test ci

* add fix and examples

* fix camel case naming

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

@@ -3,7 +3,7 @@ name: Build docker images (scheduled)
 on:
   push:
     branches:
-      - update-quantization-docker
+      - build_ci_docker_image*
   repository_dispatch:
   workflow_call:
     inputs:
@@ -18,341 +18,341 @@ concurrency:
   cancel-in-progress: false
 
 jobs:
-  # latest-docker:
-  #   name: "Latest PyTorch + TensorFlow [dev]"
-  #   runs-on:
-  #     group: aws-general-8-plus
-  #   steps:
-  #     -
-  #       name: Set up Docker Buildx
-  #       uses: docker/setup-buildx-action@v3
-  #     -
-  #       name: Check out code
-  #       uses: actions/checkout@v4
-  #     -
-  #       name: Login to DockerHub
-  #       uses: docker/login-action@v3
-  #       with:
-  #         username: ${{ secrets.DOCKERHUB_USERNAME }}
-  #         password: ${{ secrets.DOCKERHUB_PASSWORD }}
-  #     -
-  #       name: Build and push
-  #       uses: docker/build-push-action@v5
-  #       with:
-  #         context: ./docker/transformers-all-latest-gpu
-  #         build-args: |
-  #           REF=main
-  #         push: true
-  #         tags: huggingface/transformers-all-latest-gpu${{ inputs.image_postfix }}
-  #     # Push CI images still need to be re-built daily
-  #     -
-  #       name: Build and push (for Push CI) in a daily basis
-  #       # This condition allows `schedule` events, or `push` events that trigger this workflow NOT via `workflow_call`.
-  #       # The later case is useful for manual image building for debugging purpose. Use another tag in this case!
-  #       if: inputs.image_postfix != '-push-ci'
-  #       uses: docker/build-push-action@v5
-  #       with:
-  #         context: ./docker/transformers-all-latest-gpu
-  #         build-args: |
-  #           REF=main
-  #         push: true
-  #         tags: huggingface/transformers-all-latest-gpu-push-ci
+  latest-docker:
+    name: "Latest PyTorch + TensorFlow [dev]"
+    runs-on:
+      group: aws-general-8-plus
+    steps:
+      -
+        name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+      -
+        name: Check out code
+        uses: actions/checkout@v4
+      -
+        name: Login to DockerHub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_PASSWORD }}
+      -
+        name: Build and push
+        uses: docker/build-push-action@v5
+        with:
+          context: ./docker/transformers-all-latest-gpu
+          build-args: |
+            REF=main
+          push: true
+          tags: huggingface/transformers-all-latest-gpu${{ inputs.image_postfix }}
+      # Push CI images still need to be re-built daily
+      -
+        name: Build and push (for Push CI) in a daily basis
+        # This condition allows `schedule` events, or `push` events that trigger this workflow NOT via `workflow_call`.
+        # The later case is useful for manual image building for debugging purpose. Use another tag in this case!
+        if: inputs.image_postfix != '-push-ci'
+        uses: docker/build-push-action@v5
+        with:
+          context: ./docker/transformers-all-latest-gpu
+          build-args: |
+            REF=main
+          push: true
+          tags: huggingface/transformers-all-latest-gpu-push-ci
 
-  #     - name: Post to Slack
-  #       if: always()
-  #       uses: huggingface/hf-workflows/.github/actions/post-slack@main
-  #       with:
-  #         slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
-  #         title: 🤗 Results of the transformers-all-latest-gpu-push-ci docker build 
-  #         status: ${{ job.status }}
-  #         slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
+      - name: Post to Slack
+        if: always()
+        uses: huggingface/hf-workflows/.github/actions/post-slack@main
+        with:
+          slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
+          title: 🤗 Results of the transformers-all-latest-gpu-push-ci docker build 
+          status: ${{ job.status }}
+          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
-  # latest-torch-deepspeed-docker:
-  #   name: "Latest PyTorch + DeepSpeed"
-  #   runs-on:
-  #     group: aws-general-8-plus
-  #   steps:
-  #     -
-  #       name: Set up Docker Buildx
-  #       uses: docker/setup-buildx-action@v3
-  #     -
-  #       name: Check out code
-  #       uses: actions/checkout@v4
-  #     -
-  #       name: Login to DockerHub
-  #       uses: docker/login-action@v3
-  #       with:
-  #         username: ${{ secrets.DOCKERHUB_USERNAME }}
-  #         password: ${{ secrets.DOCKERHUB_PASSWORD }}
-  #     -
-  #       name: Build and push
-  #       uses: docker/build-push-action@v5
-  #       with:
-  #         context: ./docker/transformers-pytorch-deepspeed-latest-gpu
-  #         build-args: |
-  #           REF=main
-  #         push: true
-  #         tags: huggingface/transformers-pytorch-deepspeed-latest-gpu${{ inputs.image_postfix }}
+  latest-torch-deepspeed-docker:
+    name: "Latest PyTorch + DeepSpeed"
+    runs-on:
+      group: aws-general-8-plus
+    steps:
+      -
+        name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+      -
+        name: Check out code
+        uses: actions/checkout@v4
+      -
+        name: Login to DockerHub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_PASSWORD }}
+      -
+        name: Build and push
+        uses: docker/build-push-action@v5
+        with:
+          context: ./docker/transformers-pytorch-deepspeed-latest-gpu
+          build-args: |
+            REF=main
+          push: true
+          tags: huggingface/transformers-pytorch-deepspeed-latest-gpu${{ inputs.image_postfix }}
 
-  #     - name: Post to Slack
-  #       if: always()
-  #       uses: huggingface/hf-workflows/.github/actions/post-slack@main
-  #       with:
-  #         slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER}}
-  #         title: 🤗 Results of the transformers-pytorch-deepspeed-latest-gpu docker build 
-  #         status: ${{ job.status }}
-  #         slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
+      - name: Post to Slack
+        if: always()
+        uses: huggingface/hf-workflows/.github/actions/post-slack@main
+        with:
+          slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER}}
+          title: 🤗 Results of the transformers-pytorch-deepspeed-latest-gpu docker build 
+          status: ${{ job.status }}
+          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
-  # # Can't build 2 images in a single job `latest-torch-deepspeed-docker` (for `nvcr.io/nvidia`)
-  # latest-torch-deepspeed-docker-for-push-ci-daily-build:
-  #   name: "Latest PyTorch + DeepSpeed (Push CI - Daily Build)"
-  #   runs-on:
-  #     group: aws-general-8-plus
-  #   steps:
-  #     -
-  #       name: Set up Docker Buildx
-  #       uses: docker/setup-buildx-action@v3
-  #     -
-  #       name: Check out code
-  #       uses: actions/checkout@v4
-  #     -
-  #       name: Login to DockerHub
-  #       uses: docker/login-action@v3
-  #       with:
-  #         username: ${{ secrets.DOCKERHUB_USERNAME }}
-  #         password: ${{ secrets.DOCKERHUB_PASSWORD }}
-  #     # Push CI images still need to be re-built daily
-  #     -
-  #       name: Build and push (for Push CI) in a daily basis
-  #       # This condition allows `schedule` events, or `push` events that trigger this workflow NOT via `workflow_call`.
-  #       # The later case is useful for manual image building for debugging purpose. Use another tag in this case!
-  #       if: inputs.image_postfix != '-push-ci'
-  #       uses: docker/build-push-action@v5
-  #       with:
-  #         context: ./docker/transformers-pytorch-deepspeed-latest-gpu
-  #         build-args: |
-  #           REF=main
-  #         push: true
-  #         tags: huggingface/transformers-pytorch-deepspeed-latest-gpu-push-ci
+  # Can't build 2 images in a single job `latest-torch-deepspeed-docker` (for `nvcr.io/nvidia`)
+  latest-torch-deepspeed-docker-for-push-ci-daily-build:
+    name: "Latest PyTorch + DeepSpeed (Push CI - Daily Build)"
+    runs-on:
+      group: aws-general-8-plus
+    steps:
+      -
+        name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+      -
+        name: Check out code
+        uses: actions/checkout@v4
+      -
+        name: Login to DockerHub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_PASSWORD }}
+      # Push CI images still need to be re-built daily
+      -
+        name: Build and push (for Push CI) in a daily basis
+        # This condition allows `schedule` events, or `push` events that trigger this workflow NOT via `workflow_call`.
+        # The later case is useful for manual image building for debugging purpose. Use another tag in this case!
+        if: inputs.image_postfix != '-push-ci'
+        uses: docker/build-push-action@v5
+        with:
+          context: ./docker/transformers-pytorch-deepspeed-latest-gpu
+          build-args: |
+            REF=main
+          push: true
+          tags: huggingface/transformers-pytorch-deepspeed-latest-gpu-push-ci
 
-  #     - name: Post to Slack
-  #       if: always()
-  #       uses: huggingface/hf-workflows/.github/actions/post-slack@main
-  #       with:
-  #         slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
-  #         title: 🤗 Results of the transformers-pytorch-deepspeed-latest-gpu-push-ci docker build 
-  #         status: ${{ job.status }}
-  #         slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
+      - name: Post to Slack
+        if: always()
+        uses: huggingface/hf-workflows/.github/actions/post-slack@main
+        with:
+          slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
+          title: 🤗 Results of the transformers-pytorch-deepspeed-latest-gpu-push-ci docker build 
+          status: ${{ job.status }}
+          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
-  # doc-builder:
-  #   name: "Doc builder"
-  #   # Push CI doesn't need this image
-  #   if: inputs.image_postfix != '-push-ci'
-  #   runs-on:
-  #     group: aws-general-8-plus
-  #   steps:
-  #     -
-  #       name: Set up Docker Buildx
-  #       uses: docker/setup-buildx-action@v3
-  #     -
-  #       name: Check out code
-  #       uses: actions/checkout@v4
-  #     -
-  #       name: Login to DockerHub
-  #       uses: docker/login-action@v3
-  #       with:
-  #         username: ${{ secrets.DOCKERHUB_USERNAME }}
-  #         password: ${{ secrets.DOCKERHUB_PASSWORD }}
-  #     -
-  #       name: Build and push
-  #       uses: docker/build-push-action@v5
-  #       with:
-  #         context: ./docker/transformers-doc-builder
-  #         push: true
-  #         tags: huggingface/transformers-doc-builder
+  doc-builder:
+    name: "Doc builder"
+    # Push CI doesn't need this image
+    if: inputs.image_postfix != '-push-ci'
+    runs-on:
+      group: aws-general-8-plus
+    steps:
+      -
+        name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+      -
+        name: Check out code
+        uses: actions/checkout@v4
+      -
+        name: Login to DockerHub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_PASSWORD }}
+      -
+        name: Build and push
+        uses: docker/build-push-action@v5
+        with:
+          context: ./docker/transformers-doc-builder
+          push: true
+          tags: huggingface/transformers-doc-builder
 
-  #     - name: Post to Slack
-  #       if: always()
-  #       uses: huggingface/hf-workflows/.github/actions/post-slack@main
-  #       with:
-  #         slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
-  #         title: 🤗 Results of the huggingface/transformers-doc-builder docker build 
-  #         status: ${{ job.status }}
-  #         slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
+      - name: Post to Slack
+        if: always()
+        uses: huggingface/hf-workflows/.github/actions/post-slack@main
+        with:
+          slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
+          title: 🤗 Results of the huggingface/transformers-doc-builder docker build 
+          status: ${{ job.status }}
+          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
-  # latest-pytorch:
-  #   name: "Latest PyTorch [dev]"
-  #   # Push CI doesn't need this image
-  #   if: inputs.image_postfix != '-push-ci'
-  #   runs-on:
-  #     group: aws-general-8-plus
-  #   steps:
-  #     -
-  #       name: Set up Docker Buildx
-  #       uses: docker/setup-buildx-action@v3
-  #     -
-  #       name: Check out code
-  #       uses: actions/checkout@v4
-  #     -
-  #       name: Login to DockerHub
-  #       uses: docker/login-action@v3
-  #       with:
-  #         username: ${{ secrets.DOCKERHUB_USERNAME }}
-  #         password: ${{ secrets.DOCKERHUB_PASSWORD }}
-  #     -
-  #       name: Build and push
-  #       uses: docker/build-push-action@v5
-  #       with:
-  #         context: ./docker/transformers-pytorch-gpu
-  #         build-args: |
-  #           REF=main
-  #         push: true
-  #         tags: huggingface/transformers-pytorch-gpu
+  latest-pytorch:
+    name: "Latest PyTorch [dev]"
+    # Push CI doesn't need this image
+    if: inputs.image_postfix != '-push-ci'
+    runs-on:
+      group: aws-general-8-plus
+    steps:
+      -
+        name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+      -
+        name: Check out code
+        uses: actions/checkout@v4
+      -
+        name: Login to DockerHub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_PASSWORD }}
+      -
+        name: Build and push
+        uses: docker/build-push-action@v5
+        with:
+          context: ./docker/transformers-pytorch-gpu
+          build-args: |
+            REF=main
+          push: true
+          tags: huggingface/transformers-pytorch-gpu
 
-  #     - name: Post to Slack
-  #       if: always()
-  #       uses: huggingface/hf-workflows/.github/actions/post-slack@main
-  #       with:
-  #         slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
-  #         title: 🤗 Results of the huggingface/transformers-pytorch-gpudocker build 
-  #         status: ${{ job.status }}
-  #         slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
+      - name: Post to Slack
+        if: always()
+        uses: huggingface/hf-workflows/.github/actions/post-slack@main
+        with:
+          slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
+          title: 🤗 Results of the huggingface/transformers-pytorch-gpudocker build 
+          status: ${{ job.status }}
+          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
-  # latest-pytorch-amd:
-  #   name: "Latest PyTorch (AMD) [dev]"
-  #   runs-on:
-  #     group: aws-general-8-plus
-  #   steps:
-  #     - 
-  #       name: Set up Docker Buildx
-  #       uses: docker/setup-buildx-action@v3
-  #     - 
-  #       name: Check out code
-  #       uses: actions/checkout@v4
-  #     - 
-  #       name: Login to DockerHub
-  #       uses: docker/login-action@v3
-  #       with:
-  #         username: ${{ secrets.DOCKERHUB_USERNAME }}
-  #         password: ${{ secrets.DOCKERHUB_PASSWORD }}
-  #     - 
-  #       name: Build and push
-  #       uses: docker/build-push-action@v5
-  #       with:
-  #         context: ./docker/transformers-pytorch-amd-gpu
-  #         build-args: |
-  #           REF=main
-  #         push: true
-  #         tags: huggingface/transformers-pytorch-amd-gpu${{ inputs.image_postfix }}
-  #     # Push CI images still need to be re-built daily
-  #     -
-  #       name: Build and push (for Push CI) in a daily basis
-  #       # This condition allows `schedule` events, or `push` events that trigger this workflow NOT via `workflow_call`.
-  #       # The later case is useful for manual image building for debugging purpose. Use another tag in this case!
-  #       if: inputs.image_postfix != '-push-ci'
-  #       uses: docker/build-push-action@v5
-  #       with:
-  #         context: ./docker/transformers-pytorch-amd-gpu
-  #         build-args: |
-  #           REF=main
-  #         push: true
-  #         tags: huggingface/transformers-pytorch-amd-gpu-push-ci
+  latest-pytorch-amd:
+    name: "Latest PyTorch (AMD) [dev]"
+    runs-on:
+      group: aws-general-8-plus
+    steps:
+      - 
+        name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+      - 
+        name: Check out code
+        uses: actions/checkout@v4
+      - 
+        name: Login to DockerHub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_PASSWORD }}
+      - 
+        name: Build and push
+        uses: docker/build-push-action@v5
+        with:
+          context: ./docker/transformers-pytorch-amd-gpu
+          build-args: |
+            REF=main
+          push: true
+          tags: huggingface/transformers-pytorch-amd-gpu${{ inputs.image_postfix }}
+      # Push CI images still need to be re-built daily
+      -
+        name: Build and push (for Push CI) in a daily basis
+        # This condition allows `schedule` events, or `push` events that trigger this workflow NOT via `workflow_call`.
+        # The later case is useful for manual image building for debugging purpose. Use another tag in this case!
+        if: inputs.image_postfix != '-push-ci'
+        uses: docker/build-push-action@v5
+        with:
+          context: ./docker/transformers-pytorch-amd-gpu
+          build-args: |
+            REF=main
+          push: true
+          tags: huggingface/transformers-pytorch-amd-gpu-push-ci
 
-  #     - name: Post to Slack
-  #       if: always()
-  #       uses: huggingface/hf-workflows/.github/actions/post-slack@main
-  #       with:
-  #         slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
-  #         title: 🤗 Results of the huggingface/transformers-pytorch-amd-gpu-push-ci build 
-  #         status: ${{ job.status }}
-  #         slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
+      - name: Post to Slack
+        if: always()
+        uses: huggingface/hf-workflows/.github/actions/post-slack@main
+        with:
+          slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
+          title: 🤗 Results of the huggingface/transformers-pytorch-amd-gpu-push-ci build 
+          status: ${{ job.status }}
+          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
-  # latest-tensorflow:
-  #   name: "Latest TensorFlow [dev]"
-  #   # Push CI doesn't need this image
-  #   if: inputs.image_postfix != '-push-ci'
-  #   runs-on:
-  #     group: aws-general-8-plus
-  #   steps:
-  #     -
-  #       name: Set up Docker Buildx
-  #       uses: docker/setup-buildx-action@v3
-  #     -
-  #       name: Check out code
-  #       uses: actions/checkout@v4
-  #     -
-  #       name: Login to DockerHub
-  #       uses: docker/login-action@v3
-  #       with:
-  #         username: ${{ secrets.DOCKERHUB_USERNAME }}
-  #         password: ${{ secrets.DOCKERHUB_PASSWORD }}
-  #     -
-  #       name: Build and push
-  #       uses: docker/build-push-action@v5
-  #       with:
-  #         context: ./docker/transformers-tensorflow-gpu
-  #         build-args: |
-  #           REF=main
-  #         push: true
-  #         tags: huggingface/transformers-tensorflow-gpu
+  latest-tensorflow:
+    name: "Latest TensorFlow [dev]"
+    # Push CI doesn't need this image
+    if: inputs.image_postfix != '-push-ci'
+    runs-on:
+      group: aws-general-8-plus
+    steps:
+      -
+        name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+      -
+        name: Check out code
+        uses: actions/checkout@v4
+      -
+        name: Login to DockerHub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_PASSWORD }}
+      -
+        name: Build and push
+        uses: docker/build-push-action@v5
+        with:
+          context: ./docker/transformers-tensorflow-gpu
+          build-args: |
+            REF=main
+          push: true
+          tags: huggingface/transformers-tensorflow-gpu
 
-  #     - name: Post to Slack
-  #       if: always()
-  #       uses: huggingface/hf-workflows/.github/actions/post-slack@main
-  #       with:
-  #         slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
-  #         title: 🤗 Results of the huggingface/transformers-tensorflow-gpu build 
-  #         status: ${{ job.status }}
-  #         slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
+      - name: Post to Slack
+        if: always()
+        uses: huggingface/hf-workflows/.github/actions/post-slack@main
+        with:
+          slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
+          title: 🤗 Results of the huggingface/transformers-tensorflow-gpu build 
+          status: ${{ job.status }}
+          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
-  # latest-pytorch-deepspeed-amd:
-  #   name: "PyTorch + DeepSpeed (AMD) [dev]"
-  #   runs-on:
-  #     group: aws-general-8-plus
-  #   steps:
-  #     - 
-  #       name: Set up Docker Buildx
-  #       uses: docker/setup-buildx-action@v3
-  #     - 
-  #       name: Check out code
-  #       uses: actions/checkout@v4
-  #     - 
-  #       name: Login to DockerHub
-  #       uses: docker/login-action@v3
-  #       with:
-  #         username: ${{ secrets.DOCKERHUB_USERNAME }}
-  #         password: ${{ secrets.DOCKERHUB_PASSWORD }}
-  #     - 
-  #       name: Build and push
-  #       uses: docker/build-push-action@v5
-  #       with:
-  #         context: ./docker/transformers-pytorch-deepspeed-amd-gpu
-  #         build-args: |
-  #           REF=main
-  #         push: true
-  #         tags: huggingface/transformers-pytorch-deepspeed-amd-gpu${{ inputs.image_postfix }}
-  #     # Push CI images still need to be re-built daily
-  #     -
-  #       name: Build and push (for Push CI) in a daily basis
-  #       # This condition allows `schedule` events, or `push` events that trigger this workflow NOT via `workflow_call`.
-  #       # The later case is useful for manual image building for debugging purpose. Use another tag in this case!
-  #       if: inputs.image_postfix != '-push-ci'
-  #       uses: docker/build-push-action@v5
-  #       with:
-  #         context: ./docker/transformers-pytorch-deepspeed-amd-gpu
-  #         build-args: |
-  #           REF=main
-  #         push: true
-  #         tags: huggingface/transformers-pytorch-deepspeed-amd-gpu-push-ci
+  latest-pytorch-deepspeed-amd:
+    name: "PyTorch + DeepSpeed (AMD) [dev]"
+    runs-on:
+      group: aws-general-8-plus
+    steps:
+      - 
+        name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+      - 
+        name: Check out code
+        uses: actions/checkout@v4
+      - 
+        name: Login to DockerHub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_PASSWORD }}
+      - 
+        name: Build and push
+        uses: docker/build-push-action@v5
+        with:
+          context: ./docker/transformers-pytorch-deepspeed-amd-gpu
+          build-args: |
+            REF=main
+          push: true
+          tags: huggingface/transformers-pytorch-deepspeed-amd-gpu${{ inputs.image_postfix }}
+      # Push CI images still need to be re-built daily
+      -
+        name: Build and push (for Push CI) in a daily basis
+        # This condition allows `schedule` events, or `push` events that trigger this workflow NOT via `workflow_call`.
+        # The later case is useful for manual image building for debugging purpose. Use another tag in this case!
+        if: inputs.image_postfix != '-push-ci'
+        uses: docker/build-push-action@v5
+        with:
+          context: ./docker/transformers-pytorch-deepspeed-amd-gpu
+          build-args: |
+            REF=main
+          push: true
+          tags: huggingface/transformers-pytorch-deepspeed-amd-gpu-push-ci
 
-  #     - name: Post to Slack
-  #       if: always()
-  #       uses: huggingface/hf-workflows/.github/actions/post-slack@main
-  #       with:
-  #         slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
-  #         title: 🤗 Results of the transformers-pytorch-deepspeed-amd-gpu build 
-  #         status: ${{ job.status }}
-  #         slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
+      - name: Post to Slack
+        if: always()
+        uses: huggingface/hf-workflows/.github/actions/post-slack@main
+        with:
+          slack_channel: ${{ secrets.CI_SLACK_CHANNEL_DOCKER }}
+          title: 🤗 Results of the transformers-pytorch-deepspeed-amd-gpu build 
+          status: ${{ job.status }}
+          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
 
   latest-quantization-torch-docker:
     name: "Latest Pytorch + Quantization [dev]"

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

@@ -7,64 +7,64 @@ on:
     - cron: "17 2 * * *"
   push:
     branches:
-      - run_scheduled_ci*
+      - fix_gptq_tests
 
 jobs:
-  model-ci:
-    name: Model CI
-    uses: ./.github/workflows/self-scheduled.yml
-    with:
-      job: run_models_gpu
-      slack_report_channel: "#transformers-ci-daily-models"
-      runner: daily-ci
-      docker: huggingface/transformers-all-latest-gpu
-      ci_event: Daily CI
-    secrets: inherit
+  # model-ci:
+  #   name: Model CI
+  #   uses: ./.github/workflows/self-scheduled.yml
+  #   with:
+  #     job: run_models_gpu
+  #     slack_report_channel: "#transformers-ci-daily-models"
+  #     runner: daily-ci
+  #     docker: huggingface/transformers-all-latest-gpu
+  #     ci_event: Daily CI
+  #   secrets: inherit
 
-  torch-pipeline:
-    name: Torch pipeline CI
-    uses: ./.github/workflows/self-scheduled.yml
-    with:
-      job: run_pipelines_torch_gpu
-      slack_report_channel: "#transformers-ci-daily-pipeline-torch"
-      runner: daily-ci
-      docker: huggingface/transformers-pytorch-gpu
-      ci_event: Daily CI
-    secrets: inherit
+  # torch-pipeline:
+  #   name: Torch pipeline CI
+  #   uses: ./.github/workflows/self-scheduled.yml
+  #   with:
+  #     job: run_pipelines_torch_gpu
+  #     slack_report_channel: "#transformers-ci-daily-pipeline-torch"
+  #     runner: daily-ci
+  #     docker: huggingface/transformers-pytorch-gpu
+  #     ci_event: Daily CI
+  #   secrets: inherit
 
-  tf-pipeline:
-    name: TF pipeline CI
-    uses: ./.github/workflows/self-scheduled.yml
-    with:
-      job: run_pipelines_tf_gpu
-      slack_report_channel: "#transformers-ci-daily-pipeline-tf"
-      runner: daily-ci
-      docker: huggingface/transformers-tensorflow-gpu
-      ci_event: Daily CI
-    secrets: inherit
+  # tf-pipeline:
+  #   name: TF pipeline CI
+  #   uses: ./.github/workflows/self-scheduled.yml
+  #   with:
+  #     job: run_pipelines_tf_gpu
+  #     slack_report_channel: "#transformers-ci-daily-pipeline-tf"
+  #     runner: daily-ci
+  #     docker: huggingface/transformers-tensorflow-gpu
+  #     ci_event: Daily CI
+  #   secrets: inherit
 
-  example-ci:
-    name: Example CI
-    uses: ./.github/workflows/self-scheduled.yml
-    with:
-      job: run_examples_gpu
-      slack_report_channel: "#transformers-ci-daily-examples"
-      runner: daily-ci
-      docker: huggingface/transformers-all-latest-gpu
-      ci_event: Daily CI
-    secrets: inherit
+  # example-ci:
+  #   name: Example CI
+  #   uses: ./.github/workflows/self-scheduled.yml
+  #   with:
+  #     job: run_examples_gpu
+  #     slack_report_channel: "#transformers-ci-daily-examples"
+  #     runner: daily-ci
+  #     docker: huggingface/transformers-all-latest-gpu
+  #     ci_event: Daily CI
+  #   secrets: inherit
 
-  deepspeed-ci:
-    name: DeepSpeed CI
-    uses: ./.github/workflows/self-scheduled.yml
-    with:
-      job: run_torch_cuda_extensions_gpu
-      slack_report_channel: "#transformers-ci-daily-deepspeed"
-      runner: daily-ci
-      docker: huggingface/transformers-pytorch-deepspeed-latest-gpu
-      ci_event: Daily CI
-      working-directory-prefix: /workspace
-    secrets: inherit
+  # deepspeed-ci:
+  #   name: DeepSpeed CI
+  #   uses: ./.github/workflows/self-scheduled.yml
+  #   with:
+  #     job: run_torch_cuda_extensions_gpu
+  #     slack_report_channel: "#transformers-ci-daily-deepspeed"
+  #     runner: daily-ci
+  #     docker: huggingface/transformers-pytorch-deepspeed-latest-gpu
+  #     ci_event: Daily CI
+  #     working-directory-prefix: /workspace
+  #   secrets: inherit
 
   quantization-ci:
     name: Quantization CI

examples/modular-transformers/image_processing_new_imgproc_model.py

@@ -0,0 +1,287 @@
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+#           This file was automatically generated from examples/modular-transformers/modular_new_imgproc_model.py.
+#               Do NOT edit this file manually as any edits will be overwritten by the generation of
+#             the file from the modular. If any change should be done, please apply the change to the
+#                          modular_new_imgproc_model.py file directly. One of our CI enforces this.
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+from typing import Dict, List, Optional, Union
+
+import numpy as np
+import torch
+
+from ...image_processing_utils import BaseImageProcessor, BatchFeature, get_size_dict
+from ...image_transforms import convert_to_rgb, resize, to_channel_dimension_format
+from ...image_utils import (
+    OPENAI_CLIP_MEAN,
+    OPENAI_CLIP_STD,
+    ChannelDimension,
+    ImageInput,
+    PILImageResampling,
+    infer_channel_dimension_format,
+    is_scaled_image,
+    make_list_of_images,
+    to_numpy_array,
+    valid_images,
+    validate_preprocess_arguments,
+)
+from ...utils import TensorType, filter_out_non_signature_kwargs, is_vision_available, logging
+
+
+if is_vision_available():
+    import PIL
+
+
+logger = logging.get_logger(__name__)
+
+
+class ImgprocModelImageProcessor(BaseImageProcessor):
+    r"""
+    Constructs a NEW_IMGPROC_MODEL image processor.
+
+    Args:
+        do_resize (`bool`, *optional*, defaults to `True`):
+            Whether to resize the image's (height, width) dimensions to the specified `size`. Can be overridden by the
+            `do_resize` parameter in the `preprocess` method.
+        size (`dict`, *optional*, defaults to `{"height": 384, "width": 384}`):
+            Size of the output image after resizing. Can be overridden by the `size` parameter in the `preprocess`
+            method.
+        resample (`PILImageResampling`, *optional*, defaults to `Resampling.BICUBIC`):
+            Resampling filter to use if resizing the image. Only has an effect if `do_resize` is set to `True`. Can be
+            overridden by the `resample` parameter in the `preprocess` method.
+        do_rescale (`bool`, *optional*, defaults to `True`):
+            Whether to rescale the image by the specified scale `rescale_factor`. Can be overridden by the
+            `do_rescale` parameter in the `preprocess` method.
+        rescale_factor (`int` or `float`, *optional*, defaults to `1/255`):
+            Scale factor to use if rescaling the image. Only has an effect if `do_rescale` is set to `True`. Can be
+            overridden by the `rescale_factor` parameter in the `preprocess` method.
+        do_normalize (`bool`, *optional*, defaults to `True`):
+            Whether to normalize the image. Can be overridden by the `do_normalize` parameter in the `preprocess`
+            method. Can be overridden by the `do_normalize` parameter in the `preprocess` method.
+        image_mean (`float` or `List[float]`, *optional*, defaults to `IMAGENET_STANDARD_MEAN`):
+            Mean to use if normalizing the image. This is a float or list of floats the length of the number of
+            channels in the image. Can be overridden by the `image_mean` parameter in the `preprocess` method. Can be
+            overridden by the `image_mean` parameter in the `preprocess` method.
+        image_std (`float` or `List[float]`, *optional*, defaults to `IMAGENET_STANDARD_STD`):
+            Standard deviation to use if normalizing the image. This is a float or list of floats the length of the
+            number of channels in the image. Can be overridden by the `image_std` parameter in the `preprocess` method.
+            Can be overridden by the `image_std` parameter in the `preprocess` method.
+        do_convert_rgb (`bool`, *optional*, defaults to `True`):
+            Whether to convert the image to RGB.
+    """
+
+    model_input_names = ["pixel_values"]
+
+    def __init__(
+        self,
+        do_resize: bool = True,
+        size: Dict[str, int] = None,
+        resample: PILImageResampling = PILImageResampling.BICUBIC,
+        do_rescale: bool = True,
+        rescale_factor: Union[int, float] = 1 / 255,
+        do_normalize: bool = True,
+        image_mean: Optional[Union[float, List[float]]] = None,
+        image_std: Optional[Union[float, List[float]]] = None,
+        do_convert_rgb: bool = True,
+        **kwargs,
+    ) -> None:
+        super().__init__(**kwargs)
+        size = size if size is not None else {"height": 384, "width": 384}
+        size = get_size_dict(size, default_to_square=True)
+
+        self.do_resize = do_resize
+        self.size = size
+        self.resample = resample
+        self.do_rescale = do_rescale
+        self.rescale_factor = rescale_factor
+        self.do_normalize = do_normalize
+        self.image_mean = image_mean if image_mean is not None else OPENAI_CLIP_MEAN
+        self.image_std = image_std if image_std is not None else OPENAI_CLIP_STD
+        self.do_convert_rgb = do_convert_rgb
+
+    def resize(
+        self,
+        image: np.ndarray,
+        size: Dict[str, int],
+        resample: PILImageResampling = PILImageResampling.BICUBIC,
+        data_format: Optional[Union[str, ChannelDimension]] = None,
+        input_data_format: Optional[Union[str, ChannelDimension]] = None,
+        **kwargs,
+    ) -> np.ndarray:
+        """
+        Resize an image to `(size["height"], size["width"])`.
+
+        Args:
+            image (`np.ndarray`):
+                Image to resize.
+            size (`Dict[str, int]`):
+                Dictionary in the format `{"height": int, "width": int}` specifying the size of the output image.
+            resample (`PILImageResampling`, *optional*, defaults to `PILImageResampling.BICUBIC`):
+                `PILImageResampling` filter to use when resizing the image e.g. `PILImageResampling.BICUBIC`.
+            data_format (`ChannelDimension` or `str`, *optional*):
+                The channel dimension format for the output image. If unset, the channel dimension format of the input
+                image is used. Can be one of:
+                - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
+                - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
+                - `"none"` or `ChannelDimension.NONE`: image in (height, width) format.
+            input_data_format (`ChannelDimension` or `str`, *optional*):
+                The channel dimension format for the input image. If unset, the channel dimension format is inferred
+                from the input image. Can be one of:
+                - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
+                - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
+                - `"none"` or `ChannelDimension.NONE`: image in (height, width) format.
+
+        Returns:
+            `np.ndarray`: The resized image.
+        """
+        size = get_size_dict(size)
+        if "height" not in size or "width" not in size:
+            raise ValueError(f"The `size` dictionary must contain the keys `height` and `width`. Got {size.keys()}")
+        output_size = (size["height"], size["width"])
+        return resize(
+            image,
+            size=output_size,
+            resample=resample,
+            data_format=data_format,
+            input_data_format=input_data_format,
+            **kwargs,
+        )
+
+    @filter_out_non_signature_kwargs()
+    def preprocess(
+        self,
+        images: ImageInput,
+        do_resize: Optional[bool] = None,
+        size: Optional[Dict[str, int]] = None,
+        resample: PILImageResampling = None,
+        do_rescale: Optional[bool] = None,
+        rescale_factor: Optional[float] = None,
+        do_normalize: Optional[bool] = None,
+        image_mean: Optional[Union[float, List[float]]] = None,
+        image_std: Optional[Union[float, List[float]]] = None,
+        return_tensors: Optional[Union[str, TensorType]] = None,
+        do_convert_rgb: bool = None,
+        data_format: ChannelDimension = ChannelDimension.FIRST,
+        input_data_format: Optional[Union[str, ChannelDimension]] = None,
+    ) -> PIL.Image.Image:
+        """
+        Preprocess an image or batch of images.
+
+        Args:
+            images (`ImageInput`):
+                Image to preprocess. Expects a single or batch of images with pixel values ranging from 0 to 255. If
+                passing in images with pixel values between 0 and 1, set `do_rescale=False`.
+            do_resize (`bool`, *optional*, defaults to `self.do_resize`):
+                Whether to resize the image.
+            size (`Dict[str, int]`, *optional*, defaults to `self.size`):
+                Controls the size of the image after `resize`. The shortest edge of the image is resized to
+                `size["shortest_edge"]` whilst preserving the aspect ratio. If the longest edge of this resized image
+                is > `int(size["shortest_edge"] * (1333 / 800))`, then the image is resized again to make the longest
+                edge equal to `int(size["shortest_edge"] * (1333 / 800))`.
+            resample (`PILImageResampling`, *optional*, defaults to `self.resample`):
+                Resampling filter to use if resizing the image. Only has an effect if `do_resize` is set to `True`.
+            do_rescale (`bool`, *optional*, defaults to `self.do_rescale`):
+                Whether to rescale the image values between [0 - 1].
+            rescale_factor (`float`, *optional*, defaults to `self.rescale_factor`):
+                Rescale factor to rescale the image by if `do_rescale` is set to `True`.
+            do_normalize (`bool`, *optional*, defaults to `self.do_normalize`):
+                Whether to normalize the image.
+            image_mean (`float` or `List[float]`, *optional*, defaults to `self.image_mean`):
+                Image mean to normalize the image by if `do_normalize` is set to `True`.
+            image_std (`float` or `List[float]`, *optional*, defaults to `self.image_std`):
+                Image standard deviation to normalize the image by if `do_normalize` is set to `True`.
+            do_convert_rgb (`bool`, *optional*, defaults to `self.do_convert_rgb`):
+                Whether to convert the image to RGB.
+            return_tensors (`str` or `TensorType`, *optional*):
+                The type of tensors to return. Can be one of:
+                    - Unset: Return a list of `np.ndarray`.
+                    - `TensorType.TENSORFLOW` or `'tf'`: Return a batch of type `tf.Tensor`.
+                    - `TensorType.PYTORCH` or `'pt'`: Return a batch of type `torch.Tensor`.
+                    - `TensorType.NUMPY` or `'np'`: Return a batch of type `np.ndarray`.
+                    - `TensorType.JAX` or `'jax'`: Return a batch of type `jax.numpy.ndarray`.
+            data_format (`ChannelDimension` or `str`, *optional*, defaults to `ChannelDimension.FIRST`):
+                The channel dimension format for the output image. Can be one of:
+                - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
+                - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
+                - Unset: Use the channel dimension format of the input image.
+            input_data_format (`ChannelDimension` or `str`, *optional*):
+                The channel dimension format for the input image. If unset, the channel dimension format is inferred
+                from the input image. Can be one of:
+                - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
+                - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
+                - `"none"` or `ChannelDimension.NONE`: image in (height, width) format.
+        """
+        do_resize = do_resize if do_resize is not None else self.do_resize
+        resample = resample if resample is not None else self.resample
+        do_rescale = do_rescale if do_rescale is not None else self.do_rescale
+        rescale_factor = rescale_factor if rescale_factor is not None else self.rescale_factor
+        do_normalize = do_normalize if do_normalize is not None else self.do_normalize
+        image_mean = image_mean if image_mean is not None else self.image_mean
+        image_std = image_std if image_std is not None else self.image_std
+        do_convert_rgb = do_convert_rgb if do_convert_rgb is not None else self.do_convert_rgb
+
+        size = size if size is not None else self.size
+        size = get_size_dict(size, default_to_square=False)
+
+        images = make_list_of_images(images)
+
+        if not valid_images(images):
+            raise ValueError(
+                "Invalid image type. Must be of type PIL.Image.Image, numpy.ndarray, "
+                "torch.Tensor, tf.Tensor or jax.ndarray."
+            )
+
+        validate_preprocess_arguments(
+            do_rescale=do_rescale,
+            rescale_factor=rescale_factor,
+            do_normalize=do_normalize,
+            image_mean=image_mean,
+            image_std=image_std,
+            do_resize=do_resize,
+            size=size,
+            resample=resample,
+        )
+        # PIL RGBA images are converted to RGB
+        if do_convert_rgb:
+            images = [convert_to_rgb(image) for image in images]
+
+        # All transformations expect numpy arrays.
+        images = [to_numpy_array(image) for image in images]
+
+        if is_scaled_image(images[0]) and do_rescale:
+            logger.warning_once(
+                "It looks like you are trying to rescale already rescaled images. If the input"
+                " images have pixel values between 0 and 1, set `do_rescale=False` to avoid rescaling them again."
+            )
+
+        if input_data_format is None:
+            # We assume that all images have the same channel dimension format.
+            input_data_format = infer_channel_dimension_format(images[0])
+
+        if do_resize:
+            images = [
+                self.resize(image=image, size=size, resample=resample, input_data_format=input_data_format)
+                for image in images
+            ]
+
+        if do_rescale:
+            images = [
+                self.rescale(image=image, scale=rescale_factor, input_data_format=input_data_format)
+                for image in images
+            ]
+
+        if do_normalize:
+            images = [
+                self.normalize(image=image, mean=image_mean, std=image_std, input_data_format=input_data_format)
+                for image in images
+            ]
+
+        images = [
+            to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format) for image in images
+        ]
+
+        encoded_outputs = BatchFeature(data={"pixel_values": images}, tensor_type=return_tensors)
+
+        return encoded_outputs
+
+    def new_image_processing_method(self, pixel_values: torch.FloatTensor):
+        return pixel_values / 2

examples/modular-transformers/modular_new_imgproc_model.py

@@ -0,0 +1,9 @@
+import torch
+import torch.utils.checkpoint
+
+from transformers.models.blip.image_processing_blip import BlipImageProcessor
+
+
+class ImgprocModelImageProcessor(BlipImageProcessor):
+    def new_image_processing_method(self, pixel_values: torch.FloatTensor):
+        return pixel_values / 2

src/transformers/generation/utils.py

@@ -15,6 +15,7 @@
 # limitations under the License.
 import copy
 import inspect
+import os
 import warnings
 from dataclasses import dataclass
 from typing import TYPE_CHECKING, Any, Callable, Dict, List, Optional, Tuple, Union
@@ -1029,10 +1030,6 @@ class GenerationMixin:
                 "You have explicitly specified `forced_decoder_ids`. Please remove the `forced_decoder_ids` argument "
                 "in favour of `input_ids` or `decoder_input_ids` respectively.",
             )
-        if generation_config.watermarking_config is not None:
-            processors.append(
-                generation_config.watermarking_config.construct_processor(self.config.vocab_size, device)
-            )
 
         # TODO (joao): find a strategy to specify the order of the processors
         processors = self._merge_criteria_processor_list(processors, logits_processor)
@@ -1085,6 +1082,12 @@ class GenerationMixin:
                     )
                 )
 
+        # Watermarking should be after all logits processing is finished (see #34630)
+        if generation_config.watermarking_config is not None:
+            processors.append(
+                generation_config.watermarking_config.construct_processor(self.config.vocab_size, device)
+            )
+
         # `LogitNormalization` should always be the last logit processor, when present
         if generation_config.renormalize_logits is True:
             processors.append(LogitNormalization())
@@ -3222,6 +3225,16 @@ class GenerationMixin:
         unfinished_sequences = torch.ones(batch_size, dtype=torch.long, device=input_ids.device)
         model_kwargs = self._get_initial_cache_position(input_ids, model_kwargs)
 
+        def model_forward(model, *args, **kwargs):
+            return model.forward(*args, **kwargs)
+
+        if isinstance(model_kwargs.get("past_key_values"), StaticCache):
+            if self.device.type == "cuda":
+                logger.warning_once("Using `torch.compile`.")
+                os.environ["TOKENIZERS_PARALLELISM"] = "0"
+                model_forward = torch.compile(model_forward, mode="reduce-overhead", fullgraph=True)
+
+        i = 0
         while self._has_unfinished_sequences(
             this_peer_finished, synced_gpus, device=input_ids.device, cur_len=cur_len, max_length=max_length
         ):
@@ -3232,8 +3245,11 @@ class GenerationMixin:
             model_inputs.update({"output_attentions": output_attentions} if output_attentions else {})
             model_inputs.update({"output_hidden_states": output_hidden_states} if output_hidden_states else {})
 
-            # forward pass to get next token
-            outputs = self(**model_inputs, return_dict=True)
+            if i == 0:
+                outputs = self(**model_inputs, return_dict=True)
+                i += 1
+            else:
+                outputs = model_forward(self, return_dict=True, **model_inputs)
 
             # synced_gpus: don't waste resources running the code we don't need; kwargs must be updated before skipping
             model_kwargs = self._update_model_kwargs_for_generation(

src/transformers/modeling_utils.py

@@ -52,7 +52,6 @@ from .pytorch_utils import (  # noqa: F401
     find_pruneable_heads_and_indices,
     id_tensor_storage,
     is_torch_greater_or_equal_than_1_13,
-    is_torch_greater_or_equal_than_2_4,
     prune_conv1d_layer,
     prune_layer,
     prune_linear_layer,
@@ -90,6 +89,7 @@ from .utils import (
     is_peft_available,
     is_remote_url,
     is_safetensors_available,
+    is_torch_greater_or_equal,
     is_torch_sdpa_available,
     is_torch_xla_available,
     logging,
@@ -5032,7 +5032,7 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
             device_mesh (`torch.distributed.DeviceMesh`):
                 The device mesh to use for tensor parallelism.
         """
-        if not is_torch_greater_or_equal_than_2_4:
+        if not is_torch_greater_or_equal("2.5"):
             raise EnvironmentError("tensor parallel is only supported for `torch>=2.5`.")
 
         # Tensor parallelize a nn.Module based on the `_tp_plan` attribute of the module.

src/transformers/models/idefics3/image_processing_idefics3.py

@@ -38,6 +38,7 @@ from ...utils import TensorType, is_vision_available, logging
 
 
 logger = logging.get_logger(__name__)
+MAX_IMAGE_SIZE = 4096  # 4k resolution as absolute maximum
 
 
 if is_vision_available():
@@ -116,7 +117,6 @@ def _resize_output_size_scale_below_upper_bound(
 def get_resize_output_image_size(
     image,
     resolution_max_side: int,
-    max_image_size: int = 1820,
     input_data_format: Optional[Union[str, ChannelDimension]] = None,
 ) -> Tuple[int, int]:
     """
@@ -126,24 +126,18 @@ def get_resize_output_image_size(
             Image to resize.
         resolution_max_side (`int`):
             The longest edge of the image will be resized to this value. The shortest edge will be resized to keep the
-            input aspect ratio, with a lower bound of `min_image_size`.
-        max_image_size (`int`, *optional*, defaults to 1820):
-            Maximum image resolution. If the image is larger than this size, the longest edge will be resized to this
-            value, with the shortest edge resized to keep the input aspect ratio, with a lower bound of `min_image_size`.
+            input aspect ratio.
         input_data_format (`ChannelDimension` or `str`):
             The channel dimension format of the input image.
     Returns:
         The output size of the image after resizing.
     """
-    if resolution_max_side > max_image_size:
-        raise ValueError("`resolution_max_side` cannot be larger than `max_image_size`")
-
     height, width = get_image_size(image, channel_dim=input_data_format)
 
     # Find the output size, when rescaling the longest edge to max_len and preserving the aspect ratio
     height, width = _resize_output_size_rescale_to_max_len(height, width, max_len=resolution_max_side)
-    # Find the output size when scaling the image to be below the max_image_size
-    height, width = _resize_output_size_scale_below_upper_bound(height, width, max_len=max_image_size)
+    # Find the output size when scaling the image to be below the MAX_IMAGE_SIZE
+    height, width = _resize_output_size_scale_below_upper_bound(height, width, max_len=MAX_IMAGE_SIZE)
     return height, width
 
 
@@ -251,7 +245,7 @@ def convert_to_rgb(
     data_format = input_data_format if data_format is None else data_format
 
     mode = "P" if palette is not None else None
-    image = to_pil_image(image, image_mode=mode)
+    image = to_pil_image(image, image_mode=mode, input_data_format=input_data_format)
     if image.mode == "P" and palette is not None:
         image.putpalette(palette)
 
@@ -404,7 +398,7 @@ class Idefics3ImageProcessor(BaseImageProcessor):
         image_mode = None
         if image.ndim == 2 or image.shape[-1] == 1:
             image_mode = "P"
-        image = to_pil_image(image, image_mode=image_mode)
+        image = to_pil_image(image, image_mode=image_mode, input_data_format=input_data_format)
 
         resized_image = image.resize((size[1], size[0]), resample=resample)
         resized_image = np.array(resized_image)
@@ -754,6 +748,16 @@ class Idefics3ImageProcessor(BaseImageProcessor):
         # All transformations expect numpy arrays.
         images_list = [[to_numpy_array(image) for image in images] for images in images_list]
 
+        # Extra channel dimension for grayscale images
+        if input_data_format in [ChannelDimension.LAST, None]:
+            images_list = [
+                [np.expand_dims(img, axis=-1) if img.ndim == 2 else img for img in images] for images in images_list
+            ]
+        elif input_data_format == ChannelDimension.FIRST:
+            images_list = [
+                [np.expand_dims(img, axis=0) if img.ndim == 2 else img for img in images] for images in images_list
+            ]
+
         if is_scaled_image(images_list[0][0]) and do_rescale:
             logger.warning_once(
                 "It looks like you are trying to rescale already rescaled images. If the input"
@@ -764,18 +768,6 @@ class Idefics3ImageProcessor(BaseImageProcessor):
         if input_data_format is None:
             input_data_format = infer_channel_dimension_format(images_list[0][0], num_channels=(1, 3, 4))
 
-        # Extra channel dimension for grayscale images
-        if input_data_format == ChannelDimension.LAST:
-            images_list = [
-                [np.expand_dims(img, axis=-1) if img.ndim == 2 else img for img in images] for images in images_list
-            ]
-        elif input_data_format == ChannelDimension.FIRST:
-            images_list = [
-                [np.expand_dims(img, axis=0) if img.ndim == 2 else img for img in images] for images in images_list
-            ]
-        else:
-            raise ValueError(f"Invalid channel dimension format {input_data_format}.")
-
         if do_resize:
             images_list = [
                 [

src/transformers/pytorch_utils.py

@@ -21,7 +21,7 @@ from packaging import version
 from safetensors.torch import storage_ptr, storage_size
 from torch import nn
 
-from .utils import is_torch_xla_available, logging
+from .utils import is_torch_greater_or_equal, is_torch_xla_available, logging
 
 
 ALL_LAYERNORM_LAYERS = [nn.LayerNorm]
@@ -39,7 +39,7 @@ is_torch_greater_or_equal_than_1_13 = parsed_torch_version_base >= version.parse
 is_torch_greater_or_equal_than_1_12 = parsed_torch_version_base >= version.parse("1.12")
 
 
-if is_torch_greater_or_equal_than_2_4:
+if is_torch_greater_or_equal("2.5"):
     from torch.distributed.tensor import Replicate
     from torch.distributed.tensor.parallel import (
         ColwiseParallel,

src/transformers/quantizers/base.py

@@ -215,6 +215,9 @@ class HfQuantizer(ABC):
 
         # Delete quantizer and quantization config
         del model.hf_quantizer
+        del model.config.quantization_config
+        del model.config._pre_quantization_dtype
+        model.is_quantized = False
 
         return model
 

src/transformers/quantizers/quantizer_eetq.py

@@ -53,6 +53,20 @@ class EetqHfQuantizer(HfQuantizer):
                 "Please install the latest version of eetq from : https://github.com/NetEase-FuXi/EETQ"
             )
 
+        try:
+            import eetq  # noqa: F401
+        except ImportError as exc:
+            if "shard_checkpoint" in str(exc):
+                # EETQ 1.0.0 is currently broken with the latest transformers because it tries to import the removed
+                # shard_checkpoint function, see https://github.com/NetEase-FuXi/EETQ/issues/34.
+                # TODO: Update message once eetq releases a fix
+                raise ImportError(
+                    "You are using a version of EETQ that is incompatible with the current transformers version. "
+                    "Either downgrade transformers to <= v4.46.3 or, if available, upgrade EETQ to > v1.0.0."
+                ) from exc
+            else:
+                raise
+
         if not is_accelerate_available():
             raise ImportError("Loading an EETQ quantized model requires accelerate (`pip install accelerate`)")
 

src/transformers/testing_utils.py

@@ -1143,7 +1143,17 @@ def require_eetq(test_case):
     """
     Decorator marking a test that requires eetq
     """
-    return unittest.skipUnless(is_eetq_available(), "test requires eetq")(test_case)
+    eetq_available = is_eetq_available()
+    if eetq_available:
+        try:
+            import eetq  # noqa: F401
+        except ImportError as exc:
+            if "shard_checkpoint" in str(exc):
+                # EETQ 1.0.0 is currently broken with the latest transformers because it tries to import the removed
+                # shard_checkpoint function, see https://github.com/NetEase-FuXi/EETQ/issues/34.
+                # TODO: Remove once eetq releases a fix and this release is used in CI
+                eetq_available = False
+    return unittest.skipUnless(eetq_available, "test requires eetq")(test_case)
 
 
 def require_av(test_case):
@@ -1184,9 +1194,19 @@ def require_tensorboard(test_case):
 
 def require_auto_gptq(test_case):
     """
-    Decorator for auto_gptq dependency
+    Decorator for auto_gptq dependency 
     """
-    return unittest.skipUnless(is_auto_gptq_available(), "test requires auto-gptq")(test_case)
+    eetq_available = is_eetq_available()
+    if eetq_available:
+        try:
+            import eetq  # noqa: F401
+        except ImportError as exc:
+            if "shard_checkpoint" in str(exc):
+                # EETQ 1.0.0 is currently broken with the latest transformers because it tries to import the removed
+                # shard_checkpoint function, see https://github.com/NetEase-FuXi/EETQ/issues/34.
+                # TODO: Remove once eetq releases a fix and this release is used in CI
+                eetq_available = False
+    return unittest.skipUnless(is_auto_gptq_available() and eetq_available, "test requires auto-gptq")(test_case)
 
 
 def require_auto_awq(test_case):

tests/generation/test_utils.py

@@ -14,6 +14,7 @@
 # limitations under the License.
 
 
+import collections
 import copy
 import gc
 import inspect
@@ -2450,6 +2451,58 @@ class UtilsFunctionsTest(unittest.TestCase):
         self.assertTrue(n_matches.item() == 2)
         self.assertTrue(validated_tokens.tolist()[0] == [1, 4, 8])
 
+    def test_speculative_sampling_target_distribution(self):
+        """
+        Asserts that the target distribution is preserved.
+        Should help with catching issues like #32867.
+        """
+        # assume vocab size 10, input length 5 + 3 generated candidates
+        candidate_input_ids = torch.tensor([[8, 0, 3, 9, 8, 1, 4, 5]])  # input tokens
+        candidate_logits = torch.tensor(
+            [
+                [
+                    [-10.0, 10.0, -10.0, -10.0, -10.0, -10.0, -10.0, -10.0, -10.0, -10.0],  # generated 1
+                    [-10.0, -10.0, -10.0, -10.0, 10.0, -10.0, -10.0, -10.0, -10.0, -10.0],  # generated 4
+                    [-10.0, -10.0, -10.0, -10.0, -10.0, 10.0, -10.0, -10.0, -10.0, -10.0],  # generated 5
+                ]
+            ]
+        )
+        candidate_length = 3
+        inf = float("inf")
+        new_logits = torch.tensor(
+            [
+                [
+                    # accepts 1:
+                    [-inf, 10.0, -inf, -inf, -inf, -inf, -inf, -inf, -inf, -inf],
+                    # accepts 4:
+                    [-inf, -inf, -inf, -inf, 10.0, -inf, -inf, -inf, -inf, -inf],
+                    # most likely to be 1 or 8, less likely to be 3, then 7, and should never be any other value:
+                    [-inf, 2.0, -inf, 1.0, -inf, -inf, -inf, -0.01, 2.0, -inf],
+                    # N/A:
+                    [-inf, -inf, -inf, -inf, -inf, -inf, -inf, -inf, -inf, -inf],
+                ]
+            ]
+        )
+        last_assistant_token_is_eos = False
+        last_validated_token = []
+        for _ in range(10_000):
+            validated_tokens, n_matches = _speculative_sampling(
+                candidate_input_ids,
+                candidate_logits,
+                candidate_length,
+                new_logits,
+                last_assistant_token_is_eos,
+            )
+            self.assertTrue(n_matches.item() == 2)
+            self.assertTrue(validated_tokens.tolist()[0][0] == 1)
+            self.assertTrue(validated_tokens.tolist()[0][1] == 4)
+            self.assertTrue(validated_tokens.tolist()[0][2] in [1, 3, 7, 8])
+            last_validated_token.append(validated_tokens.tolist()[0][2])
+        # check that the most likely tokens are selected more often than the less likely ones
+        last_token_counts = collections.Counter(last_validated_token)
+        self.assertTrue(last_token_counts[1] > last_token_counts[3] > last_token_counts[7] > 0)
+        self.assertTrue(last_token_counts[8] > last_token_counts[3])
+
 
 @pytest.mark.generate
 @require_torch

tests/quantization/bitnet_integration/test_bitnet.py

@@ -65,7 +65,7 @@ class BitNetTest(unittest.TestCase):
         """
         Load the model
         """
-        cls.tokenizer = AutoTokenizer.from_pretrained("meta-llama/Meta-Llama-3-8B-Instruct")
+        cls.tokenizer = AutoTokenizer.from_pretrained(cls.model_name)
         cls.quantized_model = AutoModelForCausalLM.from_pretrained(cls.model_name, device_map=cls.device)
 
     def tearDown(self):

utils/modular_model_converter.py

@@ -1192,7 +1192,7 @@ class ModularFileMapper(ModuleMapper):
         self.visited_modules = {}
         self.renamers = {}
         for file, module in self.model_specific_modules.items():
-            file_model_name = re.search(r"models\.\w*?\.\w*?_(\S*)", file).groups()[0]
+            file_model_name = file.split(".")[-2]
             renamer = ReplaceNameTransformer(
                 file_model_name, self.model_name, self.given_old_name, self.given_new_name
             )