training loop. Note that the `backward` call is handled by 🤗 Accelerate. We can also see how gradient accumulation works: we normalize the loss so we get the average at the end of accumulation and once we have enough steps we run the optimization. Now the question is: does this use the same amount of memory as the previous steps? Let's check:
 
@@ -719,14 +719,17 @@ For some applications, such as pretraining large language models, applying all t
 Another use case for training on many GPUs is if the model does not fit on a single GPU with all the mentioned tricks. There are still more methods we can apply although life starts to get a bit more complicated. This usually involves some form of pipeline or tensor parallelism where the model itself is distributed across several GPUs. One can also make use of DeepSpeed which implements some of these parallelism strategies along with some more optimization to reduce the memory footprint such as partitioning the optimizer states. You can read more about this in the ["Multi-GPU training" section](perf_train_gpu_many).
 
 ## Inference with torchdynamo
+
 TorchDynamo is a new tracer that uses Python’s frame evaluation API to automatically create FX traces from existing PyTorch programs. After capturing the FX graph, different backends can be deployed to lower the graph to an optimized engine. One solution is using the [TensorRT](https://developer.nvidia.com/tensorrt) or NVFuser as backend. You can choose one option below for performance boost.
+
 ```
 TrainingArguments(torchdynamo="eager")      #enable eager model GPU. No performance boost
 TrainingArguments(torchdynamo="nvfuser")    #enable nvfuser
 TrainingArguments(torchdynamo="fx2trt")     #enable tensorRT fp32
 TrainingArguments(torchdynamo="fx2trt-f16") #enable tensorRT fp16
 ```
+
 This feature involves 3 different libraries. To install them, please follow the instructions below:  
 - [Torchdynamo installation](https://github.com/pytorch/torchdynamo#requirements-and-setup)  
 - [Functorch installation](https://github.com/pytorch/functorch#install)  
-- [Torch-TensorRT(FX) installation](https://github.com/pytorch/TensorRT/blob/master/docsrc/tutorials/getting_started_with_fx_path.rst#installation)  
+- [Torch-TensorRT(FX) installation](https://github.com/pytorch/TensorRT/blob/master/docsrc/tutorials/getting_started_with_fx_path.rst#installation)  

docs/source/en/perplexity.mdx

@@ -101,22 +101,32 @@ from tqdm import tqdm
 
 max_length = model.config.n_positions
 stride = 512
+seq_len = encodings.input_ids.size(1)
 
 nlls = []
-for i in tqdm(range(0, encodings.input_ids.size(1), stride)):
-    begin_loc = max(i + stride - max_length, 0)
-    end_loc = min(i + stride, encodings.input_ids.size(1))
-    trg_len = end_loc - i  # may be different from stride on last loop
+prev_end_loc = 0
+for begin_loc in tqdm(range(0, seq_len, stride)):
+    end_loc = min(begin_loc + max_length, seq_len)
+    trg_len = end_loc - prev_end_loc  # may be different from stride on last loop
     input_ids = encodings.input_ids[:, begin_loc:end_loc].to(device)
     target_ids = input_ids.clone()
     target_ids[:, :-trg_len] = -100
 
     with torch.no_grad():
         outputs = model(input_ids, labels=target_ids)
-        neg_log_likelihood = outputs[0] * trg_len
+
+        # loss is calculated using CrossEntropyLoss which averages over input tokens.
+        # Multiply it with trg_len to get the summation instead of average.
+        # We will take average over all the tokens to get the true average
+        # in the last step of this example.
+        neg_log_likelihood = outputs.loss * trg_len
 
     nlls.append(neg_log_likelihood)
 
+    prev_end_loc = end_loc
+    if end_loc == seq_len:
+        break
+
 ppl = torch.exp(torch.stack(nlls).sum() / end_loc)
 ```
 
@@ -126,5 +136,5 @@ and the better the reported perplexity will typically be.
 
 When we run the above with `stride = 1024`, i.e. no overlap, the resulting PPL is `19.64`, which is about the same
 as the `19.93` reported in the GPT-2 paper. By using `stride = 512` and thereby employing our striding window
-strategy, this jumps down to `16.53`. This is not only a more favorable score, but is calculated in a way that is
+strategy, this jumps down to `16.44`. This is not only a more favorable score, but is calculated in a way that is
 closer to the true autoregressive decomposition of a sequence likelihood.

docs/source/en/philosophy.mdx

@@ -14,29 +14,28 @@ specific language governing permissions and limitations under the License.
 
 🤗 Transformers is an opinionated library built for:
 
-- NLP researchers and educators seeking to use/study/extend large-scale transformers models
-- hands-on practitioners who want to fine-tune those models and/or serve them in production
-- engineers who just want to download a pretrained model and use it to solve a given NLP task.
+- machine learning researchers and educators seeking to use, study or extend large-scale Transformers models.
+- hands-on practitioners who want to fine-tune those models or serve them in production, or both.
+- engineers who just want to download a pretrained model and use it to solve a given machine learning task.
 
 The library was designed with two strong goals in mind:
 
-- Be as easy and fast to use as possible:
+1. Be as easy and fast to use as possible:
 
   - We strongly limited the number of user-facing abstractions to learn, in fact, there are almost no abstractions,
     just three standard classes required to use each model: [configuration](main_classes/configuration),
-    [models](main_classes/model) and [tokenizer](main_classes/tokenizer).
+    [models](main_classes/model), and a preprocessing class ([tokenizer](main_classes/tokenizer) for NLP, [feature extractor](main_classes/feature_extractor) for vision and audio, and [processor](main_classes/processors) for multimodal inputs).
   - All of these classes can be initialized in a simple and unified way from pretrained instances by using a common
-    `from_pretrained()` instantiation method which will take care of downloading (if needed), caching and
-    loading the related class instance and associated data (configurations' hyper-parameters, tokenizers' vocabulary,
+    `from_pretrained()` method which downloads (if needed), caches and
+    loads the related class instance and associated data (configurations' hyperparameters, tokenizers' vocabulary,
     and models' weights) from a pretrained checkpoint provided on [Hugging Face Hub](https://huggingface.co/models) or your own saved checkpoint.
   - On top of those three base classes, the library provides two APIs: [`pipeline`] for quickly
-    using a model (plus its associated tokenizer and configuration) on a given task and
-    [`Trainer`]/`Keras.fit` to quickly train or fine-tune a given model.
+    using a model for inference on a given task and [`Trainer`] to quickly train or fine-tune a PyTorch model (all TensorFlow models are compatible with `Keras.fit`).
   - As a consequence, this library is NOT a modular toolbox of building blocks for neural nets. If you want to
-    extend/build-upon the library, just use regular Python/PyTorch/TensorFlow/Keras modules and inherit from the base
-    classes of the library to reuse functionalities like model loading/saving.
+    extend or build upon the library, just use regular Python, PyTorch, TensorFlow, Keras modules and inherit from the base
+    classes of the library to reuse functionalities like model loading and saving. If you'd like to learn more about our coding philosophy for models, check out our [Repeat Yourself](https://huggingface.co/blog/transformers-design-philosophy) blog post.
 
-- Provide state-of-the-art models with performances as close as possible to the original models:
+2. Provide state-of-the-art models with performances as close as possible to the original models:
 
   - We provide at least one example for each architecture which reproduces a result provided by the official authors
     of said architecture.
@@ -48,33 +47,29 @@ A few other goals:
 - Expose the models' internals as consistently as possible:
 
   - We give access, using a single API, to the full hidden-states and attention weights.
-  - Tokenizer and base model's API are standardized to easily switch between models.
+  - The preprocessing classes and base model APIs are standardized to easily switch between models.
 
-- Incorporate a subjective selection of promising tools for fine-tuning/investigating these models:
+- Incorporate a subjective selection of promising tools for fine-tuning and investigating these models:
 
-  - A simple/consistent way to add new tokens to the vocabulary and embeddings for fine-tuning.
-  - Simple ways to mask and prune transformer heads.
+  - A simple and consistent way to add new tokens to the vocabulary and embeddings for fine-tuning.
+  - Simple ways to mask and prune Transformer heads.
 
-- Switch easily between PyTorch and TensorFlow 2.0, allowing training using one framework and inference using another.
+- Easily switch between PyTorch, TensorFlow 2.0 and Flax, allowing training with one framework and inference with another.
 
 ## Main concepts
 
 The library is built around three types of classes for each model:
 
-- **Model classes** such as [`BertModel`], which are 30+ PyTorch models ([torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module)) or Keras models ([tf.keras.Model](https://www.tensorflow.org/api_docs/python/tf/keras/Model)) that work with the pretrained weights provided in the
-  library.
-- **Configuration classes** such as [`BertConfig`], which store all the parameters required to build
-  a model. You don't always need to instantiate these yourself. In particular, if you are using a pretrained model
-  without any modification, creating the model will automatically take care of instantiating the configuration (which
-  is part of the model).
-- **Tokenizer classes** such as [`BertTokenizer`], which store the vocabulary for each model and
-  provide methods for encoding/decoding strings in a list of token embeddings indices to be fed to a model.
+- **Model classes** can be PyTorch models ([torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module)), Keras models ([tf.keras.Model](https://www.tensorflow.org/api_docs/python/tf/keras/Model)) or JAX/Flax models ([flax.linen.Module](https://flax.readthedocs.io/en/latest/api_reference/flax.linen.html)) that work with the pretrained weights provided in the library.
+- **Configuration classes** store the hyperparameters required to build a model (such as the number of layers and hidden size). You don't always need to instantiate these yourself. In particular, if you are using a pretrained model without any modification, creating the model will automatically take care of instantiating the configuration (which is part of the model).
+- **Preprocessing classes** convert the raw data into a format accepted by the model. A [tokenizer](main_classes/tokenizer) stores the vocabulary for each model and provide methods for encoding and decoding strings in a list of token embedding indices to be fed to a model. [Feature extractors](main_classes/feature_extractor) preprocess audio or vision inputs, and a [processor](main_classes/processors) handles multimodal inputs.
 
-All these classes can be instantiated from pretrained instances and saved locally using two methods:
+All these classes can be instantiated from pretrained instances, saved locally, and shared on the Hub with three methods:
 
-- `from_pretrained()` lets you instantiate a model/configuration/tokenizer from a pretrained version either
+- `from_pretrained()` lets you instantiate a model, configuration, and preprocessing class from a pretrained version either
   provided by the library itself (the supported models can be found on the [Model Hub](https://huggingface.co/models)) or
-  stored locally (or on a server) by the user,
-- `save_pretrained()` lets you save a model/configuration/tokenizer locally so that it can be reloaded using
+  stored locally (or on a server) by the user.
+- `save_pretrained()` lets you save a model, configuration, and preprocessing class locally so that it can be reloaded using
   `from_pretrained()`.
+- `push_to_hub()` lets you share a model, configuration, and a preprocessing class to the Hub, so it is easily accessible to everyone.
 

docs/source/en/pipeline_tutorial.mdx

@@ -12,21 +12,21 @@ specific language governing permissions and limitations under the License.
 
 # Pipelines for inference
 
-The [`pipeline`] makes it simple to use any model from the [Model Hub](https://huggingface.co/models) for inference on a variety of tasks such as text generation, image segmentation and audio classification. Even if you don't have experience with a specific modality or understand the code powering the models, you can still use them with the [`pipeline`]! This tutorial will teach you to:
+The [`pipeline`] makes it simple to use any model from the [Hub](https://huggingface.co/models) for inference on any language, computer vision, speech, and multimodal tasks. Even if you don't have experience with a specific modality or aren't familiar with the underlying code behind the models, you can still use them for inference with the [`pipeline`]! This tutorial will teach you to:
 
 * Use a [`pipeline`] for inference.
 * Use a specific tokenizer or model.
-* Use a [`pipeline`] for audio and vision tasks.
+* Use a [`pipeline`] for audio, vision, and multimodal tasks.
 
 <Tip>
 
-Take a look at the [`pipeline`] documentation for a complete list of supported tasks.
+Take a look at the [`pipeline`] documentation for a complete list of supported tasks and available parameters.
 
 </Tip>
 
 ## Pipeline usage
 
-While each task has an associated [`pipeline`], it is simpler to use the general [`pipeline`] abstraction which contains all the specific task pipelines. The [`pipeline`] automatically loads a default model and tokenizer capable of inference for your task. 
+While each task has an associated [`pipeline`], it is simpler to use the general [`pipeline`] abstraction which contains all the task-specific pipelines. The [`pipeline`] automatically loads a default model and a preprocessing class capable of inference for your task.
 
 1. Start by creating a [`pipeline`] and specify an inference task:
 
@@ -67,7 +67,7 @@ Any additional parameters for your task can also be included in the [`pipeline`]
 
 ### Choose a model and tokenizer
 
-The [`pipeline`] accepts any model from the [Model Hub](https://huggingface.co/models). There are tags on the Model Hub that allow you to filter for a model you'd like to use for your task. Once you've picked an appropriate model, load it with the corresponding `AutoModelFor` and [`AutoTokenizer'] class. For example, load the [`AutoModelForCausalLM`] class for a causal language modeling task:
+The [`pipeline`] accepts any model from the [Hub](https://huggingface.co/models). There are tags on the Hub that allow you to filter for a model you'd like to use for your task. Once you've picked an appropriate model, load it with the corresponding `AutoModelFor` and [`AutoTokenizer`] class. For example, load the [`AutoModelForCausalLM`] class for a causal language modeling task:
 
 ```py
 >>> from transformers import AutoTokenizer, AutoModelForCausalLM
@@ -95,7 +95,7 @@ Pass your input text to the [`pipeline`] to generate some text:
 
 ## Audio pipeline
 
-The flexibility of the [`pipeline`] means it can also be extended to audio tasks.
+The [`pipeline`] also supports audio tasks like audio classification and automatic speech recognition.
 
 For example, let's classify the emotion in this audio clip:
 
@@ -129,9 +129,9 @@ Pass the audio file to the [`pipeline`]:
 
 ## Vision pipeline
 
-Finally, using a [`pipeline`] for vision tasks is practically identical.
+Using a [`pipeline`] for vision tasks is practically identical.
 
-Specify your vision task and pass your image to the classifier. The imaage can be a link or a local path to the image. For example, what species of cat is shown below?
+Specify your task and pass your image to the classifier. The image can be a link or a local path to the image. For example, what species of cat is shown below?
 
 ![pipeline-cat-chonk](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/pipeline-cat-chonk.jpeg)
 
@@ -146,3 +146,26 @@ Specify your vision task and pass your image to the classifier. The imaage can b
 >>> preds
 [{'score': 0.4335, 'label': 'lynx, catamount'}, {'score': 0.0348, 'label': 'cougar, puma, catamount, mountain lion, painter, panther, Felis concolor'}, {'score': 0.0324, 'label': 'snow leopard, ounce, Panthera uncia'}, {'score': 0.0239, 'label': 'Egyptian cat'}, {'score': 0.0229, 'label': 'tiger cat'}]
 ```
+
+## Multimodal pipeline
+
+The [`pipeline`] supports more than one modality. For example, a visual question answering (VQA) task combines text and image. Feel free to use any image link you like and a question you want to ask about the image. The image can be a URL or a local path to the image.
+
+For example, if you use the same image from the vision pipeline above:
+
+```py
+>>> image = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/pipeline-cat-chonk.jpeg"
+>>> question = "Where is the cat?"
+```
+
+Create a pipeline for `vqa` and pass it the image and question:
+
+```py
+>>> from transformers import pipeline
+
+>>> vqa = pipeline(task="vqa")
+>>> preds = vqa(image=image, question=question)
+>>> preds = [{"score": round(pred["score"], 4), "answer": pred["answer"]} for pred in preds]
+>>> preds
+[{'score': 0.911, 'answer': 'snow'}, {'score': 0.8786, 'answer': 'in snow'}, {'score': 0.6714, 'answer': 'outside'}, {'score': 0.0293, 'answer': 'on ground'}, {'score': 0.0272, 'answer': 'ground'}]
+```

docs/source/en/pr_checks.mdx

@@ -65,13 +65,9 @@ Just in case anything slipped through the cracks, the full test suite is also ru
 
 ## Documentation build
 
-The job `ci/circleci: build_doc` runs a build of the documentation just to make sure everything will be okay once your PR is merged. If that steps fails, you can inspect it locally by going into the `docs` folder of the Transformers repo and then typing
+The `build_pr_documentation` job builds and generates a preview of the documentation to make sure everything looks okay once your PR is merged. A bot will add a link to preview the documentation in your PR. Any changes you make to the PR are automatically updated in the preview. If the documentation fails to build, click on **Details** next to the failed job to see where things went wrong. Often, the error is as simple as a missing file in the `toctree`.
 
-```bash
-make html
-```
-
-Sphinx is not known for its helpful error messages, so you might have to try a few things to really find the source of the error.
+If you're interested in building or previewing the documentation locally, take a look at the [`README.md`](https://github.com/huggingface/transformers/tree/main/docs) in the docs folder.
 
 ## Code and documentation style
 

docs/source/en/preprocessing.mdx

@@ -14,17 +14,29 @@ specific language governing permissions and limitations under the License.
 
 [[open-in-colab]]
 
-Before you can use your data in a model, the data needs to be processed into an acceptable format for the model. A model does not understand raw text, images or audio. These inputs need to be converted into numbers and assembled into tensors. In this tutorial, you will:
+Before you can train a model on a dataset, it needs to be preprocessed into the expected model input format. Whether your data is text, images, or audio, they need to be converted and assembled into batches of tensors. 🤗 Transformers provides a set of preprocessing classes to help prepare your data for the model. In this tutorial, you'll learn that for:
 
-* Preprocess textual data with a tokenizer.
-* Preprocess image or audio data with a feature extractor.
-* Preprocess data for a multimodal task with a processor.
+* Text, use a [Tokenizer](./main_classes/tokenizer) to convert text into a sequence of tokens, create a numerical representation of the tokens, and assemble them into tensors.
+* Computer vision and speech, use a [Feature extractor](./main_classes/feature_extractor) to extract sequential features from audio waveforms and images and convert them into tensors.
+* Multimodal inputs, use a [Processor](./main_classes/processors) to combine a tokenizer and a feature extractor.
 
-## NLP
+<Tip>
+
+`AutoProcessor` **always** works and automatically chooses the correct class for the model you're using, whether you're using a tokenizer, feature extractor or processor.
+
+</Tip>
+
+Before you begin, install 🤗 Datasets so you can load some datasets to experiment with:
+
+```bash
+pip install datasets
+```
+
+## Natural Language Processing
 
 <Youtube id="Yffk5aydLzg"/>
 
-The main tool for processing textual data is a [tokenizer](main_classes/tokenizer). A tokenizer starts by splitting text into *tokens* according to a set of rules. The tokens are converted into numbers, which are used to build tensors as input to a model. Any additional inputs required by a model are also added by the tokenizer.
+The main tool for preprocessing textual data is a [tokenizer](main_classes/tokenizer). A tokenizer splits text into *tokens* according to a set of rules. The tokens are converted into numbers and then tensors, which become the model inputs. Any additional inputs required by the model are added by the tokenizer.
 
 <Tip>
 
@@ -32,11 +44,7 @@ If you plan on using a pretrained model, it's important to use the associated pr
 
 </Tip>
 
-Get started quickly by loading a pretrained tokenizer with the [`AutoTokenizer`] class. This downloads the *vocab* used when a model is pretrained.
-
-### Tokenize
-
-Load a pretrained tokenizer with [`AutoTokenizer.from_pretrained`]:
+Get started by loading a pretrained tokenizer with the [`AutoTokenizer.from_pretrained`] method. This downloads the *vocab* a model was pretrained with:
 
 ```py
 >>> from transformers import AutoTokenizer
@@ -44,7 +52,7 @@ Load a pretrained tokenizer with [`AutoTokenizer.from_pretrained`]:
 >>> tokenizer = AutoTokenizer.from_pretrained("bert-base-cased")
 ```
 
-Then pass your sentence to the tokenizer:
+Then pass your text to the tokenizer:
 
 ```py
 >>> encoded_input = tokenizer("Do not meddle in the affairs of wizards, for they are subtle and quick to anger.")
@@ -54,13 +62,13 @@ Then pass your sentence to the tokenizer:
  'attention_mask': [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]}
 ```
 
-The tokenizer returns a dictionary with three important itmes:
+The tokenizer returns a dictionary with three important items:
 
 * [input_ids](glossary#input-ids) are the indices corresponding to each token in the sentence.
 * [attention_mask](glossary#attention-mask) indicates whether a token should be attended to or not.
 * [token_type_ids](glossary#token-type-ids) identifies which sequence a token belongs to when there is more than one sequence.
 
-You can decode the `input_ids` to return the original input:
+Return your input by decoding the `input_ids`:
 
 ```py
 >>> tokenizer.decode(encoded_input["input_ids"])
@@ -68,9 +76,9 @@ You can decode the `input_ids` to return the original input:
 ```
 
 As you can see, the tokenizer added two special tokens - `CLS` and `SEP` (classifier and separator) - to the sentence. Not all models need
-special tokens, but if they do, the tokenizer will automatically add them for you.
+special tokens, but if they do, the tokenizer automatically adds them for you.
 
-If there are several sentences you want to process, pass the sentences as a list to the tokenizer:
+If there are several sentences you want to preprocess, pass them as a list to the tokenizer:
 
 ```py
 >>> batch_sentences = [
@@ -93,7 +101,7 @@ If there are several sentences you want to process, pass the sentences as a list
 
 ### Pad
 
-This brings us to an important topic. When you process a batch of sentences, they aren't always the same length. This is a problem because tensors, the input to the model, need to have a uniform shape. Padding is a strategy for ensuring tensors are rectangular by adding a special *padding token* to sentences with fewer tokens.
+Sentences aren't always the same length which can be an issue because tensors, the model inputs, need to have a uniform shape. Padding is a strategy for ensuring tensors are rectangular by adding a special *padding token* to shorter sentences.
 
 Set the `padding` parameter to `True` to pad the shorter sequences in the batch to match the longest sequence:
 
@@ -116,11 +124,11 @@ Set the `padding` parameter to `True` to pad the shorter sequences in the batch
                     [1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0]]}
 ```
 
-Notice the tokenizer padded the first and third sentences with a `0` because they are shorter!
+The first and third sentences are now padded with `0`'s because they are shorter.
 
 ### Truncation
 
-On the other end of the spectrum, sometimes a sequence may be too long for a model to handle. In this case, you will need to truncate the sequence to a shorter length.
+On the other end of the spectrum, sometimes a sequence may be too long for a model to handle. In this case, you'll need to truncate the sequence to a shorter length.
 
 Set the `truncation` parameter to `True` to truncate a sequence to the maximum length accepted by the model:
 
@@ -143,9 +151,15 @@ Set the `truncation` parameter to `True` to truncate a sequence to the maximum l
                     [1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0]]}
 ```
 
+<Tip>
+
+Check out the [Padding and truncation](./pad_truncation) concept guide to learn more different padding and truncation arguments.
+
+</Tip>
+
 ### Build tensors
 
-Finally, you want the tokenizer to return the actual tensors that are fed to the model.
+Finally, you want the tokenizer to return the actual tensors that get fed to the model.
 
 Set the `return_tensors` parameter to either `pt` for PyTorch, or `tf` for TensorFlow:
 
@@ -199,13 +213,9 @@ array([[1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0],
 
 ## Audio
 
-Audio inputs are preprocessed differently than textual inputs, but the end goal remains the same: create numerical sequences the model can understand. A [feature extractor](main_classes/feature_extractor) is designed for the express purpose of extracting features from raw image or audio data and converting them into tensors. Before you begin, install 🤗 Datasets to load an audio dataset to experiment with:
+For audio tasks, you'll need a [feature extractor](main_classes/feature_extractor) to prepare your dataset for the model. The feature extractor is designed to extract features from raw audio data, and convert them into tensors.
 
-```bash
-pip install datasets
-```
-
-Load the [MInDS-14](https://huggingface.co/datasets/PolyAI/minds14) dataset (see the 🤗 [Datasets tutorial](https://huggingface.co/docs/datasets/load_hub.html) for more details on how to load a dataset):
+Load the [MInDS-14](https://huggingface.co/datasets/PolyAI/minds14) dataset (see the 🤗 [Datasets tutorial](https://huggingface.co/docs/datasets/load_hub.html) for more details on how to load a dataset) to see how you can use a feature extractor with audio datasets:
 
 ```py
 >>> from datasets import load_dataset, Audio
@@ -213,7 +223,7 @@ Load the [MInDS-14](https://huggingface.co/datasets/PolyAI/minds14) dataset (see
 >>> dataset = load_dataset("PolyAI/minds14", name="en-US", split="train")
 ```
 
-Access the first element of the `audio` column to take a look at the input. Calling the `audio` column will automatically load and resample the audio file:
+Access the first element of the `audio` column to take a look at the input. Calling the `audio` column automatically loads and resamples the audio file:
 
 ```py
 >>> dataset[0]["audio"]
@@ -229,20 +239,7 @@ This returns three items:
 * `path` points to the location of the audio file.
 * `sampling_rate` refers to how many data points in the speech signal are measured per second.
 
-### Resample
-
-For this tutorial, you will use the [Wav2Vec2](https://huggingface.co/facebook/wav2vec2-base) model. As you can see from the model card, the Wav2Vec2 model is pretrained on 16kHz sampled speech audio. It is important your audio data's sampling rate matches the sampling rate of the dataset used to pretrain the model. If your data's sampling rate isn't the same, then you need to resample your audio data. 
-
-For example, the [MInDS-14](https://huggingface.co/datasets/PolyAI/minds14) dataset has a sampling rate of 8000kHz. In order to use the Wav2Vec2 model with this dataset, upsample the sampling rate to 16kHz:
-
-```py
->>> dataset = load_dataset("PolyAI/minds14", name="en-US", split="train")
->>> dataset[0]["audio"]
-{'array': array([ 0.        ,  0.00024414, -0.00024414, ..., -0.00024414,
-         0.        ,  0.        ], dtype=float32),
- 'path': '/root/.cache/huggingface/datasets/downloads/extracted/f14948e0e84be638dd7943ac36518a4cf3324e8b7aa331c5ab11541518e9368c/en-US~JOINT_ACCOUNT/602ba55abb1e6d0fbce92065.wav',
- 'sampling_rate': 8000}
-```
+For this tutorial, you'll use the [Wav2Vec2](https://huggingface.co/facebook/wav2vec2-base) model. Take a look at the model card, and you'll learn Wav2Vec2 is pretrained on 16kHz sampled speech audio. It is important your audio data's sampling rate matches the sampling rate of the dataset used to pretrain the model. If your data's sampling rate isn't the same, then you need to resample your data. 
 
 1. Use 🤗 Datasets' [`~datasets.Dataset.cast_column`] method to upsample the sampling rate to 16kHz:
 
@@ -250,7 +247,7 @@ For example, the [MInDS-14](https://huggingface.co/datasets/PolyAI/minds14) data
 >>> dataset = dataset.cast_column("audio", Audio(sampling_rate=16_000))
 ```
 
-2. Load the audio file:
+2. Call the `audio` column again to resample the audio file:
 
 ```py
 >>> dataset[0]["audio"]
@@ -260,11 +257,7 @@ For example, the [MInDS-14](https://huggingface.co/datasets/PolyAI/minds14) data
  'sampling_rate': 16000}
 ```
 
-As you can see, the `sampling_rate` is now 16kHz!
-
-### Feature extractor
-
-The next step is to load a feature extractor to normalize and pad the input. When padding textual data, a `0` is added for shorter sequences. The same idea applies to audio data, and the audio feature extractor will add a `0` - interpreted as silence - to `array`.
+Next, load a feature extractor to normalize and pad the input. When padding textual data, a `0` is added for shorter sequences. The same idea applies to audio data. The feature extractor adds a `0` - interpreted as silence - to `array`.
 
 Load the feature extractor with [`AutoFeatureExtractor.from_pretrained`]:
 
@@ -283,8 +276,6 @@ Pass the audio `array` to the feature extractor. We also recommend adding the `s
         5.6335266e-04,  4.6588284e-06, -1.7142107e-04], dtype=float32)]}
 ```
 
-### Pad and truncate
-
 Just like the tokenizer, you can apply padding or truncation to handle variable sequences in a batch. Take a look at the sequence length of these two audio samples:
 
 ```py
@@ -295,7 +286,7 @@ Just like the tokenizer, you can apply padding or truncation to handle variable
 (106496,)
 ```
 
-As you can see, the first sample has a longer sequence than the second sample. Let's create a function that will preprocess the dataset. Specify a maximum sample length, and the feature extractor will either pad or truncate the sequences to match it:
+Create a function to preprocess the dataset so the audio samples are the same lengths. Specify a maximum sample length, and the feature extractor will either pad or truncate the sequences to match it:
 
 ```py
 >>> def preprocess_function(examples):
@@ -310,13 +301,13 @@ As you can see, the first sample has a longer sequence than the second sample. L
 ...     return inputs
 ```
 
-Apply the function to the the first few examples in the dataset:
+Apply the `preprocess_function` to the the first few examples in the dataset:
 
 ```py
 >>> processed_dataset = preprocess_function(dataset[:5])
 ```
 
-Now take another look at the processed sample lengths:
+The sample lengths are now the same and match the specified maximum length. You can pass your processed dataset to the model now!
 
 ```py
 >>> processed_dataset["input_values"][0].shape
@@ -326,13 +317,17 @@ Now take another look at the processed sample lengths:
 (100000,)
 ```
 
-The lengths of the first two samples now match the maximum length you specified.
+## Computer vision
 
-## Vision
+For computer vision tasks, you'll need a [feature extractor](main_classes/feature_extractor) to prepare your dataset for the model. The feature extractor is designed to extract features from images, and convert them into tensors.
 
-A feature extractor is also used to process images for vision tasks. Once again, the goal is to convert the raw image into a batch of tensors as input.
+Load the [food101](https://huggingface.co/datasets/food101) dataset (see the 🤗 [Datasets tutorial](https://huggingface.co/docs/datasets/load_hub.html) for more details on how to load a dataset) to see how you can use a feature extractor with computer vision datasets: 
 
-Let's load the [food101](https://huggingface.co/datasets/food101) dataset for this tutorial. Use 🤗 Datasets `split` parameter to only load a small sample from the training split since the dataset is quite large:
+<Tip>
+
+Use 🤗 Datasets `split` parameter to only load a small sample from the training split since the dataset is quite large!
+
+</Tip>
 
 ```py
 >>> from datasets import load_dataset
@@ -346,9 +341,9 @@ Next, take a look at the image with 🤗 Datasets [`Image`](https://huggingface.
 >>> dataset[0]["image"]
 ```
 
-![vision-preprocess-tutorial.png](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/vision-preprocess-tutorial.png)
-
-### Feature extractor
+<div class="flex justify-center">
+    <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/vision-preprocess-tutorial.png"/>
+</div>
 
 Load the feature extractor with [`AutoFeatureExtractor.from_pretrained`]:
 
@@ -358,11 +353,9 @@ Load the feature extractor with [`AutoFeatureExtractor.from_pretrained`]:
 >>> feature_extractor = AutoFeatureExtractor.from_pretrained("google/vit-base-patch16-224")
 ```
 
-### Data augmentation
+For computer vision tasks, it is common to add some type of data augmentation to the images as a part of preprocessing. You can add augmentations with any library you'd like, but in this tutorial, you'll use torchvision's [`transforms`](https://pytorch.org/vision/stable/transforms.html) module. If you're interested in using another data augmentation library, learn how in the [Albumentations](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/image_classification_albumentations.ipynb) or [Kornia notebooks](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/image_classification_kornia.ipynb).
 
-For vision tasks, it is common to add some type of data augmentation to the images as a part of preprocessing. You can add augmentations with any library you'd like, but in this tutorial, you will use torchvision's [`transforms`](https://pytorch.org/vision/stable/transforms.html) module.
-
-1. Normalize the image and use [`Compose`](https://pytorch.org/vision/master/generated/torchvision.transforms.Compose.html) to chain some transforms - [`RandomResizedCrop`](https://pytorch.org/vision/main/generated/torchvision.transforms.RandomResizedCrop.html) and [`ColorJitter`](https://pytorch.org/vision/main/generated/torchvision.transforms.ColorJitter.html) - together:
+1. Normalize the image with the feature extractor and use [`Compose`](https://pytorch.org/vision/master/generated/torchvision.transforms.Compose.html) to chain some transforms - [`RandomResizedCrop`](https://pytorch.org/vision/main/generated/torchvision.transforms.RandomResizedCrop.html) and [`ColorJitter`](https://pytorch.org/vision/main/generated/torchvision.transforms.ColorJitter.html) - together:
 
 ```py
 >>> from torchvision.transforms import Compose, Normalize, RandomResizedCrop, ColorJitter, ToTensor
@@ -373,7 +366,7 @@ For vision tasks, it is common to add some type of data augmentation to the imag
 ... )
 ```
 
-2. The model accepts [`pixel_values`](model_doc/visionencoderdecoder#transformers.VisionEncoderDecoderModel.forward.pixel_values) as it's input. This value is generated by the feature extractor. Create a function that generates `pixel_values` from the transforms:
+2. The model accepts [`pixel_values`](model_doc/visionencoderdecoder#transformers.VisionEncoderDecoderModel.forward.pixel_values) as its input, which is generated by the feature extractor. Create a function that generates `pixel_values` from the transforms:
 
 ```py
 >>> def transforms(examples):
@@ -381,13 +374,13 @@ For vision tasks, it is common to add some type of data augmentation to the imag
 ...     return examples
 ```
 
-3. Then use 🤗 Datasets [`set_transform`](https://huggingface.co/docs/datasets/process.html#format-transform) to apply the transforms on-the-fly:
+3. Then use 🤗 Datasets [`set_transform`](https://huggingface.co/docs/datasets/process.html#format-transform) to apply the transforms on the fly:
 
 ```py
 >>> dataset.set_transform(transforms)
 ```
 
-4. Now when you access the image, you will notice the feature extractor has added the model input `pixel_values`:
+4. Now when you access the image, you'll notice the feature extractor has added `pixel_values`. You can pass your processed dataset to the model now!
 
 ```py
 >>> dataset[0]["image"]
@@ -418,7 +411,7 @@ For vision tasks, it is common to add some type of data augmentation to the imag
           [-0.1922, -0.1922, -0.1922,  ..., -0.2941, -0.2863, -0.3412]]])}
 ```
 
-Here is what the image looks like after you preprocess it. Just as you'd expect from the applied transforms, the image has been randomly cropped and it's color properties are different.
+Here is what the image looks like after the transforms are applied. The image has been randomly cropped and it's color properties are different.
 
 ```py
 >>> import numpy as np
@@ -428,16 +421,15 @@ Here is what the image looks like after you preprocess it. Just as you'd expect
 >>> plt.imshow(img.permute(1, 2, 0))
 ```
 
-![preprocessed_image](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/preprocessed_image.png)
+<div class="flex justify-center">
+    <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/preprocessed_image.png"/>
+</div>
 
 ## Multimodal
 
-For multimodal tasks. you will use a combination of everything you've learned so far and apply your skills to a automatic speech recognition (ASR) task. This means you will need a:
+For tasks involving multimodal inputs, you'll need a [processor](main_classes/processors) to prepare your dataset for the model. A processor couples a tokenizer and feature extractor.
 
-* Feature extractor to preprocess the audio data.
-* Tokenizer to process the text.
-
-Let's return to the [LJ Speech](https://huggingface.co/datasets/lj_speech) dataset:
+Load the [LJ Speech](https://huggingface.co/datasets/lj_speech) dataset (see the 🤗 [Datasets tutorial](https://huggingface.co/docs/datasets/load_hub.html) for more details on how to load a dataset) to see how you can use a processor for automatic speech recognition (ASR):
 
 ```py
 >>> from datasets import load_dataset
@@ -445,7 +437,7 @@ Let's return to the [LJ Speech](https://huggingface.co/datasets/lj_speech) datas
 >>> lj_speech = load_dataset("lj_speech", split="train")
 ```
 
-Since you are mainly interested in the `audio` and `text` column, remove the other columns:
+For ASR, you're mainly focused on `audio` and `text` so you can remove the other columns:
 
 ```py
 >>> lj_speech = lj_speech.map(remove_columns=["file", "id", "normalized_text"])
@@ -464,15 +456,13 @@ Now take a look at the `audio` and `text` columns:
 'Printing, in the only sense with which we are at present concerned, differs from most if not from all the arts and crafts represented in the Exhibition'
 ```
 
-Remember from the earlier section on processing audio data, you should always [resample](preprocessing#audio) your audio data's sampling rate to match the sampling rate of the dataset used to pretrain a model:
+Remember you should always [resample](preprocessing#audio) your audio dataset's sampling rate to match the sampling rate of the dataset used to pretrain a model!
 
 ```py
 >>> lj_speech = lj_speech.cast_column("audio", Audio(sampling_rate=16_000))
 ```
 
-### Processor
-
-A processor combines a feature extractor and tokenizer. Load a processor with [`AutoProcessor.from_pretrained]:
+Load a processor with [`AutoProcessor.from_pretrained`]:
 
 ```py
 >>> from transformers import AutoProcessor
@@ -480,7 +470,7 @@ A processor combines a feature extractor and tokenizer. Load a processor with [`
 >>> processor = AutoProcessor.from_pretrained("facebook/wav2vec2-base-960h")
 ```
 
-1. Create a function to process the audio data to `input_values`, and tokenizes the text to `labels`. These are your inputs to the model:
+1. Create a function to process the audio data contained in `array` to `input_values`, and tokenize `text` to `labels`. These are the inputs to the model:
 
 ```py
 >>> def prepare_dataset(example):
@@ -497,6 +487,4 @@ A processor combines a feature extractor and tokenizer. Load a processor with [`
 >>> prepare_dataset(lj_speech[0])
 ```
 
-Notice the processor has added `input_values` and `labels`. The sampling rate has also been correctly downsampled to 16kHz.
-
-Awesome, you should now be able to preprocess data for any modality and even combine different modalities! In the next tutorial, learn how to fine-tune a model on your newly preprocessed data.
+The processor has now added `input_values` and `labels`, and the sampling rate has also been correctly downsampled to 16kHz. You can pass your processed dataset to the model now!

docs/source/en/quicktour.mdx

@@ -14,53 +14,15 @@ specific language governing permissions and limitations under the License.
 
 [[open-in-colab]]
 
-Get up and running with 🤗 Transformers! Start using the [`pipeline`] for rapid inference, and quickly load a pretrained model and tokenizer with an [AutoClass](./model_doc/auto) to solve your text, vision or audio task.
+Get up and running with 🤗 Transformers! Whether you're a developer or an everyday user, this quick tour will help you get started and show you how to use the [`pipeline`] for inference, load a pretrained model and preprocessor with an [AutoClass](./model_doc/auto), and quickly train a model with PyTorch or TensorFlow. If you're a beginner, we recommend checking out our tutorials or [course](https://huggingface.co/course/chapter1/1) next for more in-depth explanations of the concepts introduced here.
 
-<Tip>
+Before you begin, make sure you have all the necessary libraries installed:
 
-All code examples presented in the documentation have a toggle on the top left for PyTorch and TensorFlow. If
-not, the code is expected to work for both backends without any change.
+```bash
+!pip install transformers datasets
+```
 
-</Tip>
-
-## Pipeline
-
-[`pipeline`] is the easiest way to use a pretrained model for a given task.
-
-<Youtube id="tiZFewofSLM"/>
-
-The [`pipeline`] supports many common tasks out-of-the-box:
-
-**Text**:
-* Sentiment analysis: classify the polarity of a given text.
-* Text generation (in English): generate text from a given input.
-* Name entity recognition (NER): label each word with the entity it represents (person, date, location, etc.).
-* Question answering: extract the answer from the context, given some context and a question.
-* Fill-mask: fill in the blank given a text with masked words.
-* Summarization: generate a summary of a long sequence of text or document.
-* Translation: translate text into another language.
-* Feature extraction: create a tensor representation of the text.
-
-**Image**:
-* Image classification: classify an image.
-* Image segmentation: classify every pixel in an image.
-* Object detection: detect objects within an image.
-
-**Audio**:
-* Audio classification: assign a label to a given segment of audio.
-* Automatic speech recognition (ASR): transcribe audio data into text.
-
-<Tip>
-
-For more details about the [`pipeline`] and associated tasks, refer to the documentation [here](./main_classes/pipelines).
-
-</Tip>
-
-### Pipeline usage
-
-In the following example, you will use the [`pipeline`] for sentiment analysis.
-
-Install the following dependencies if you haven't already:
+You'll also need to install your preferred machine learning framework:
 
 <frameworkcontent>
 <pt>
@@ -75,7 +37,29 @@ pip install tensorflow
 </tf>
 </frameworkcontent>
 
-Import [`pipeline`] and specify the task you want to complete:
+## Pipeline
+
+<Youtube id="tiZFewofSLM"/>
+
+The [`pipeline`] is the easiest way to use a pretrained model for inference. You can use the [`pipeline`] out-of-the-box for many tasks across different modalities. Take a look at the table below for some supported tasks:
+
+| **Task**                     | **Description**                                                                                              | **Modality**    | **Pipeline identifier**                       |
+|------------------------------|--------------------------------------------------------------------------------------------------------------|-----------------|-----------------------------------------------|
+| Text classification          | assign a label to a given sequence of text                                                                   | NLP             | pipeline(task="sentiment-analysis")           |
+| Text generation              | generate text that follows a given prompt                                                                    | NLP             | pipeline(task="text-generation")              |
+| Name entity recognition      | assign a label to each token in a sequence (people, organization, location, etc.)                            | NLP             | pipeline(task="ner")                          |
+| Question answering           | extract an answer from the text given some context and a question                                            | NLP             | pipeline(task="question-answering")           |
+| Fill-mask                    | predict the correct masked token in a sequence                                                               | NLP             | pipeline(task="fill-mask")                    |
+| Summarization                | generate a summary of a sequence of text or document                                                         | NLP             | pipeline(task="summarization")                |
+| Translation                  | translate text from one language into another                                                                | NLP             | pipeline(task="translation")                  |
+| Image classification         | assign a label to an image                                                                                   | Computer vision | pipeline(task="image-classification")         |
+| Image segmentation           | assign a label to each individual pixel of an image (supports semantic, panoptic, and instance segmentation) | Computer vision | pipeline(task="image-segmentation")           |
+| Object detection             | predict the bounding boxes and classes of objects in an image                                                | Computer vision | pipeline(task="object-detection")             |
+| Audio classification         | assign a label to an audio file                                                                              | Audio           | pipeline(task="audio-classification")         |
+| Automatic speech recognition | extract speech from an audio file into text                                                                  | Audio           | pipeline(task="automatic-speech-recognition") |
+| Visual question answering    | given an image and a question, correctly answer a question about the image                                   | Multimodal      | pipeline(task="vqa")                          |
+
+Start by creating an instance of [`pipeline`] and specifying a task you want to use it for. You can use the [`pipeline`] for any of the previously mentioned tasks, and for a complete list of supported tasks, take a look at the [pipeline API reference](./main_classes/pipelines). In this guide though, you'll use the [`pipeline`] for sentiment analysis as an example:
 
 ```py
 >>> from transformers import pipeline
@@ -83,14 +67,14 @@ Import [`pipeline`] and specify the task you want to complete:
 >>> classifier = pipeline("sentiment-analysis")
 ```
 
-The pipeline downloads and caches a default [pretrained model](https://huggingface.co/distilbert-base-uncased-finetuned-sst-2-english) and tokenizer for sentiment analysis. Now you can use the `classifier` on your target text:
+The [`pipeline`] downloads and caches a default [pretrained model](https://huggingface.co/distilbert-base-uncased-finetuned-sst-2-english) and tokenizer for sentiment analysis. Now you can use the `classifier` on your target text:
 
 ```py
 >>> classifier("We are very happy to show you the 🤗 Transformers library.")
 [{'label': 'POSITIVE', 'score': 0.9998}]
 ```
 
-For more than one sentence, pass a list of sentences to the [`pipeline`] which returns a list of dictionaries:
+If you have more than one input, pass your inputs as a list to the [`pipeline`] to return a list of dictionaries:
 
 ```py
 >>> results = classifier(["We are very happy to show you the 🤗 Transformers library.", "We hope you don't hate it."])
@@ -100,13 +84,7 @@ label: POSITIVE, with score: 0.9998
 label: NEGATIVE, with score: 0.5309
 ```
 
-The [`pipeline`] can also iterate over an entire dataset. Start by installing the [🤗 Datasets](https://huggingface.co/docs/datasets/) library:
-
-```bash
-pip install datasets 
-```
-
-Create a [`pipeline`] with the task you want to solve for and the model you want to use.
+The [`pipeline`] can also iterate over an entire dataset for any task you like. For this example, let's choose automatic speech recognition as our task:
 
 ```py
 >>> import torch
@@ -115,7 +93,7 @@ Create a [`pipeline`] with the task you want to solve for and the model you want
 >>> speech_recognizer = pipeline("automatic-speech-recognition", model="facebook/wav2vec2-base-960h")
 ```
 
-Next, load a dataset (see the 🤗 Datasets [Quick Start](https://huggingface.co/docs/datasets/quickstart.html) for more details) you'd like to iterate over. For example, let's load the [MInDS-14](https://huggingface.co/datasets/PolyAI/minds14) dataset:
+Load an audio dataset (see the 🤗 Datasets [Quick Start](https://huggingface.co/docs/datasets/quickstart#audio) for more details) you'd like to iterate over. For example, load the [MInDS-14](https://huggingface.co/datasets/PolyAI/minds14) dataset:
 
 ```py
 >>> from datasets import load_dataset, Audio
@@ -123,27 +101,27 @@ Next, load a dataset (see the 🤗 Datasets [Quick Start](https://huggingface.co
 >>> dataset = load_dataset("PolyAI/minds14", name="en-US", split="train")  # doctest: +IGNORE_RESULT
 ```
 
-We need to make sure that the sampling rate of the dataset matches the sampling 
-rate `facebook/wav2vec2-base-960h` was trained on.
+You need to make sure the sampling rate of the dataset matches the sampling 
+rate [`facebook/wav2vec2-base-960h`](https://huggingface.co/facebook/wav2vec2-base-960h) was trained on:
 
 ```py
 >>> dataset = dataset.cast_column("audio", Audio(sampling_rate=speech_recognizer.feature_extractor.sampling_rate))
 ```
 
-Audio files are automatically loaded and resampled when calling the `"audio"` column.
-Let's extract the raw waveform arrays of the first 4 samples and pass it as a list to the pipeline:
+The audio files are automatically loaded and resampled when calling the `"audio"` column.
+Extract the raw waveform arrays from the first 4 samples and pass it as a list to the pipeline:
 
 ```py
 >>> result = speech_recognizer(dataset[:4]["audio"])
 >>> print([d["text"] for d in result])
-['I WOULD LIKE TO SET UP A JOINT ACCOUNT WITH MY PARTNER HOW DO I PROCEED WITH DOING THAT', "FONDERING HOW I'D SET UP A JOIN TO HET WITH MY WIFE AND WHERE THE AP MIGHT BE", "I I'D LIKE TOY SET UP A JOINT ACCOUNT WITH MY PARTNER I'M NOT SEEING THE OPTION TO DO IT ON THE APSO I CALLED IN TO GET SOME HELP CAN I JUST DO IT OVER THE PHONE WITH YOU AND GIVE YOU THE INFORMATION OR SHOULD I DO IT IN THE AP AND I'M MISSING SOMETHING UQUETTE HAD PREFERRED TO JUST DO IT OVER THE PHONE OF POSSIBLE THINGS", 'HOW DO I TURN A JOIN A COUNT']
+['I WOULD LIKE TO SET UP A JOINT ACCOUNT WITH MY PARTNER HOW DO I PROCEED WITH DOING THAT', "FODING HOW I'D SET UP A JOIN TO HET WITH MY WIFE AND WHERE THE AP MIGHT BE", "I I'D LIKE TOY SET UP A JOINT ACCOUNT WITH MY PARTNER I'M NOT SEEING THE OPTION TO DO IT ON THE AP SO I CALLED IN TO GET SOME HELP CAN I JUST DO IT OVER THE PHONE WITH YOU AND GIVE YOU THE INFORMATION OR SHOULD I DO IT IN THE AP AND I'M MISSING SOMETHING UQUETTE HAD PREFERRED TO JUST DO IT OVER THE PHONE OF POSSIBLE THINGS", 'HOW DO I THURN A JOIN A COUNT']
 ```
 
-For a larger dataset where the inputs are big (like in speech or vision), you will want to pass along a generator instead of a list that loads all the inputs in memory. See the [pipeline documentation](./main_classes/pipelines) for more information.
+For larger datasets where the inputs are big (like in speech or vision), you'll want to pass a generator instead of a list to load all the inputs in memory. Take a look at the [pipeline API reference](./main_classes/pipelines) for more information.
 
 ### Use another model and tokenizer in the pipeline
 
-The [`pipeline`] can accommodate any model from the [Model Hub](https://huggingface.co/models), making it easy to adapt the [`pipeline`] for other use-cases. For example, if you'd like a model capable of handling French text, use the tags on the Model Hub to filter for an appropriate model. The top filtered result returns a multilingual [BERT model](https://huggingface.co/nlptown/bert-base-multilingual-uncased-sentiment) fine-tuned for sentiment analysis. Great, let's use this model!
+The [`pipeline`] can accommodate any model from the [Hub](https://huggingface.co/models), making it easy to adapt the [`pipeline`] for other use-cases. For example, if you'd like a model capable of handling French text, use the tags on the Hub to filter for an appropriate model. The top filtered result returns a multilingual [BERT model](https://huggingface.co/nlptown/bert-base-multilingual-uncased-sentiment) finetuned for sentiment analysis you can use for French text:
 
 ```py
 >>> model_name = "nlptown/bert-base-multilingual-uncased-sentiment"
@@ -151,7 +129,7 @@ The [`pipeline`] can accommodate any model from the [Model Hub](https://huggingf
 
 <frameworkcontent>
 <pt>
-Use the [`AutoModelForSequenceClassification`] and [`AutoTokenizer`] to load the pretrained model and it's associated tokenizer (more on an `AutoClass` below):
+Use [`AutoModelForSequenceClassification`] and [`AutoTokenizer`] to load the pretrained model and it's associated tokenizer (more on an `AutoClass` in the next section):
 
 ```py
 >>> from transformers import AutoTokenizer, AutoModelForSequenceClassification
@@ -161,7 +139,7 @@ Use the [`AutoModelForSequenceClassification`] and [`AutoTokenizer`] to load the
 ```
 </pt>
 <tf>
-Use the [`TFAutoModelForSequenceClassification`] and [`AutoTokenizer`] to load the pretrained model and it's associated tokenizer (more on an `TFAutoClass` below):
+Use [`TFAutoModelForSequenceClassification`] and [`AutoTokenizer`] to load the pretrained model and it's associated tokenizer (more on an `TFAutoClass` in the next section):
 
 ```py
 >>> from transformers import AutoTokenizer, TFAutoModelForSequenceClassification
@@ -172,7 +150,7 @@ Use the [`TFAutoModelForSequenceClassification`] and [`AutoTokenizer`] to load t
 </tf>
 </frameworkcontent>
 
-Then you can specify the model and tokenizer in the [`pipeline`], and apply the `classifier` on your target text:
+Specify the model and tokenizer in the [`pipeline`], and now you can apply the `classifier` on French text:
 
 ```py
 >>> classifier = pipeline("sentiment-analysis", model=model, tokenizer=tokenizer)
@@ -180,19 +158,19 @@ Then you can specify the model and tokenizer in the [`pipeline`], and apply the
 [{'label': '5 stars', 'score': 0.7273}]
 ```
 
-If you can't find a model for your use-case, you will need to fine-tune a pretrained model on your data. Take a look at our [fine-tuning tutorial](./training) to learn how. Finally, after you've fine-tuned your pretrained model, please consider sharing it (see tutorial [here](./model_sharing)) with the community on the Model Hub to democratize NLP for everyone! 🤗
+If you can't find a model for your use-case, you'll need to finetune a pretrained model on your data. Take a look at our [finetuning tutorial](./training) to learn how. Finally, after you've finetuned your pretrained model, please consider [sharing](./model_sharing) the model with the community on the Hub to democratize machine learning for everyone! 🤗
 
 ## AutoClass
 
 <Youtube id="AhChOFRegn4"/>
 
-Under the hood, the [`AutoModelForSequenceClassification`] and [`AutoTokenizer`] classes work together to power the [`pipeline`]. An [AutoClass](./model_doc/auto) is a shortcut that automatically retrieves the architecture of a pretrained model from it's name or path. You only need to select the appropriate `AutoClass` for your task and it's associated tokenizer with [`AutoTokenizer`]. 
+Under the hood, the [`AutoModelForSequenceClassification`] and [`AutoTokenizer`] classes work together to power the [`pipeline`] you used above. An [AutoClass](./model_doc/auto) is a shortcut that automatically retrieves the architecture of a pretrained model from it's name or path. You only need to select the appropriate `AutoClass` for your task and it's associated preprocessing class. 
 
-Let's return to our example and see how you can use the `AutoClass` to replicate the results of the [`pipeline`].
+Let's return to the example from the previous section and see how you can use the `AutoClass` to replicate the results of the [`pipeline`].
 
 ### AutoTokenizer
 
-A tokenizer is responsible for preprocessing text into a format that is understandable to the model. First, the tokenizer will split the text into words called *tokens*. There are multiple rules that govern the tokenization process, including how to split a word and at what level (learn more about tokenization [here](./tokenizer_summary)). The most important thing to remember though is you need to instantiate the tokenizer with the same model name to ensure you're using the same tokenization rules a model was pretrained with.
+A tokenizer is responsible for preprocessing text into an array of numbers as inputs to a model. There are multiple rules that govern the tokenization process, including how to split a word and at what level words should be split (learn more about tokenization in the [tokenizer summary](./tokenizer_summary)). The most important thing to remember is you need to instantiate a tokenizer with the same model name to ensure you're using the same tokenization rules a model was pretrained with.
 
 Load a tokenizer with [`AutoTokenizer`]:
 
@@ -203,8 +181,6 @@ Load a tokenizer with [`AutoTokenizer`]:
 >>> tokenizer = AutoTokenizer.from_pretrained(model_name)
 ```
 
-Next, the tokenizer converts the tokens into numbers in order to construct a tensor as input to the model. This is known as the model's *vocabulary*.
-
 Pass your text to the tokenizer:
 
 ```py
@@ -215,12 +191,12 @@ Pass your text to the tokenizer:
  'attention_mask': [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]}
 ```
 
-The tokenizer will return a dictionary containing:
+The tokenizer returns a dictionary containing:
 
 * [input_ids](./glossary#input-ids): numerical representions of your tokens.
 * [atttention_mask](.glossary#attention-mask): indicates which tokens should be attended to.
 
-Just like the [`pipeline`], the tokenizer will accept a list of inputs. In addition, the tokenizer can also pad and truncate the text to return a batch with uniform length:
+A tokenizer can also accept a list of inputs, and pad and truncate the text to return a batch with uniform length:
 
 <frameworkcontent>
 <pt>
@@ -247,13 +223,17 @@ Just like the [`pipeline`], the tokenizer will accept a list of inputs. In addit
 </tf>
 </frameworkcontent>
 
-Read the [preprocessing](./preprocessing) tutorial for more details about tokenization.
+<Tip>
+
+Check out the [preprocess](./preprocessing) tutorial for more details about tokenization, and how to use an [`AutoFeatureExtractor`] and [`AutoProcessor`] to preprocess image, audio, and multimodal inputs.
+
+</Tip>
 
 ### AutoModel
 
 <frameworkcontent>
 <pt>
-🤗 Transformers provides a simple and unified way to load pretrained instances. This means you can load an [`AutoModel`] like you would load an [`AutoTokenizer`]. The only difference is selecting the correct [`AutoModel`] for the task. Since you are doing text - or sequence - classification, load [`AutoModelForSequenceClassification`]:
+🤗 Transformers provides a simple and unified way to load pretrained instances. This means you can load an [`AutoModel`] like you would load an [`AutoTokenizer`]. The only difference is selecting the correct [`AutoModel`] for the task. For text (or sequence) classification, you should load [`AutoModelForSequenceClassification`]:
 
 ```py
 >>> from transformers import AutoModelForSequenceClassification
@@ -264,11 +244,11 @@ Read the [preprocessing](./preprocessing) tutorial for more details about tokeni
 
 <Tip>
 
-See the [task summary](./task_summary) for which [`AutoModel`] class to use for which task.
+See the [task summary](./task_summary) for tasks supported by an [`AutoModel`] class.
 
 </Tip>
 
-Now you can pass your preprocessed batch of inputs directly to the model. You just have to unpack the dictionary by adding `**`:
+Now pass your preprocessed batch of inputs directly to the model. You just have to unpack the dictionary by adding `**`:
 
 ```py
 >>> pt_outputs = pt_model(**pt_batch)
@@ -286,7 +266,7 @@ tensor([[0.0021, 0.0018, 0.0115, 0.2121, 0.7725],
 ```
 </pt>
 <tf>
-🤗 Transformers provides a simple and unified way to load pretrained instances. This means you can load an [`TFAutoModel`] like you would load an [`AutoTokenizer`]. The only difference is selecting the correct [`TFAutoModel`] for the task. Since you are doing text - or sequence - classification, load [`TFAutoModelForSequenceClassification`]:
+🤗 Transformers provides a simple and unified way to load pretrained instances. This means you can load an [`TFAutoModel`] like you would load an [`AutoTokenizer`]. The only difference is selecting the correct [`TFAutoModel`] for the task. For text (or sequence) classification, you should load [`TFAutoModelForSequenceClassification`]:
 
 ```py
 >>> from transformers import TFAutoModelForSequenceClassification
@@ -297,11 +277,11 @@ tensor([[0.0021, 0.0018, 0.0115, 0.2121, 0.7725],
 
 <Tip>
 
-See the [task summary](./task_summary) for which [`AutoModel`] class to use for which task.
+See the [task summary](./task_summary) for tasks supported by an [`AutoModel`] class.
 
 </Tip>
 
-Now you can pass your preprocessed batch of inputs directly to the model by passing the dictionary keys directly to the tensors:
+Now pass your preprocessed batch of inputs directly to the model by passing the dictionary keys directly to the tensors:
 
 ```py
 >>> tf_outputs = tf_model(tf_batch)
@@ -320,17 +300,8 @@ The model outputs the final activations in the `logits` attribute. Apply the sof
 
 <Tip>
 
-All 🤗 Transformers models (PyTorch or TensorFlow) outputs the tensors *before* the final activation
-function (like softmax) because the final activation function is often fused with the loss.
-
-</Tip>
-
-Models are a standard [`torch.nn.Module`](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) or a [`tf.keras.Model`](https://www.tensorflow.org/api_docs/python/tf/keras/Model) so you can use them in your usual training loop. However, to make things easier, 🤗 Transformers provides a [`Trainer`] class for PyTorch that adds functionality for distributed training, mixed precision, and more. For TensorFlow, you can use the `fit` method from [Keras](https://keras.io/). Refer to the [training tutorial](./training) for more details.
-
-<Tip>
-
-🤗 Transformers model outputs are special dataclasses so their attributes are autocompleted in an IDE.
-The model outputs also behave like a tuple or a dictionary (e.g., you can index with an integer, a slice or a string) in which case the attributes that are `None` are ignored.
+All 🤗 Transformers models (PyTorch or TensorFlow) output the tensors *before* the final activation
+function (like softmax) because the final activation function is often fused with the loss. Model outputs are special dataclasses so their attributes are autocompleted in an IDE. The model outputs behave like a tuple or a dictionary (you can index with an integer, a slice or a string) in which case, attributes that are None are ignored.
 
 </Tip>
 
@@ -425,6 +396,133 @@ Create a model from your custom configuration with [`TFAutoModel.from_config`]:
 
 Take a look at the [Create a custom architecture](./create_a_model) guide for more information about building custom configurations.
 
+## Trainer - a PyTorch optimized training loop
+
+All models are a standard [`torch.nn.Module`](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) so you can use them in any typical training loop. While you can write your own training loop, 🤗 Transformers provides a [`Trainer`] class for PyTorch, which contains the basic training loop and adds additional functionality for features like distributed training, mixed precision, and more.
+
+Depending on your task, you'll typically pass the following parameters to [`Trainer`]:
+
+1. A [`PreTrainedModel`] or a [`torch.nn.Module`](https://pytorch.org/docs/stable/nn.html#torch.nn.Module):
+
+   ```py
+   >>> from transformers import AutoModelForSequenceClassification
+
+   >>> model = AutoModelForSequenceClassification.from_pretrained("distilbert-base-uncased")
+   ```
+
+2. [`TrainingArguments`] contains the model hyperparameters you can change like learning rate, batch size, and the number of epochs to train for. The default values are used if you don't specify any training arguments:
+
+   ```py
+   >>> from transformers import TrainingArguments
+
+   >>> training_args = TrainingArguments(
+   ...     output_dir="path/to/save/folder/",
+   ...     learning_rate=2e-5,
+   ...     per_device_train_batch_size=8,
+   ...     per_device_eval_batch_size=8,
+   ...     num_train_epochs=2,
+   ... )
+   ```
+
+3. A preprocessing class like a tokenizer, feature extractor, or processor:
+
+   ```py
+   >>> from transformers import AutoTokenizer
+
+   >>> tokenizer = AutoTokenizer.from_pretrained("distilbert-base-uncased")
+   ```
+
+4. Your preprocessed train and test datasets:
+
+   ```py
+   >>> train_dataset = dataset["train"]  # doctest: +SKIP
+   >>> eval_dataset = dataset["eval"]  # doctest: +SKIP
+   ```
+
+5. A [`DataCollator`] to create a batch of examples from your dataset:
+
+   ```py
+   >>> from transformers import DefaultDataCollator
+
+   >>> data_collator = DefaultDataCollator()
+   ```
+
+Now gather all these classes in [`Trainer`]:
+
+```py
+>>> from transformers import Trainer
+
+>>> trainer = Trainer(
+...     model=model,
+...     args=training_args,
+...     train_dataset=dataset["train"],
+...     eval_dataset=dataset["test"],
+...     tokenizer=tokenizer,
+...     data_collator=data_collator,
+... )  # doctest: +SKIP
+```
+
+When you're ready, call [`~Trainer.train`] to start training:
+
+```py
+>>> trainer.train()  # doctest: +SKIP
+```
+
+<Tip>
+
+For tasks - like translation or summarization - that use a sequence-to-sequence model, use the [`Seq2SeqTrainer`] and [`Seq2SeqTrainingArguments`] classes instead.
+
+</Tip>
+
+You can customize the training loop behavior by subclassing the methods inside [`Trainer`]. This allows you to customize features such as the loss function, optimizer, and scheduler. Take a look at the [`Trainer`] reference for which methods can be subclassed. 
+
+The other way to customize the training loop is by using [Callbacks](./main_classes/callbacks). You can use callbacks to integrate with other libraries and inspect the training loop to report on progress or stop the training early. Callbacks do not modify anything in the training loop itself. To customize something like the loss function, you need to subclass the [`Trainer`] instead.
+
+## Train with TensorFlow
+
+All models are a standard [`tf.keras.Model`](https://www.tensorflow.org/api_docs/python/tf/keras/Model) so they can be trained in TensorFlow with the [Keras](https://keras.io/) API. 🤗 Transformers provides the [`~TFPreTrainedModel.prepare_tf_dataset`] method to easily load your dataset as a `tf.data.Dataset` so you can start training right away with Keras' [`compile`](https://keras.io/api/models/model_training_apis/#compile-method) and [`fit`](https://keras.io/api/models/model_training_apis/#fit-method) methods.
+
+1. You'll start with a [`TFPreTrainedModel`] or a [`tf.keras.Model`](https://www.tensorflow.org/api_docs/python/tf/keras/Model):
+
+   ```py
+   >>> from transformers import TFAutoModelForSequenceClassification
+
+   >>> model = TFAutoModelForSequenceClassification.from_pretrained("distilbert-base-uncased")
+   ```
+
+2. A preprocessing class like a tokenizer, feature extractor, or processor:
+
+   ```py
+   >>> from transformers import AutoTokenizer
+
+   >>> tokenizer = AutoTokenizer.from_pretrained("distilbert-base-uncased")
+   ```
+
+3. Create a function to tokenize the dataset:
+
+   ```py
+   >>> def tokenize_dataset(dataset):
+   ...     return tokenizer(dataset["text"])  # doctest: +SKIP
+   ```
+
+4. Apply the tokenizer over the entire dataset with [`~datasets.Dataset.map`] and then pass the dataset and tokenizer to [`~TFPreTrainedModel.prepare_tf_dataset`]. You can also change the batch size and shuffle the dataset here if you'd like:
+
+   ```py
+   >>> dataset = dataset.map(tokenize_dataset)  # doctest: +SKIP
+   >>> tf_dataset = model.prepare_tf_dataset(
+   ...     dataset, batch_size=16, shuffle=True, tokenizer=tokenizer
+   ... )  # doctest: +SKIP
+   ```
+
+5. When you're ready, you can call `compile` and `fit` to start training:
+
+   ```py
+   >>> from tensorflow.keras.optimizers import Adam
+
+   >>> model.compile(optimizer=Adam(3e-5))
+   >>> model.fit(dataset)  # doctest: +SKIP
+   ```
+
 ## What's next?
 
 Now that you've completed the 🤗 Transformers quick tour, check out our guides and learn how to do more specific things like writing a custom model, fine-tuning a model for a task, and how to train a model with a script. If you're interested in learning more about 🤗 Transformers core concepts, grab a cup of coffee and take a look at our Conceptual Guides!

docs/source/en/run_scripts.mdx

@@ -187,7 +187,7 @@ python run_summarization.py  \
 
 ## Run a script with 🤗 Accelerate
 
-🤗 [Accelerate](https://huggingface.co/docs/accelerate/index.html) is a PyTorch-only library that offers a unified method for training a model on several types of setups (CPU-only, multiple GPUs, TPUs) while maintaining complete visibility into the PyTorch training loop. Make sure you have 🤗 Accelerate installed if you don't already have it:
+🤗 [Accelerate](https://huggingface.co/docs/accelerate) is a PyTorch-only library that offers a unified method for training a model on several types of setups (CPU-only, multiple GPUs, TPUs) while maintaining complete visibility into the PyTorch training loop. Make sure you have 🤗 Accelerate installed if you don't already have it:
 
 > Note: As Accelerate is rapidly developing, the git version of accelerate must be installed to run the scripts
 ```bash

docs/source/en/serialization.mdx

@@ -10,36 +10,36 @@ an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express o
 specific language governing permissions and limitations under the License.
 -->
 
-# Export 🤗 Transformers Models
+# Export to ONNX
 
-If you need to deploy 🤗 Transformers models in production environments, we
-recommend exporting them to a serialized format that can be loaded and executed
-on specialized runtimes and hardware. In this guide, we'll show you how to
-export 🤗 Transformers models in two widely used formats: ONNX and TorchScript.
+If you need to deploy 🤗 Transformers models in production environments, we recommend
+exporting them to a serialized format that can be loaded and executed on specialized
+runtimes and hardware. In this guide, we'll show you how to export 🤗 Transformers
+models to [ONNX (Open Neural Network eXchange)](http://onnx.ai).
 
-Once exported, a model can optimized for inference via techniques such as
-quantization and pruning. If you are interested in optimizing your models to run
-with maximum efficiency, check out the [🤗 Optimum
+<Tip>
+
+Once exported, a model can be optimized for inference via techniques such as
+quantization and pruning. If you are interested in optimizing your models to run with
+maximum efficiency, check out the [🤗 Optimum
 library](https://github.com/huggingface/optimum).
 
-## ONNX
+</Tip>
 
-The [ONNX (Open Neural Network eXchange)](http://onnx.ai) project is an open
-standard that defines a common set of operators and a common file format to
-represent deep learning models in a wide variety of frameworks, including
-PyTorch and TensorFlow. When a model is exported to the ONNX format, these
-operators are used to construct a computational graph (often called an
-_intermediate representation_) which represents the flow of data through the
-neural network.
+ONNX is an open standard that defines a common set of operators and a common file format
+to represent deep learning models in a wide variety of frameworks, including PyTorch and
+TensorFlow. When a model is exported to the ONNX format, these operators are used to
+construct a computational graph (often called an _intermediate representation_) which
+represents the flow of data through the neural network.
 
-By exposing a graph with standardized operators and data types, ONNX makes it
-easy to switch between frameworks. For example, a model trained in PyTorch can
-be exported to ONNX format and then imported in TensorFlow (and vice versa).
+By exposing a graph with standardized operators and data types, ONNX makes it easy to
+switch between frameworks. For example, a model trained in PyTorch can be exported to
+ONNX format and then imported in TensorFlow (and vice versa).
 
-🤗 Transformers provides a `transformers.onnx` package that enables you to
-convert model checkpoints to an ONNX graph by leveraging configuration objects.
-These configuration objects come ready made for a number of model architectures,
-and are designed to be easily extendable to other architectures.
+🤗 Transformers provides a [`transformers.onnx`](main_classes/onnx) package that enables
+you to convert model checkpoints to an ONNX graph by leveraging configuration objects.
+These configuration objects come ready made for a number of model architectures, and are
+designed to be easily extendable to other architectures.
 
 Ready-made configurations include the following architectures:
 
@@ -55,7 +55,9 @@ Ready-made configurations include the following architectures:
 - BlenderbotSmall
 - BLOOM
 - CamemBERT
+- CLIP
 - CodeGen
+- Conditional DETR
 - ConvBERT
 - ConvNeXT
 - Data2VecText
@@ -66,27 +68,35 @@ Ready-made configurations include the following architectures:
 - DETR
 - DistilBERT
 - ELECTRA
+- ERNIE
 - FlauBERT
 - GPT Neo
 - GPT-J
+- GroupViT
 - I-BERT
 - LayoutLM
 - LayoutLMv3
 - LeViT
+- Longformer
 - LongT5
 - M2M100
 - Marian
 - mBART
 - MobileBERT
 - MobileViT
+- MT5
 - OpenAI GPT-2
+- OWL-ViT
 - Perceiver
 - PLBart
 - ResNet
 - RoBERTa
 - RoFormer
+- SegFormer
 - SqueezeBERT
+- Swin Transformer
 - T5
+- Vision Encoder decoder
 - ViT
 - XLM
 - XLM-RoBERTa
@@ -98,10 +108,10 @@ In the next two sections, we'll show you how to:
 * Export a supported model using the `transformers.onnx` package.
 * Export a custom model for an unsupported architecture.
 
-### Exporting a model to ONNX
+## Exporting a model to ONNX
 
-To export a 🤗 Transformers model to ONNX, you'll first need to install some
-extra dependencies:
+To export a 🤗 Transformers model to ONNX, you'll first need to install some extra
+dependencies:
 
 ```bash
 pip install transformers[onnx]
@@ -133,7 +143,7 @@ Exporting a checkpoint using a ready-made configuration can be done as follows:
 python -m transformers.onnx --model=distilbert-base-uncased onnx/
 ```
 
-which should show the following logs:
+You should see the following logs:
 
 ```bash
 Validating ONNX model...
@@ -144,13 +154,13 @@ Validating ONNX model...
 All good, model saved at: onnx/model.onnx
 ```
 
-This exports an ONNX graph of the checkpoint defined by the `--model` argument.
-In this example it is `distilbert-base-uncased`, but it can be any checkpoint on
-the Hugging Face Hub or one that's stored locally.
+This exports an ONNX graph of the checkpoint defined by the `--model` argument. In this
+example, it is `distilbert-base-uncased`, but it can be any checkpoint on the Hugging
+Face Hub or one that's stored locally.
 
 The resulting `model.onnx` file can then be run on one of the [many
-accelerators](https://onnx.ai/supported-tools.html#deployModel) that support the
-ONNX standard. For example, we can load and run the model with [ONNX
+accelerators](https://onnx.ai/supported-tools.html#deployModel) that support the ONNX
+standard. For example, we can load and run the model with [ONNX
 Runtime](https://onnxruntime.ai/) as follows:
 
 ```python
@@ -164,9 +174,8 @@ Runtime](https://onnxruntime.ai/) as follows:
 >>> outputs = session.run(output_names=["last_hidden_state"], input_feed=dict(inputs))
 ```
 
-The required output names (i.e. `["last_hidden_state"]`) can be obtained by
-taking a look at the ONNX configuration of each model. For example, for
-DistilBERT we have:
+The required output names (like `["last_hidden_state"]`) can be obtained by taking a
+look at the ONNX configuration of each model. For example, for DistilBERT we have:
 
 ```python
 >>> from transformers.models.distilbert import DistilBertConfig, DistilBertOnnxConfig
@@ -177,20 +186,19 @@ DistilBERT we have:
 ["last_hidden_state"]
 ```
 
-The process is identical for TensorFlow checkpoints on the Hub. For example, we
-can export a pure TensorFlow checkpoint from the [Keras
+The process is identical for TensorFlow checkpoints on the Hub. For example, we can
+export a pure TensorFlow checkpoint from the [Keras
 organization](https://huggingface.co/keras-io) as follows:
 
 ```bash
 python -m transformers.onnx --model=keras-io/transformers-qa onnx/
 ```
 
-To export a model that's stored locally, you'll need to have the model's weights
-and tokenizer files stored in a directory. For example, we can load and save a
-checkpoint as follows:
+To export a model that's stored locally, you'll need to have the model's weights and
+tokenizer files stored in a directory. For example, we can load and save a checkpoint as
+follows:
 
-<frameworkcontent>
-<pt>
+<frameworkcontent> <pt>
 ```python
 >>> from transformers import AutoTokenizer, AutoModelForSequenceClassification
 
@@ -208,8 +216,7 @@ argument of the `transformers.onnx` package to the desired directory:
 ```bash
 python -m transformers.onnx --model=local-pt-checkpoint onnx/
 ```
-</pt>
-<tf>
+</pt> <tf>
 ```python
 >>> from transformers import AutoTokenizer, TFAutoModelForSequenceClassification
 
@@ -227,14 +234,13 @@ argument of the `transformers.onnx` package to the desired directory:
 ```bash
 python -m transformers.onnx --model=local-tf-checkpoint onnx/
 ```
-</tf>
-</frameworkcontent>
+</tf> </frameworkcontent>
 
-### Selecting features for different model topologies
+## Selecting features for different model tasks
 
-Each ready-made configuration comes with a set of _features_ that enable you to
-export models for different types of topologies or tasks. As shown in the table
-below, each feature is associated with a different auto class:
+Each ready-made configuration comes with a set of _features_ that enable you to export
+models for different types of tasks. As shown in the table below, each feature is
+associated with a different `AutoClass`:
 
 | Feature                              | Auto Class                           |
 | ------------------------------------ | ------------------------------------ |
@@ -247,7 +253,7 @@ below, each feature is associated with a different auto class:
 | `token-classification`               | `AutoModelForTokenClassification`    |
 
 For each configuration, you can find the list of supported features via the
-`FeaturesManager`. For example, for DistilBERT we have:
+[`~transformers.onnx.FeaturesManager`]. For example, for DistilBERT we have:
 
 ```python
 >>> from transformers.onnx.features import FeaturesManager
@@ -258,15 +264,15 @@ For each configuration, you can find the list of supported features via the
 ```
 
 You can then pass one of these features to the `--feature` argument in the
-`transformers.onnx` package. For example, to export a text-classification model
-we can pick a fine-tuned model from the Hub and run:
+`transformers.onnx` package. For example, to export a text-classification model we can
+pick a fine-tuned model from the Hub and run:
 
 ```bash
 python -m transformers.onnx --model=distilbert-base-uncased-finetuned-sst-2-english \
                             --feature=sequence-classification onnx/
 ```
 
-which will display the following logs:
+This displays the following logs:
 
 ```bash
 Validating ONNX model...
@@ -277,37 +283,42 @@ Validating ONNX model...
 All good, model saved at: onnx/model.onnx
 ```
 
-Notice that in this case, the output names from the fine-tuned model are
-`logits` instead of the `last_hidden_state` we saw with the
-`distilbert-base-uncased` checkpoint earlier. This is expected since the
-fine-tuned model has a sequence classification head.
+Notice that in this case, the output names from the fine-tuned model are `logits`
+instead of the `last_hidden_state` we saw with the `distilbert-base-uncased` checkpoint
+earlier. This is expected since the fine-tuned model has a sequence classification head.
 
 <Tip>
 
-The features that have a `with-past` suffix (e.g. `causal-lm-with-past`)
-correspond to model topologies with precomputed hidden states (key and values
-in the attention blocks) that can be used for fast autoregressive decoding.
+The features that have a `with-past` suffix (like `causal-lm-with-past`) correspond to
+model classes with precomputed hidden states (key and values in the attention blocks)
+that can be used for fast autoregressive decoding.
+
+</Tip>
+
+<Tip>
+
+For `VisionEncoderDecoder` type models, the encoder and decoder parts are
+exported separately as two ONNX files named `encoder_model.onnx` and `decoder_model.onnx` respectively.
 
 </Tip>
 
 
-### Exporting a model for an unsupported architecture
+## Exporting a model for an unsupported architecture
 
-If you wish to export a model whose architecture is not natively supported by
-the library, there are three main steps to follow:
+If you wish to export a model whose architecture is not natively supported by the
+library, there are three main steps to follow:
 
 1. Implement a custom ONNX configuration.
 2. Export the model to ONNX.
 3. Validate the outputs of the PyTorch and exported models.
 
-In this section, we'll look at how DistilBERT was implemented to show what's
-involved with each step.
+In this section, we'll look at how DistilBERT was implemented to show what's involved
+with each step.
 
-#### Implementing a custom ONNX configuration
+### Implementing a custom ONNX configuration
 
-Let's start with the ONNX configuration object. We provide three abstract
-classes that you should inherit from, depending on the type of model
-architecture you wish to export:
+Let's start with the ONNX configuration object. We provide three abstract classes that
+you should inherit from, depending on the type of model architecture you wish to export:
 
 * Encoder-based models inherit from [`~onnx.config.OnnxConfig`]
 * Decoder-based models inherit from [`~onnx.config.OnnxConfigWithPast`]
@@ -339,25 +350,24 @@ Since DistilBERT is an encoder-based model, its configuration inherits from
 ...         )
 ```
 
-Every configuration object must implement the `inputs` property and return a
-mapping, where each key corresponds to an expected input, and each value
-indicates the axis of that input. For DistilBERT, we can see that two inputs are
-required: `input_ids` and `attention_mask`. These inputs have the same shape of
-`(batch_size, sequence_length)` which is why we see the same axes used in the
-configuration.
+Every configuration object must implement the `inputs` property and return a mapping,
+where each key corresponds to an expected input, and each value indicates the axis of
+that input. For DistilBERT, we can see that two inputs are required: `input_ids` and
+`attention_mask`. These inputs have the same shape of `(batch_size, sequence_length)`
+which is why we see the same axes used in the configuration.
 
 <Tip>
 
-Notice that `inputs` property for `DistilBertOnnxConfig` returns an
-`OrderedDict`. This ensures that the inputs are matched with their relative
-position within the `PreTrainedModel.forward()` method when tracing the graph.
-We recommend using an `OrderedDict` for the `inputs` and `outputs` properties
-when implementing custom ONNX configurations.
+Notice that `inputs` property for `DistilBertOnnxConfig` returns an `OrderedDict`. This
+ensures that the inputs are matched with their relative position within the
+`PreTrainedModel.forward()` method when tracing the graph. We recommend using an
+`OrderedDict` for the `inputs` and `outputs` properties when implementing custom ONNX
+configurations.
 
 </Tip>
 
-Once you have implemented an ONNX configuration, you can instantiate it by
-providing the base model's configuration as follows:
+Once you have implemented an ONNX configuration, you can instantiate it by providing the
+base model's configuration as follows:
 
 ```python
 >>> from transformers import AutoConfig
@@ -366,8 +376,8 @@ providing the base model's configuration as follows:
 >>> onnx_config = DistilBertOnnxConfig(config)
 ```
 
-The resulting object has several useful properties. For example you can view the
-ONNX operator set that will be used during the export:
+The resulting object has several useful properties. For example, you can view the ONNX
+operator set that will be used during the export:
 
 ```python
 >>> print(onnx_config.default_onnx_opset)
@@ -381,15 +391,14 @@ You can also view the outputs associated with the model as follows:
 OrderedDict([("last_hidden_state", {0: "batch", 1: "sequence"})])
 ```
 
-Notice that the outputs property follows the same structure as the inputs; it
-returns an `OrderedDict` of named outputs and their shapes. The output structure
-is linked to the choice of feature that the configuration is initialised with.
-By default, the ONNX configuration is initialized with the `default` feature
-that corresponds to exporting a model loaded with the `AutoModel` class. If you
-want to export a different model topology, just provide a different feature to
-the `task` argument when you initialize the ONNX configuration. For example, if
-we wished to export DistilBERT with a sequence classification head, we could
-use:
+Notice that the outputs property follows the same structure as the inputs; it returns an
+`OrderedDict` of named outputs and their shapes. The output structure is linked to the
+choice of feature that the configuration is initialised with. By default, the ONNX
+configuration is initialized with the `default` feature that corresponds to exporting a
+model loaded with the `AutoModel` class. If you want to export a model for another task,
+just provide a different feature to the `task` argument when you initialize the ONNX
+configuration. For example, if we wished to export DistilBERT with a sequence
+classification head, we could use:
 
 ```python
 >>> from transformers import AutoConfig
@@ -402,18 +411,18 @@ OrderedDict([('logits', {0: 'batch'})])
 
 <Tip>
 
-All of the base properties and methods associated with [`~onnx.config.OnnxConfig`] and the
-other configuration classes can be overriden if needed. Check out
-[`BartOnnxConfig`] for an advanced example.
+All of the base properties and methods associated with [`~onnx.config.OnnxConfig`] and
+the other configuration classes can be overriden if needed. Check out [`BartOnnxConfig`]
+for an advanced example.
 
 </Tip>
 
-#### Exporting the model
+### Exporting the model
 
-Once you have implemented the ONNX configuration, the next step is to export the
-model. Here we can use the `export()` function provided by the
-`transformers.onnx` package. This function expects the ONNX configuration, along
-with the base model and tokenizer, and the path to save the exported file:
+Once you have implemented the ONNX configuration, the next step is to export the model.
+Here we can use the `export()` function provided by the `transformers.onnx` package.
+This function expects the ONNX configuration, along with the base model and tokenizer,
+and the path to save the exported file:
 
 ```python
 >>> from pathlib import Path
@@ -428,10 +437,9 @@ with the base model and tokenizer, and the path to save the exported file:
 >>> onnx_inputs, onnx_outputs = export(tokenizer, base_model, onnx_config, onnx_config.default_onnx_opset, onnx_path)
 ```
 
-The `onnx_inputs` and `onnx_outputs` returned by the `export()` function are
-lists of the keys defined in the `inputs` and `outputs` properties of the
-configuration. Once the model is exported, you can test that the model is well
-formed as follows:
+The `onnx_inputs` and `onnx_outputs` returned by the `export()` function are lists of
+the keys defined in the `inputs` and `outputs` properties of the configuration. Once the
+model is exported, you can test that the model is well formed as follows:
 
 ```python
 >>> import onnx
@@ -442,21 +450,20 @@ formed as follows:
 
 <Tip>
 
-If your model is larger than 2GB, you will see that many additional files are
-created during the export. This is _expected_ because ONNX uses [Protocol
-Buffers](https://developers.google.com/protocol-buffers/) to store the model and
-these have a size limit of 2GB. See the [ONNX
-documentation](https://github.com/onnx/onnx/blob/master/docs/ExternalData.md)
-for instructions on how to load models with external data.
+If your model is larger than 2GB, you will see that many additional files are created
+during the export. This is _expected_ because ONNX uses [Protocol
+Buffers](https://developers.google.com/protocol-buffers/) to store the model and these
+have a size limit of 2GB. See the [ONNX
+documentation](https://github.com/onnx/onnx/blob/master/docs/ExternalData.md) for
+instructions on how to load models with external data.
 
 </Tip>
 
-#### Validating the model outputs
+### Validating the model outputs
 
-The final step is to validate that the outputs from the base and exported model
-agree within some absolute tolerance. Here we can use the
-`validate_model_outputs()` function provided by the `transformers.onnx` package
-as follows:
+The final step is to validate that the outputs from the base and exported model agree
+within some absolute tolerance. Here we can use the `validate_model_outputs()` function
+provided by the `transformers.onnx` package as follows:
 
 ```python
 >>> from transformers.onnx import validate_model_outputs
@@ -466,220 +473,23 @@ as follows:
 ... )
 ```
 
-This function uses the `OnnxConfig.generate_dummy_inputs()` method to generate
-inputs for the base and exported model, and the absolute tolerance can be
-defined in the configuration. We generally find numerical agreement in the 1e-6
-to 1e-4 range, although anything smaller than 1e-3 is likely to be OK.
+This function uses the [`~transformers.onnx.OnnxConfig.generate_dummy_inputs`] method to
+generate inputs for the base and exported model, and the absolute tolerance can be
+defined in the configuration. We generally find numerical agreement in the 1e-6 to 1e-4
+range, although anything smaller than 1e-3 is likely to be OK.
 
-### Contributing a new configuration to 🤗 Transformers
+## Contributing a new configuration to 🤗 Transformers
 
-We are looking to expand the set of ready-made configurations and welcome
-contributions from the community! If you would like to contribute your addition
-to the library, you will need to:
+We are looking to expand the set of ready-made configurations and welcome contributions
+from the community! If you would like to contribute your addition to the library, you
+will need to:
 
 * Implement the ONNX configuration in the corresponding `configuration_<model_name>.py`
 file
-* Include the model architecture and corresponding features in [`~onnx.features.FeatureManager`]
+* Include the model architecture and corresponding features in
+  [`~onnx.features.FeatureManager`]
 * Add your model architecture to the tests in `test_onnx_v2.py`
 
 Check out how the configuration for [IBERT was
-contributed](https://github.com/huggingface/transformers/pull/14868/files) to
-get an idea of what's involved.
-
-## TorchScript
-
-<Tip>
-
-This is the very beginning of our experiments with TorchScript and we are still exploring its capabilities with
-variable-input-size models. It is a focus of interest to us and we will deepen our analysis in upcoming releases,
-with more code examples, a more flexible implementation, and benchmarks comparing python-based codes with compiled
-TorchScript.
-
-</Tip>
-
-According to Pytorch's documentation: "TorchScript is a way to create serializable and optimizable models from PyTorch
-code". Pytorch's two modules [JIT and TRACE](https://pytorch.org/docs/stable/jit.html) allow the developer to export
-their model to be re-used in other programs, such as efficiency-oriented C++ programs.
-
-We have provided an interface that allows the export of 🤗 Transformers models to TorchScript so that they can be reused
-in a different environment than a Pytorch-based python program. Here we explain how to export and use our models using
-TorchScript.
-
-Exporting a model requires two things:
-
-- a forward pass with dummy inputs.
-- model instantiation with the `torchscript` flag.
-
-These necessities imply several things developers should be careful about. These are detailed below.
-
-### TorchScript flag and tied weights
-
-This flag is necessary because most of the language models in this repository have tied weights between their
-`Embedding` layer and their `Decoding` layer. TorchScript does not allow the export of models that have tied
-weights, therefore it is necessary to untie and clone the weights beforehand.
-
-This implies that models instantiated with the `torchscript` flag have their `Embedding` layer and `Decoding`
-layer separate, which means that they should not be trained down the line. Training would de-synchronize the two
-layers, leading to unexpected results.
-
-This is not the case for models that do not have a Language Model head, as those do not have tied weights. These models
-can be safely exported without the `torchscript` flag.
-
-### Dummy inputs and standard lengths
-
-The dummy inputs are used to do a model forward pass. While the inputs' values are propagating through the layers,
-Pytorch keeps track of the different operations executed on each tensor. These recorded operations are then used to
-create the "trace" of the model.
-
-The trace is created relatively to the inputs' dimensions. It is therefore constrained by the dimensions of the dummy
-input, and will not work for any other sequence length or batch size. When trying with a different size, an error such
-as:
-
-`The expanded size of the tensor (3) must match the existing size (7) at non-singleton dimension 2`
-
-will be raised. It is therefore recommended to trace the model with a dummy input size at least as large as the largest
-input that will be fed to the model during inference. Padding can be performed to fill the missing values. As the model
-will have been traced with a large input size however, the dimensions of the different matrix will be large as well,
-resulting in more calculations.
-
-It is recommended to be careful of the total number of operations done on each input and to follow performance closely
-when exporting varying sequence-length models.
-
-### Using TorchScript in Python
-
-Below is an example, showing how to save, load models as well as how to use the trace for inference.
-
-#### Saving a model
-
-This snippet shows how to use TorchScript to export a `BertModel`. Here the `BertModel` is instantiated according
-to a `BertConfig` class and then saved to disk under the filename `traced_bert.pt`
-
-```python
-from transformers import BertModel, BertTokenizer, BertConfig
-import torch
-
-enc = BertTokenizer.from_pretrained("bert-base-uncased")
-
-# Tokenizing input text
-text = "[CLS] Who was Jim Henson ? [SEP] Jim Henson was a puppeteer [SEP]"
-tokenized_text = enc.tokenize(text)
-
-# Masking one of the input tokens
-masked_index = 8
-tokenized_text[masked_index] = "[MASK]"
-indexed_tokens = enc.convert_tokens_to_ids(tokenized_text)
-segments_ids = [0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1]
-
-# Creating a dummy input
-tokens_tensor = torch.tensor([indexed_tokens])
-segments_tensors = torch.tensor([segments_ids])
-dummy_input = [tokens_tensor, segments_tensors]
-
-# Initializing the model with the torchscript flag
-# Flag set to True even though it is not necessary as this model does not have an LM Head.
-config = BertConfig(
-    vocab_size_or_config_json_file=32000,
-    hidden_size=768,
-    num_hidden_layers=12,
-    num_attention_heads=12,
-    intermediate_size=3072,
-    torchscript=True,
-)
-
-# Instantiating the model
-model = BertModel(config)
-
-# The model needs to be in evaluation mode
-model.eval()
-
-# If you are instantiating the model with *from_pretrained* you can also easily set the TorchScript flag
-model = BertModel.from_pretrained("bert-base-uncased", torchscript=True)
-
-# Creating the trace
-traced_model = torch.jit.trace(model, [tokens_tensor, segments_tensors])
-torch.jit.save(traced_model, "traced_bert.pt")
-```
-
-#### Loading a model
-
-This snippet shows how to load the `BertModel` that was previously saved to disk under the name `traced_bert.pt`.
-We are re-using the previously initialised `dummy_input`.
-
-```python
-loaded_model = torch.jit.load("traced_bert.pt")
-loaded_model.eval()
-
-all_encoder_layers, pooled_output = loaded_model(*dummy_input)
-```
-
-#### Using a traced model for inference
-
-Using the traced model for inference is as simple as using its `__call__` dunder method:
-
-```python
-traced_model(tokens_tensor, segments_tensors)
-```
-
-### Deploying HuggingFace TorchScript models on AWS using the Neuron SDK
-
-AWS introduced the [Amazon EC2 Inf1](https://aws.amazon.com/ec2/instance-types/inf1/)
-instance family for low cost, high performance machine learning inference in the cloud.
-The Inf1 instances are powered by the AWS Inferentia chip, a custom-built hardware accelerator,
-specializing in deep learning inferencing workloads.
-[AWS Neuron](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/#)
-is the SDK for Inferentia that supports tracing and optimizing transformers models for
-deployment on Inf1. The Neuron SDK provides:
-
-
-1. Easy-to-use API with one line of code change to trace and optimize a TorchScript model for inference in the cloud.
-2. Out of the box performance optimizations for [improved cost-performance](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/neuron-guide/benchmark/>)
-3. Support for HuggingFace transformers models built with either [PyTorch](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/src/examples/pytorch/bert_tutorial/tutorial_pretrained_bert.html)
-   or [TensorFlow](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/src/examples/tensorflow/huggingface_bert/huggingface_bert.html).
-
-#### Implications
-
-Transformers Models based on the [BERT (Bidirectional Encoder Representations from Transformers)](https://huggingface.co/docs/transformers/main/model_doc/bert)
-architecture, or its variants such as [distilBERT](https://huggingface.co/docs/transformers/main/model_doc/distilbert)
- and [roBERTa](https://huggingface.co/docs/transformers/main/model_doc/roberta)
- will run best on Inf1 for non-generative tasks such as Extractive Question Answering,
- Sequence Classification, Token Classification. Alternatively, text generation
-tasks can be adapted to run on Inf1, according to this [AWS Neuron MarianMT tutorial](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/src/examples/pytorch/transformers-marianmt.html).
-More information about models that can be converted out of the box on Inferentia can be
-found in the [Model Architecture Fit section of the Neuron documentation](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/neuron-guide/models/models-inferentia.html#models-inferentia).
-
-#### Dependencies
-
-Using AWS Neuron to convert models requires the following dependencies and environment:
-
-* A [Neuron SDK environment](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/neuron-guide/neuron-frameworks/pytorch-neuron/index.html#installation-guide),
-  which comes pre-configured on [AWS Deep Learning AMI](https://docs.aws.amazon.com/dlami/latest/devguide/tutorial-inferentia-launching.html).
-
-#### Converting a Model for AWS Neuron
-
-Using the same script as in [Using TorchScript in Python](https://huggingface.co/docs/transformers/main/en/serialization#using-torchscript-in-python)
-to trace a "BertModel", you import `torch.neuron` framework extension to access
-the components of the Neuron SDK through a Python API.
-
-```python
-from transformers import BertModel, BertTokenizer, BertConfig
-import torch
-import torch.neuron
-```
-And only modify the tracing line of code
-
-from:
-
-```python
-torch.jit.trace(model, [tokens_tensor, segments_tensors])
-```
-
-to:
-
-```python
-torch.neuron.trace(model, [token_tensor, segments_tensors])
-```
-
-This change enables Neuron SDK to trace the model and optimize it to run in Inf1 instances.
-
-To learn more about AWS Neuron SDK features, tools, example tutorials and latest updates,
-please see the [AWS NeuronSDK documentation](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/index.html).
+contributed](https://github.com/huggingface/transformers/pull/14868/files) to get an
+idea of what's involved.