Merge branch 'main' into update-for-cb

Co-authored-by: Kashif Rasul <kashif.rasul@gmail.com>
Co-authored-by: Shirin Yamani <75791599+shirinyamani@users.noreply.github.com>
Co-authored-by: Quentin Gallouédec <45557362+qgallouedec@users.noreply.github.com>
Co-authored-by: Quentin Gallouédec <gallouedec.quentin@gmail.com>

.github/ISSUE_TEMPLATE/bug-report.yml

@@ -7,36 +7,7 @@ body:
       value: |
         Thanks for taking the time to fill out this bug report! 🤗
 
-        Before you submit your bug report:
-
-          - If it is your first time submitting, be sure to check our [bug report guidelines](https://github.com/huggingface/trl/blob/main/CONTRIBUTING.md#did-you-find-a-bug)
-
-  - type: textarea
-    id: system-info
-    attributes:
-      label: System Info
-      description: Please share your system info with us. You can run the command `transformers-cli env` and copy-paste its output below.
-      placeholder: trl version, transformers version, platform, python version, ...
-    validations:
-      required: true
-
-  - type: checkboxes
-    id: information-scripts-examples
-    attributes:
-      label: Information
-      description: 'The problem arises when using:'
-      options:
-        - label: "The official example scripts"
-        - label: "My own modified scripts"
-
-  - type: checkboxes
-    id: information-tasks
-    attributes:
-      label: Tasks
-      description: "The tasks I am working on are:"
-      options:
-        - label: "An officially supported task in the `examples` folder"
-        - label: "My own task or dataset (give details below)"
+        🚩 If it is your first time submitting, be sure to check our [bug report guidelines](https://github.com/huggingface/trl/blob/main/CONTRIBUTING.md#did-you-find-a-bug)
 
   - type: textarea
     id: reproduction
@@ -50,18 +21,47 @@ body:
         Important! Use code tags to correctly format your code. See https://help.github.com/en/github/writing-on-github/creating-and-highlighting-code-blocks#syntax-highlighting
         Do not use screenshots, as they are hard to read and (more importantly) don't allow others to copy-and-paste your code.
 
-      placeholder: |
-        Steps to reproduce the behavior:
+      value: |
+        ```python
+        from trl import ...
 
-          1.
-          2.
-          3.
+        ```
 
+        outputs:
+
+        ```
+        Traceback (most recent call last):
+          File "example.py", line 42, in <module>
+            ...
+        ```
 
   - type: textarea
-    id: expected-behavior
+    id: system-info
+    attributes:
+      label: System Info
+      description: |
+        Please provide information about your system: platform, Python version, PyTorch version, Transformers version, devices, TRL version, ...
+        You can get this information by running `trl env` in your terminal.
+
+      placeholder: Copy-paste the output of `trl env`
     validations:
       required: true
+
+  - type: checkboxes
+    id: terms
     attributes:
-      label: Expected behavior
-      description: "A clear and concise description of what you would expect to happen."
+      label: Checklist
+      description: |
+        Before submitting, please confirm that you've completed each of the following.
+        If an item doesn't apply to your issue, check it anyway to show you've reviewed it.
+      options:
+        - label: "I have checked that my issue isn't already filed (see [open issues](https://github.com/huggingface/trl/issues?q=is%3Aissue))"
+          required: true
+        - label: "I have included my system information"
+          required: true
+        - label: "Any code provided is minimal, complete, and reproducible ([more on MREs](https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/creating-and-highlighting-code-blocks))"
+          required: true
+        - label: "Any code provided is properly formatted in code blocks, (no screenshot, [more on code blocks](https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/creating-and-highlighting-code-blocks))"
+          required: true
+        - label: "Any traceback provided is complete"
+          required: true

.github/PULL_REQUEST_TEMPLATE.md

@@ -21,8 +21,7 @@ Fixes # (issue)
       Pull Request section?
 - [ ] Was this discussed/approved via a GitHub issue? Please add a link
       to it if that's the case.
-- [ ] Did you make sure to update the documentation with your changes? Here are the
-      [documentation guidelines](https://github.com/huggingface/trl/tree/main/docs).
+- [ ] Did you make sure to update the documentation with your changes?
 - [ ] Did you write any new necessary tests?
 
 

.github/codeql/custom-queries.qls

@@ -0,0 +1,19 @@
+import codeql
+
+from WorkflowString interpolation, Workflow workflow
+where 
+  interpolation.getStringValue().matches("${{ github.event.issue.title }}") or
+  interpolation.getStringValue().matches("${{ github.event.issue.body }}") or
+  interpolation.getStringValue().matches("${{ github.event.pull_request.title }}") or
+  interpolation.getStringValue().matches("${{ github.event.pull_request.body }}") or
+  interpolation.getStringValue().matches("${{ github.event.review.body }}") or
+  interpolation.getStringValue().matches("${{ github.event.comment.body }}") or
+  interpolation.getStringValue().matches("${{ github.event.inputs.* }}") or
+  interpolation.getStringValue().matches("${{ github.event.head_commit.message }}")
+  interpolation.getStringValue().matches("${{ github.event.* }}") and
+  (
+    step.getKey() = "run" or  // Injection in run
+    step.getKey() = "env" or  // Injection via env
+    step.getKey() = "with"    // Injection via with
+  )
+select workflow, "🚨 Do not use directly as input of action"

.github/workflows/build_pr_documentation.yml

@@ -9,6 +9,7 @@ concurrency:
 
 jobs:
   build:
+    if: github.event.pull_request.draft == false
     uses: huggingface/doc-builder/.github/workflows/build_pr_documentation.yml@main
     with:
       commit_sha: ${{ github.event.pull_request.head.sha }}

.github/workflows/codeQL.yml

@@ -0,0 +1,26 @@
+name: "CodeQL Analysis - Workflows"
+
+on:
+  workflow_dispatch:
+
+jobs:
+  analyze:
+    name: "Analyze GitHub Workflows"
+    runs-on: ubuntu-latest
+    permissions:
+      security-events: write
+      actions: read
+      contents: read
+
+    steps:
+      - name: "Checkout repository"
+        uses: actions/checkout@v4
+
+      - name: "Initialize CodeQL"
+        uses: github/codeql-action/init@v2
+        with:
+          languages: "yaml"
+          queries: +security-and-quality, ./.github/codeql/custom-queries.qls
+
+      - name: "Perform CodeQL Analysis"
+        uses: github/codeql-action/analyze@v2

.github/workflows/issue_auto_labeller.yml

@@ -0,0 +1,15 @@
+name: "Hugging Face Issue Labeler"
+on:
+  issues:
+    types: opened
+
+jobs:
+  triage:
+    runs-on: ubuntu-latest
+    permissions:
+      issues: write
+    steps:
+      - uses: actions/checkout@v3
+      - uses: August-murr/auto-labeler@main
+        with:
+            hf-api-key: ${{ secrets.CI_HF_API_TOKEN }}

.github/workflows/pr_style_bot.yml

@@ -0,0 +1,127 @@
+name: PR Style Bot
+
+on:
+  workflow_dispatch:
+
+
+permissions:
+  contents: write
+  pull-requests: write
+
+jobs:
+  run-style-bot:
+    if: >
+      contains(github.event.comment.body, '@bot /style') &&
+      github.event.issue.pull_request != null
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Extract PR details
+        id: pr_info
+        uses: actions/github-script@v6
+        with:
+          script: |
+            const prNumber = context.payload.issue.number;
+            const { data: pr } = await github.rest.pulls.get({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              pull_number: prNumber
+            });
+            
+            // We capture both the branch ref and the "full_name" of the head repo
+            // so that we can check out the correct repository & branch (including forks).
+            core.setOutput("prNumber", prNumber);
+            core.setOutput("headRef", pr.head.ref);
+            core.setOutput("headRepoFullName", pr.head.repo.full_name);
+
+      - name: Check out PR branch
+        uses: actions/checkout@v3
+        env: 
+          HEADREPOFULLNAME: ${{ steps.pr_info.outputs.headRepoFullName }}
+          HEADREF: ${{ steps.pr_info.outputs.headRef }}
+        with:
+          # Instead of checking out the base repo, use the contributor's repo name
+          repository: ${{ env.HEADREPOFULLNAME }}
+          ref: ${{ env.HEADREF }}
+          # You may need fetch-depth: 0 for being able to push
+          fetch-depth: 0
+          token: ${{ secrets.GITHUB_TOKEN }}
+      
+      - name: Debug
+        env: 
+          HEADREPOFULLNAME: ${{ steps.pr_info.outputs.headRepoFullName }}
+          HEADREF: ${{ steps.pr_info.outputs.headRef }}
+          PRNUMBER: ${{ steps.pr_info.outputs.prNumber }}
+        run: |
+          echo "PR number: ${{ env.PRNUMBER }}"
+          echo "Head Ref: ${{ env.HEADREF }}"
+          echo "Head Repo Full Name: ${{ env.HEADREPOFULLNAME }}"
+
+      - name: Set up Python
+        uses: actions/setup-python@v4
+
+      - name: Install dependencies
+        run: |
+          pip install ruff pre-commit
+
+      - name: Download Makefile from main branch
+        run: |
+          curl -o main_Makefile https://raw.githubusercontent.com/huggingface/trl/main/Makefile
+        
+      - name: Compare Makefiles
+        run: |
+          if ! diff -q main_Makefile Makefile; then
+            echo "Error: The Makefile has changed. Please ensure it matches the main branch."
+            exit 1
+          fi
+          echo "No changes in Makefile. Proceeding..."
+          rm -rf main_Makefile
+
+      - name: Run make style and make quality
+        run: |
+          make precommit || true
+
+      - name: Commit and push changes
+        id: commit_and_push
+        env: 
+          HEADREPOFULLNAME: ${{ steps.pr_info.outputs.headRepoFullName }}
+          HEADREF: ${{ steps.pr_info.outputs.headRef }}
+          PRNUMBER: ${{ steps.pr_info.outputs.prNumber }}
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          echo "HEADREPOFULLNAME: ${{ env.HEADREPOFULLNAME }}, HEADREF: ${{ env.HEADREF }}"
+          # Configure git with the Actions bot user
+          git config user.name "github-actions[bot]"
+          git config user.email "github-actions[bot]@users.noreply.github.com"
+
+          # Make sure your 'origin' remote is set to the contributor's fork
+          git remote set-url origin "https://x-access-token:${GITHUB_TOKEN}@github.com/${{ env.HEADREPOFULLNAME }}.git"
+
+          # If there are changes after running style/quality, commit them
+          if [ -n "$(git status --porcelain)" ]; then
+            git add .
+            git commit -m "Apply style fixes"
+            # Push to the original contributor's forked branch
+            git push origin HEAD:${{ env.HEADREF }}
+            echo "changes_pushed=true" >> $GITHUB_OUTPUT
+          else
+            echo "No changes to commit."
+            echo "changes_pushed=false" >> $GITHUB_OUTPUT
+          fi
+
+      - name: Comment on PR with workflow run link
+        if: steps.commit_and_push.outputs.changes_pushed == 'true'
+        uses: actions/github-script@v6
+        with:
+          script: |
+            const prNumber = parseInt(process.env.prNumber, 10);
+            const runUrl = `${process.env.GITHUB_SERVER_URL}/${process.env.GITHUB_REPOSITORY}/actions/runs/${process.env.GITHUB_RUN_ID}`
+
+            await github.rest.issues.createComment({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              issue_number: prNumber,
+              body: `Style fixes have been applied. [View the workflow run here](${runUrl}).`
+            });
+        env:
+          prNumber: ${{ steps.pr_info.outputs.prNumber }}

.github/workflows/slow-tests.yml

@@ -19,7 +19,8 @@ jobs:
       fail-fast: false
       matrix:
         docker-image-name: ["huggingface/trl-latest-gpu:latest", "huggingface/trl-source-gpu:latest"]
-    runs-on: [self-hosted, single-gpu, nvidia-gpu, t4, ci]
+    runs-on:
+      group: aws-g4dn-2xlarge
     env:
       CUDA_VISIBLE_DEVICES: "0"
       TEST_TYPE: "single_gpu_${{ matrix.docker-image-name }}"
@@ -55,7 +56,8 @@ jobs:
       fail-fast: false
       matrix:
         docker-image-name: ["huggingface/trl-latest-gpu:latest", "huggingface/trl-source-gpu:latest"]
-    runs-on: [self-hosted, multi-gpu, nvidia-gpu, t4, ci]
+    runs-on:
+      group: aws-g4dn-2xlarge
     env:
       CUDA_VISIBLE_DEVICES: "0,1"
       TEST_TYPE: "multi_gpu_${{ matrix.docker-image-name }}"

.github/workflows/stale.yml

@@ -1,27 +0,0 @@
-name: Stale Bot
-
-on:
-  schedule:
-    - cron: "0 15 * * *"
-
-jobs:
-  close_stale_issues:
-    name: Close Stale Issues
-    if: github.repository == 'huggingface/trl'
-    runs-on: ubuntu-latest
-    env:
-      GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-    steps:
-    - uses: actions/checkout@v4
-
-    - name: Setup Python
-      uses: actions/setup-python@v5
-      with:
-        python-version: 3.8
-
-    - name: Install requirements
-      run: |
-        pip install PyGithub
-    - name: Close stale issues
-      run: |
-        python scripts/stale.py

.github/workflows/tests-main.yml

@@ -1,46 +0,0 @@
-name: tests on transformers PEFT main
-
-on:
-  push:
-    branches: [ main ]
-
-env:
-  CI_SLACK_CHANNEL: ${{ secrets.CI_PUSH_MAIN_CHANNEL }}
-
-jobs:
-  tests:
-    strategy:
-      matrix:
-        python-version: ['3.9', '3.10', '3.11']
-        os: ['ubuntu-latest', 'windows-latest']
-      fail-fast: false
-    runs-on: ${{ matrix.os }}
-    steps:
-    - uses: actions/checkout@v4
-    - name: Set up Python ${{ matrix.python-version }}
-      uses: actions/setup-python@v5
-      with:
-        python-version: ${{ matrix.python-version }}
-        cache: "pip"
-        cache-dependency-path: |
-            setup.py
-            requirements.txt
-    - name: Install dependencies
-      run: |
-        python -m pip install --upgrade pip
-        # install PEFT & transformers from source
-        pip install -U git+https://github.com/huggingface/peft.git
-        pip install -U git+https://github.com/huggingface/transformers.git 
-        # cpu version of pytorch
-        pip install ".[test, diffusers]"
-    - name: Test with pytest
-      run: |
-        make test
-    - name: Post to Slack
-      if: always()
-      uses: huggingface/hf-workflows/.github/actions/post-slack@main
-      with:
-        slack_channel: ${{ env.CI_SLACK_CHANNEL }}
-        title: 🤗 Results of the TRL CI on transformers/PEFT main
-        status: ${{ job.status }}
-        slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}

.github/workflows/tests.yml

@@ -1,88 +1,252 @@
-name: tests
+name: Tests
 
 on:
   push:
     branches: [ main ]
   pull_request:
-    branches: [ main ]
     paths:
       # Run only when relevant files are modified
-      - "trl/**.py"
+      - ".github/**.yml"
       - "examples/**.py"
       - "scripts/**.py"
-      - ".github/**.yml"
       - "tests/**.py"
+      - "trl/**.py"
+      - "setup.py"
 
 env:
   TQDM_DISABLE: 1
+  CI_SLACK_CHANNEL: ${{ secrets.CI_PUSH_MAIN_CHANNEL }}
 
 jobs:
   check_code_quality:
+    name: Check code quality
     runs-on: ubuntu-latest
-    strategy:
-      matrix:
-        python-version: [3.9]
-
+    if: github.event.pull_request.draft == false
     steps:
       - uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-          submodules: recursive
-      - name: Set up Python ${{ matrix.python-version }}
+      - name: Set up Python 3.12
         uses: actions/setup-python@v5
         with:
-          python-version: ${{ matrix.python-version }}
+          python-version: 3.12
       - uses: pre-commit/action@v3.0.1
         with:
           extra_args: --all-files
 
   tests:
-    needs: check_code_quality
+    name: Tests
     strategy:
       matrix:
-        python-version: ['3.9', '3.10', '3.11']
-        os: ['ubuntu-latest', 'windows-latest']
-    runs-on: ${{ matrix.os }}
+        python-version: ['3.9', '3.10', '3.11', '3.12', '3.13']
+      fail-fast: false
+    runs-on:
+      group: aws-g4dn-2xlarge
+    container:
+      image: pytorch/pytorch:2.6.0-cuda12.6-cudnn9-devel
+      options: --gpus all
+    defaults:
+      run:
+        shell: bash
+    if: github.event.pull_request.draft == false
     steps:
-    - uses: actions/checkout@v4
-    - name: Set up Python ${{ matrix.python-version }}
-      uses: actions/setup-python@v5
-      with:
-        python-version: ${{ matrix.python-version }}
-        cache: "pip"
-        cache-dependency-path: |
-            setup.py
-            requirements.txt
-    - name: Install dependencies
-      run: |
-        python -m pip install --upgrade pip
-        # install PEFT & transformers from source
-        pip install -U git+https://github.com/huggingface/peft.git
-        pip install -U git+https://github.com/huggingface/transformers.git 
-        # cpu version of pytorch
-        pip install ".[test, diffusers]"
-    - name: Test with pytest
-      run: |
-        make test
+      - name: Git checkout
+        uses: actions/checkout@v4
 
-  tests_no_optional_dep:
-    needs: check_code_quality
-    runs-on: 'ubuntu-latest'
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Install Make and Git
+        run: |
+          apt-get update && apt-get install -y make git curl
+
+      - name: Install uv
+        run: |
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+
+      - name: Create Python virtual environment
+        run: |
+          uv venv
+          uv pip install --upgrade setuptools wheel
+
+      - name: Install dependencies
+        run: |
+          source .venv/bin/activate
+          uv pip install ".[dev]"
+
+      - name: Test with pytest
+        run: |
+          source .venv/bin/activate
+          make test
+
+      - name: Post to Slack
+        if: github.ref == 'refs/heads/main' && always()  # Check if the branch is main
+        uses: huggingface/hf-workflows/.github/actions/post-slack@main
+        with:
+          slack_channel: ${{ env.CI_SLACK_CHANNEL }}
+          title: Results with Python ${{ matrix.python-version }} and latest dependencies
+          status: ${{ job.status }}
+          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
+
+  tests_dev:
+    name: Tests with dev dependencies
+    runs-on:
+      group: aws-g4dn-2xlarge
+    container:
+      image: pytorch/pytorch:2.6.0-cuda12.6-cudnn9-devel
+      options: --gpus all
+    defaults:
+      run:
+        shell: bash
+    if: github.event.pull_request.draft == false
     steps:
-    - uses: actions/checkout@v4
-    - name: Set up Python 3.9
-      uses: actions/setup-python@v5
-      with:
-        python-version: '3.9'
-        cache: "pip"
-        cache-dependency-path: |
-            setup.py
-            requirements.txt
-    - name: Install dependencies
-      run: |
-        python -m pip install --upgrade pip
-        # cpu version of pytorch
-        pip install .[test]
-    - name: Test with pytest
-      run: |
-        make test
+      - name: Git checkout
+        uses: actions/checkout@v4
+
+      - name: Set up Python 3.12
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+
+      - name: Install Make and Git
+        run: |
+          apt-get update && apt-get install -y make git curl
+
+      - name: Install uv
+        run: |
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+
+      - name: Create Python virtual environment
+        run: |
+          uv venv
+          uv pip install --upgrade setuptools wheel
+
+      - name: Install dependencies
+        run: |
+          source .venv/bin/activate
+          uv pip install ".[dev]"
+          uv pip install -U git+https://github.com/huggingface/accelerate.git
+          uv pip install -U git+https://github.com/huggingface/datasets.git
+          uv pip install -U git+https://github.com/huggingface/transformers.git
+          
+
+      - name: Test with pytest
+        run: |
+          source .venv/bin/activate
+          make test
+
+      - name: Post to Slack
+        if: github.ref == 'refs/heads/main' && always()  # Check if the branch is main
+        uses: huggingface/hf-workflows/.github/actions/post-slack@main
+        with:
+          slack_channel: ${{ env.CI_SLACK_CHANNEL }}
+          title: Results with Python 3.12 and dev dependencies
+          status: ${{ job.status }}
+          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
+
+  tests_wo_optional_deps:
+    name: Tests without optional dependencies
+    runs-on:
+      group: aws-g4dn-2xlarge
+    container:
+      image: pytorch/pytorch:2.6.0-cuda12.6-cudnn9-devel
+      options: --gpus all
+    defaults:
+      run:
+        shell: bash
+    if: github.event.pull_request.draft == false
+    steps:
+      - name: Git checkout
+        uses: actions/checkout@v4
+
+      - name: Set up Python 3.12
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+
+      - name: Install Make and Git
+        run: |
+          apt-get update && apt-get install -y make git curl
+
+      - name: Install uv
+        run: |
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+
+      - name: Create Python virtual environment
+        run: |
+          uv venv
+          uv pip install --upgrade setuptools wheel
+
+      - name: Install dependencies
+        run: |
+          source .venv/bin/activate
+          uv pip install ".[test]"
+
+      - name: Test with pytest
+        run: |
+          source .venv/bin/activate
+          make test
+
+      - name: Post to Slack
+        if: github.ref == 'refs/heads/main' && always()  # Check if the branch is main
+        uses: huggingface/hf-workflows/.github/actions/post-slack@main
+        with:
+          slack_channel: ${{ env.CI_SLACK_CHANNEL }}
+          title: Results with Python 3.12 without optional dependencies
+          status: ${{ job.status }}
+          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}
+
+  tests_min_versions:
+    name: Tests with minimum versions
+    runs-on:
+      group: aws-g4dn-2xlarge
+    container:
+      image: pytorch/pytorch:2.6.0-cuda12.6-cudnn9-devel
+      options: --gpus all
+    defaults:
+      run:
+        shell: bash
+    if: github.event.pull_request.draft == false
+    steps:
+      - name: Git checkout
+        uses: actions/checkout@v4
+
+      - name: Set up Python 3.12
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+
+      - name: Install Make and Git
+        run: |
+          apt-get update && apt-get install -y make git curl
+
+      - name: Install uv
+        run: |
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+
+      - name: Create Python virtual environment
+        run: |
+          uv venv
+          uv pip install --upgrade setuptools wheel
+
+      - name: Install dependencies
+        run: |
+          source .venv/bin/activate
+          uv pip install ".[dev]"
+          uv pip install accelerate==1.4.0
+          uv pip install datasets==3.0.0
+          uv pip install transformers==4.53.2
+
+      - name: Test with pytest
+        run: |
+          source .venv/bin/activate
+          make test
+
+      - name: Post to Slack
+        if: github.ref == 'refs/heads/main' && always()  # Check if the branch is main
+        uses: huggingface/hf-workflows/.github/actions/post-slack@main
+        with:
+          slack_channel: ${{ env.CI_SLACK_CHANNEL }}
+          title: Results with Python 3.12 and minimum dependencies versions
+          status: ${{ job.status }}
+          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}

.github/workflows/tests_latest.yml

@@ -0,0 +1,66 @@
+name: Tests latest TRL release with dev dependencies
+
+on:
+  schedule:
+    - cron: '0 0 * * *'  # Runs daily at midnight UTC
+
+  workflow_dispatch:
+
+env:
+  TQDM_DISABLE: 1
+  CI_SLACK_CHANNEL: ${{ secrets.CI_PUSH_MAIN_CHANNEL }}
+
+jobs:
+  tests:
+    name: Tests latest TRL release with dev dependencies
+    runs-on:
+      group: aws-g4dn-2xlarge
+    container:
+      image: pytorch/pytorch:2.6.0-cuda12.6-cudnn9-devel
+      options: --gpus all
+    defaults:
+      run:
+        shell: bash
+    steps:
+      - name: Git checkout
+        uses: actions/checkout@v4
+        with: { ref: v0.20-release }
+
+      - name: Set up Python 3.12
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+
+      - name: Install Make and Git
+        run: |
+          apt-get update && apt-get install -y make git curl
+
+      - name: Install uv
+        run: |
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+
+      - name: Create Python virtual environment
+        run: |
+          uv venv
+          uv pip install --upgrade setuptools wheel
+
+      - name: Install dependencies
+        run: |
+          source .venv/bin/activate
+          uv pip install ".[dev]"
+          uv pip install -U git+https://github.com/huggingface/accelerate.git
+          uv pip install -U git+https://github.com/huggingface/datasets.git
+          uv pip install -U git+https://github.com/huggingface/transformers.git
+
+      - name: Test with pytest
+        run: |
+          source .venv/bin/activate
+          make test
+
+      - name: Post to Slack
+        uses: huggingface/hf-workflows/.github/actions/post-slack@main
+        with:
+          slack_channel: ${{ env.CI_SLACK_CHANNEL }}
+          title: Results of latest TRL with Python 3.12 and dev dependencies
+          status: ${{ job.status }}
+          slack_token: ${{ secrets.SLACK_CIFEEDBACK_BOT_TOKEN }}

.github/workflows/trufflehog.yml

@@ -12,4 +12,7 @@ jobs:
       with:
         fetch-depth: 0
     - name: Secret Scanning
-      uses: trufflesecurity/trufflehog@main
+      uses: trufflesecurity/trufflehog@853e1e8d249fd1e29d0fcc7280d29b03df3d643d
+      with:
+        # exclude buggy postgres detector that is causing false positives and not relevant to our codebase
+        extra_args: --results=verified,unknown --exclude-detectors=postgres

.gitignore

@@ -1,4 +1,3 @@
-benchmark/trl
 *.bak
 .gitattributes
 .last_checked
@@ -143,7 +142,4 @@ checklink/cookies.txt
 # wandb files
 nbs/wandb/
 examples/notebooks/wandb/
-wandb/
-
-# cli scripts that are symlinked from `examples/scripts`
-trl/commands/scripts/
+wandb/

.pre-commit-config.yaml

@@ -1,10 +1,12 @@
 repos:
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.2.0
+    rev: v0.11.10
     hooks:
-      - id: ruff
+      - id: ruff-check
+        types_or: [ python, pyi ]
         args: [ --fix ]
       - id: ruff-format
+        types_or: [ python, pyi ]
 
   # - repo: https://github.com/codespell-project/codespell
   #   rev: v2.1.0

CITATION.cff

@@ -17,6 +17,12 @@ authors:
     family-names: Thrush
   - given-names: Nathan
     family-names: Lambert
+  - given-names: Shengyi
+    family-names: Huang
+  - given-names: Kashif
+    family-names: Rasul
+  - given-names: Quentin
+    family-names: Gallouédec
 repository-code: 'https://github.com/huggingface/trl'
 abstract: "With trl you can train transformer language models with Proximal Policy Optimization (PPO). The library is built on top of the transformers library by \U0001F917 Hugging Face. Therefore, pre-trained language models can be directly loaded via transformers. At this point, most decoder and encoder-decoder architectures are supported."
 keywords:
@@ -25,4 +31,4 @@ keywords:
   - pytorch
   - transformers
 license: Apache-2.0
-version: 0.2.1
+version: 0.20

CONTRIBUTING.md

@@ -20,10 +20,10 @@ There are several ways you can contribute to TRL:
 * Fix outstanding issues with the existing code.
 * Submit issues related to bugs or desired new features.
 * Implement trainers for new post-training algorithms.
-* Contribute to the examples or to the documentation.
+* Contribute to the examples or the documentation.
 
 If you don't know where to start, there is a special [Good First
-Issue](https://github.com/huggingface/trl/contribute) listing. It will give you a list of
+Issue](https://github.com/huggingface/trl/labels/%F0%9F%91%B6%20good%20first%20issue) listing. It will give you a list of
 open issues that are beginner-friendly and help you start contributing to open-source. The best way to do that is to open a Pull Request and link it to the issue that you'd like to work on. We try to give priority to opened PRs as we can easily track the progress of the fix, and if the contributor does not have time anymore, someone else can take the PR over.
 
 For something slightly more challenging, you can also take a look at the [Good Second Issue](https://github.com/huggingface/trl/labels/Good%20Second%20Issue) list. In general though, if you feel like you know what you're doing, go for it and we'll help you get there! 🚀
@@ -33,12 +33,12 @@ For something slightly more challenging, you can also take a look at the [Good S
 Before you start contributing make sure you have installed all the dev tools:
 
 ```bash
-make dev
+pip install -e .[dev]
 ```
 
 ## Fixing outstanding issues
 
-If you notice an issue with the existing code and have a fix in mind, feel free to [start contributing](#create-a-pull-request) and open a Pull Request!
+If you notice an issue with the existing code and have a fix in mind, feel free to [start contributing](#submitting-a-pull-request-pr) and open a Pull Request!
 
 ## Submitting a bug-related issue or feature request
 
@@ -62,7 +62,7 @@ Once you've confirmed the bug hasn't already been reported, please include the f
 To get the OS and software versions automatically, run the following command:
 
 ```bash
-transformers-cli env
+trl env
 ```
 
 ### Do you want a new feature?
@@ -74,19 +74,19 @@ If there is a new feature you'd like to see in TRL, please open an issue and des
    Whatever it is, we'd love to hear about it!
 
 2. Describe your requested feature in as much detail as possible. The more you can tell us about it, the better we'll be able to help you.
-3. Provide a *code snippet* that demonstrates the features usage.
+3. Provide a *code snippet* that demonstrates the feature's usage.
 4. If the feature is related to a paper, please include a link.
 
 If your issue is well written we're already 80% of the way there by the time you create it.
 
 ## Do you want to implement a new trainer?
 
-New post-training methods are published on a frequent basis and those which satisfy the following criteria are good candidates to be integrated in TRL:
+New post-training methods are published frequently and those that satisfy the following criteria are good candidates to be integrated into TRL:
 
-* **Simplicity:** does the new method achieve similar performance as prior methods, but with less complexity? A good example is [Direct Preference Optimization](https://arxiv.org/abs/2305.18290) (DPO), which provided a simpler and compelling alternative to RLHF methods.
-* **Efficiency:** does the new method provide a significant improvement in training efficiency? A good example is [Odds Ratio Preference Optimization](https://arxiv.org/abs/2403.07691v2), which utilises a similar objective as DPO, but requires half the GPU VRAM.
+* **Simplicity:** Does the new method achieve similar performance as prior methods, but with less complexity? A good example is Direct Preference Optimization (DPO) [[Rafailov et al, 2023]](https://huggingface.co/papers/2305.18290), which provided a simpler and compelling alternative to RLHF methods.
+* **Efficiency:** Does the new method provide a significant improvement in training efficiency? A good example is Odds Ratio Preference Optimization (ORPO) [[Hong et al, 2023]](https://huggingface.co/papers/2403.07691), which utilizes a similar objective as DPO but requires half the GPU VRAM.
 
-Methods which only provide incremental improvements at the expense of added complexity or compute costs are unlikely to be included in TRL.
+Methods that only provide incremental improvements at the expense of added complexity or compute costs are unlikely to be included in TRL.
 
 If you want to implement a trainer for a new post-training method, first open an issue and provide the following information:
 
@@ -102,7 +102,7 @@ Based on the community and maintainer feedback, the next step will be to impleme
 
 ## Do you want to add documentation?
 
-We're always looking for improvements to the documentation that make it more clear and accurate. Please let us know how the documentation can be improved, such as typos, dead links and any missing, unclear or inaccurate content.. We'll be happy to make the changes or help you make a contribution if you're interested!
+We're always looking for improvements to the documentation that make it more clear and accurate. Please let us know how the documentation can be improved, such as typos, dead links, and any missing, unclear, or inaccurate content... We'll be happy to make the changes or help you contribute if you're interested!
 
 ## Submitting a pull request (PR)
 
@@ -133,7 +133,7 @@ Follow these steps to start contributing:
 
 3. Create a new branch to hold your development changes, and do this for every new PR you work on.
 
-   Start by synchronizing your `main` branch with the `upstream/main` branch (ore details in the [GitHub Docs](https://docs.github.com/en/github/collaborating-with-issues-and-pull-requests/syncing-a-fork)):
+   Start by synchronizing your `main` branch with the `upstream/main` branch (more details in the [GitHub Docs](https://docs.github.com/en/github/collaborating-with-issues-and-pull-requests/syncing-a-fork)):
 
    ```bash
    $ git checkout main
@@ -152,7 +152,7 @@ Follow these steps to start contributing:
 4. Set up a development environment by running the following command in a conda or a virtual environment you've created for working on this library:
 
    ```bash
-   $ make dev
+   $ pip install -e .[dev]
    ```
 
    (If TRL was already installed in the virtual environment, remove
@@ -171,8 +171,7 @@ Follow these steps to start contributing:
    $ pytest tests/<TEST_TO_RUN>.py
    ```
    
-   > For the following commands leveraging the `make` utility, we recommend using the WSL system when running on
-   > Windows. More information [here](https://docs.microsoft.com/en-us/windows/wsl/about).
+   > For the following commands leveraging the `make` utility.
 
    You can also run the full suite with the following command.
 
@@ -180,18 +179,21 @@ Follow these steps to start contributing:
    $ make test
    ```
 
-   TRL relies on `ruff` to format its source code
-   consistently. After you make changes, apply automatic style corrections and code verifications
-   that can't be automated in one go with:
+    TRL relies on `ruff` for maintaining consistent code formatting across its source files. Before submitting any PR, you should apply automatic style corrections and run code verification checks.
 
-   This target is also optimized to only work with files modified by the PR you're working on.
+    We provide a `precommit` target in the `Makefile` that simplifies this process by running all required checks and optimizations on only the files modified by your PR.
 
-   If you prefer to run the checks one after the other, the following command apply the
-   style corrections:
+    To apply these checks and corrections in one step, use:
 
-   ```bash
-   $ make precommit
-   ```
+    ```bash
+    $ make precommit
+    ```
+
+    This command runs the following:
+    - Executes `pre-commit` hooks to automatically fix style issues with `ruff` and other tools.
+    - Runs additional scripts such as adding copyright information.
+
+    If you prefer to apply the style corrections separately or review them individually, the `pre-commit` hook will handle the formatting for the files in question.
 
    Once you're happy with your changes, add changed files using `git add` and
    make a commit with `git commit` to record your changes locally:
@@ -221,10 +223,7 @@ Follow these steps to start contributing:
    webpage of your fork on GitHub. Click on 'Pull request' to send your changes
    to the project maintainers for review.
 
-7. It's ok if maintainers ask you for changes. It happens to core contributors
-   too! So everyone can see the changes in the Pull request, work in your local
-   branch and push the changes to your fork. They will automatically appear in
-   the pull request.
+7. It's ok if maintainers ask you for changes. It happens to core contributors too! To ensure everyone can review your changes in the pull request, work on your local branch and push the updates to your fork. They will automatically appear in the pull request.
 
 
 ### Checklist
@@ -245,14 +244,524 @@ Follow these steps to start contributing:
 An extensive test suite is included to test the library behavior and several examples. Library tests can be found in
 the [tests folder](https://github.com/huggingface/trl/tree/main/tests).
 
-We use `pytest` in order to run the tests. From the root of the
-repository, here's how to run tests with `pytest` for the library:
+We use `pytest` to run the tests. From the root of the
+repository here's how to run tests with `pytest` for the library:
 
 ```bash
 $ python -m pytest -sv ./tests
 ```
 
-In fact, that's how `make test` is implemented (sans the `pip install` line)!
+That's how `make test` is implemented (without the `pip install` line)!
 
-You can specify a smaller set of tests in order to test only the feature
+You can specify a smaller set of tests to test only the feature
 you're working on.
+
+### Default values guidelines
+
+1. **Use defaults when appropriate**:  
+
+Provide default values unless the parameter's value varies significantly by use case. For example, datasets or models should not have defaults, but parameters like `learning_rate` should.
+
+2. **Prioritize proven defaults**:  
+
+Default values should align with those recommended in the original paper or method. Alternatives require strong evidence of superior performance in most cases.
+
+3. **Ensure safety and predictability**:  
+
+Defaults must be safe, expected and reliable. Avoid settings that could lead to surprising outcomes, such as excessive memory usage or poor performance in edge cases.
+
+4. **Balance consistency and flexibility**:  
+
+Aim for consistent defaults across similar functions or methods. However, consistency should not be preferred to point 2 or 3.
+
+5. **Opt-in for new features**:  
+
+Do not enable new features or improvements (e.g., novel loss functions) by default. Users should explicitly opt-in to use these.
+
+### Writing documentation
+
+High-quality documentation is crucial for maintaining a project that is easy to use, understand, and extend. When adding new features, ensure they are thoroughly documented to maintain consistency and clarity throughout the project.
+
+To illustrate what good documentation looks like, here’s an example of a well-documented function:
+
+````python
+def replicate_str(string: str, n: int, sep: str = " ") -> str:
+    r"""
+    Replicate a string `n` times with a separator.
+
+    Args:
+        string (`str`):
+            String to replicate.
+        n (`int`):
+            Number of times to replicate the string.
+        sep (`str`, *optional*, defaults to `" "`):
+            Separator to use between each replication.
+    
+    Returns:
+        `str`: The replicated string.
+    
+    Examples:
+    ```python
+    >>> replicate_str("hello", 3)
+    "hello hello hello"
+    >>> replicate_str("hello", 3, sep=", ")
+    "hello, hello, hello"
+    ```
+    """
+    return sep.join([string] * n)
+````
+
+* **Line Wrapping:** Applied a consistent line wrap at column 120 to improve readability.
+* **Definite Articles:** Removed definite articles where possible to streamline language. (Eg: Changed "The string to replicate" to "String to replicate")
+* **Type Annotations:**
+  * Always include type definitions, indicating if a parameter is optional and specifying the default value.
+  * Note that `Optional` means that the value can be `None`, and `*optional*` means that it is not required for the user to pass a value.
+    E.g., for arguments that can't be `None` and aren't required:
+
+    ```python
+    foo (`int`, *optional*, defaults to `4`):
+    ```
+
+    For arguments that can be `None` and are required:
+
+    ```python
+    foo (`Optional[int]`):
+    ```
+
+    for arguments that can be `None` and aren't required:
+
+    ```python
+    foo (`Optional[int]`, *optional*, defaults to `None`):
+    ```
+
+* **String Defaults:**
+  * Ensured that default string values are wrapped in double quotes:
+
+    ```python
+    defaults to `"foo"`
+    ```
+
+* **Dictionary Typing:**
+  * Replaced generic `dict` type hints with more explicit `dict[str, Any]` to clarify expected key-value pairs.
+* **Default Value Formatting:**
+  * Consistently surrounded default values with backticks for improved formatting:
+
+    ```python
+    defaults to `4`
+    ```
+
+* **Sub-sectioning:** When the number of arguments is large, consider breaking them into sub-sections for better readability.
+
+    ```python
+    def calculate_statistics(data: list[float], precision: int = 2, include_variance: bool = False) -> dict[str, float]:
+        r"""
+        Calculates basic statistics for a given dataset.
+    
+        Args:
+            > Data inputs
+    
+            data (`list[float]`):
+                A list of numerical values to analyze.
+    
+            > Configuration parameters
+    
+            precision (`int`, *optional*, defaults to `2`):
+                Number of decimal places to round the results.
+            include_variance (`bool`, *optional*, defaults to `False`):
+                Whether to include the variance of the dataset in the results.
+    
+        Returns:
+            `dict[str, float]`:
+                A dictionary containing calculated statistics such as mean, median, and optionally variance.
+        """
+        ...
+      ```
+
+### Deprecation and backward compatibility
+
+Our approach to deprecation and backward compatibility is flexible and based on the feature’s usage and impact. Each deprecation is carefully evaluated, aiming to balance innovation with user needs.
+
+When a feature or component is marked for deprecation, its use will emit a warning message. This warning will include:
+
+- **Transition Guidance**: Instructions on how to migrate to the alternative solution or replacement.
+- **Removal Version**: The target version when the feature will be removed, providing users with a clear timeframe to transition.
+
+Example:
+   
+   ```python
+   warnings.warn(
+       "The `Trainer.foo` method is deprecated and will be removed in version 0.14.0. "
+       "Please use the `Trainer.bar` class instead.",
+       FutureWarning,
+   )
+   ```
+
+The deprecation and removal schedule is based on each feature's usage and impact, with examples at two extremes:
+
+- **Experimental or Low-Use Features**: For a feature that is experimental or has limited usage, backward compatibility may not be maintained between releases. Users should therefore anticipate potential breaking changes from one version to the next.
+
+- **Widely-Used Components**: For a feature with high usage, we aim for a more gradual transition period of approximately **5 months**, generally scheduling deprecation around **5 minor releases** after the initial warning.
+
+These examples represent the two ends of a continuum. The specific timeline for each feature will be determined individually, balancing innovation with user stability needs.
+
+### Working with warnings
+
+Warnings play a critical role in guiding users toward resolving potential issues, but they should be used thoughtfully to avoid unnecessary noise. Unlike logging, which provides informational context or operational details, warnings signal conditions that require attention and action. Overusing warnings can dilute their importance, leading users to ignore them entirely.
+
+#### Definitions
+
+- **Correct**: An operation is correct if it is valid, follows the intended approach, and aligns with the current best practices or guidelines within the codebase. This is the recommended or intended way to perform the operation.
+- **Supported**: An operation is supported if it is technically valid and works within the current codebase, but it may not be the most efficient, optimal, or recommended way to perform the task. This includes deprecated features or legacy approaches that still work but may be phased out in the future.
+
+#### Choosing the right message
+
+- **Correct → No warning**:  
+   If the operation is fully valid and expected, no message should be issued. The system is working as intended, so no warning is necessary.  
+
+- **Correct but deserves attention → No warning, possibly a log message**:
+   When an operation is correct but uncommon or requires special attention, providing an informational message can be helpful. This keeps users informed without implying any issue. If available, use the logger to output this message. Example:  
+
+   ```python
+   logger.info("This is an informational message about a rare but correct operation.")
+   ```
+
+- **Correct but very likely a mistake → Warning with option to disable**:  
+   In rare cases, you may want to issue a warning for a correct operation that’s very likely a mistake. In such cases, you must provide an option to suppress the warning. This can be done with a flag in the function. Example:  
+
+   ```python
+   def my_function(foo, bar, _warn=True):
+       if foo == bar:
+           if _warn:
+               warnings.warn("foo and bar are the same, this is likely a mistake. Ignore this warning by setting `_warn=False`.")
+           # Do something
+   ```
+
+- **Supported but not correct → Warning**:  
+   If the operation is technically supported but is deprecated, suboptimal, or could cause future issues (e.g., conflicting arguments), a warning should be raised. This message should be actionable, meaning it must explain how to resolve the issue. Example:  
+
+   ```python
+   def my_function(foo, bar):
+       if foo and bar:
+           warnings.warn("Both `foo` and `bar` were provided, but only one is allowed. Ignoring `foo`. Please pass only one of these arguments.")
+           # Do something
+   ```
+
+- **Not supported → Exception**:  
+   If the operation is invalid or unsupported, raise an exception. This indicates that the operation cannot be performed and requires immediate attention. Example:  
+
+   ```python
+   def my_function(foo, bar):
+       if foo and bar:
+           raise ValueError("Both `foo` and `bar` were provided, but only one is allowed. Please pass only one of these arguments.")
+   ```
+
+By following this classification, you ensure that warnings, information, and exceptions are used appropriately, providing clear guidance to the user without cluttering the system with unnecessary messages.
+
+
+## Making a release
+
+> [!NOTE]
+> VERSION needs to be formatted following the `v{major}.{minor}.{patch}` convention. We need to follow this convention to be able to retrieve versioned scripts.
+
+#### 0. Prerequisites
+
+- Dependencies:
+   - twine: `pip install build twine`
+- Create an account in (and join the `trl` project):
+   - PyPI: https://pypi.org/
+   - Test PyPI: https://test.pypi.org/
+
+### Major/Minor Release
+
+#### 1. Ensure your local repository is up to date with the upstream repository
+
+```bash
+git checkout main
+git pull origin main
+```
+
+> [!WARNING]
+> Do not merge other pull requests into `main` until the release is done. This is to ensure that the release is stable and does not include any untested changes. Announce internally (#trl-internal) to other maintainers that you are doing a release and that they must not merge PRs until the release is done.
+
+#### 2. Create a release branch from main
+
+```bash
+git checkout -b release-v{major}.{minor}
+```
+
+#### 3. Change the version in the following files
+
+- `.github/workflows/tests_latest.yml`:
+  ```diff
+  - with: { ref: v{major}.{minor-1}-release }
+  + with: { ref: v{major}.{minor}-release }
+  ```
+- `CITATION.cff`
+  ```diff
+  - version: {major}.{minor-1}
+  + version: {major}.{minor}
+  ```    
+- `trl/__init__.py`
+  ```diff
+  - __version__ = "{major}.{minor}.0.dev0"
+  + __version__ = "{major}.{minor}.0"
+  ```
+- `setup.cfg`
+  ```diff
+  - version = {major}.{minor}.0.dev0
+  + version = {major}.{minor}.0
+  ```
+
+#### 4. Commit and push these changes
+
+```shell
+git add .github/workflows/tests_latest.yml CITATION.cff trl/__init__.py setup.cfg
+git commit -m 'Release: {major}.{minor}'
+git push origin release-v{major}.{minor}
+```
+
+#### 5. Create a pull request 
+
+from `release-v{major}.{minor}` to `main`, named `Release: v{major}.{minor}`, wait for tests to pass, and request a review.
+
+#### 6. Once the pull request is approved, merge it into `main`
+
+#### 7. Add a tag in git to mark the release
+
+```shell
+git checkout main
+git pull origin main
+git tag -a v{major}.{minor}.0 -m 'Adds tag v{major}.{minor}.0 for PyPI'
+git push origin v{major}.{minor}.0
+```
+
+#### 8. Create a branch `v{major}.{minor}-release` for future patch releases.
+
+```shell
+git checkout -b v{major}.{minor}-release
+git push origin v{major}.{minor}-release
+```
+
+This ensures that future patch releases (`v{major}.{minor}.1`, `v{major}.{minor}.2`, etc.) can be made separately from `main`.
+
+#### 9. Create the wheels for your release
+
+These are the artifacts that will be uploaded to PyPI and installed by users via `pip install trl`.
+
+Clean previous builds:
+
+```shell
+rm -rf build dist
+```
+
+At the root of your repo, run
+
+```bash
+python -m build .
+```
+
+This will create a folders named `dist` with the new versions of your package.
+
+#### 10. Upload the package to PyPI Test
+
+> [!IMPORTANT]
+> Do not skip this step. It is important to test the package before uploading it to the main PyPI server.
+
+```shell
+twine upload dist/* -r testpypi
+```
+
+Then in a fresh environment containing all dependencies you need, try to install your new package from the PyPI test server.
+
+```bash
+pip install -i https://test.pypi.org/simple/ trl
+```
+
+You might get errors for missing dependencies since the PyPI test server does not contain all packages like PyPI does. To make sure you have everything you can do:
+
+```bash
+pip install trl
+pip uninstall trl
+```
+
+(the second line will remove trl but keep all its dependencies).
+
+Also make sure you can actually use the package! Run the following line:
+
+```bash
+python -c "from trl import *"
+```
+
+along with anything that tests:
+
+- the core feature of your package
+- the new features you’re adding in the release
+
+#### 11. Publish on PyPI
+
+> [!WARNING]
+> This can't be reverted. Make sure you have tested everything before doing this step.
+
+```shell
+twine upload dist/*
+```
+
+#### 12. Create a GitHub Release
+
+1. Go to the repo’s [releases section](https://github.com/huggingface/trl/releases) on GitHub.
+2. Click **Draft a new release**.
+3. Select the `v{major}.{minor}.0` tag you just created in step 7.
+4. Add a title (`v{major}.{minor}.0`) and a short description of what’s new.
+5. Click **Publish Release**.
+
+#### 13. Bump to dev version
+
+1. Create a branch `bump-dev-version-{major}.{minor+1}` from `main` and checkout to it.
+
+   ```shell
+   git checkout -b bump-dev-version-{major}.{minor+1}
+   ```
+
+2. Change the version in the following files:
+   1. `trl/__init__.py`
+      ```diff
+      - __version__ = "{major}.{minor}.0"
+      + __version__ = "{major}.{minor+1}.0.dev0"
+      ```
+   2. `setup.cfg`
+      ```diff
+      - version = {major}.{minor}.0
+      + version = {major}.{minor+1}.0.dev0
+      ```
+
+3. Commit and push these changes
+
+   ```shell
+   git add trl/__init__.py setup.cfg
+   git commit -m '⬆️ Bump dev version'
+   git push origin bump-dev-version-{major}.{minor+1}
+   ```
+
+4. Create a pull request from `bump-dev-version-{major}.{minor+1}` to `main`, named `⬆️ Bump dev version`, and request urgent review.
+
+5. Once the pull request is approved, merge it into `main`.
+
+6. The codebase is now ready for the next development cycle, inform the team in the #trl-internal channel.
+
+
+## Making a patch release
+
+#### 1. Ensure your local repository is up to date with the upstream repository
+
+```bash
+git checkout v{major}.{minor}-release
+git pull origin main
+```
+
+#### 2. Cherry-pick the changes you want to include in the patch release
+
+```bash
+git cherry-pick <commit-hash-0>
+git cherry-pick <commit-hash-1>
+...
+```
+
+#### 3. Change the version in the following files
+
+- `trl/__init__.py`
+  ```diff
+  - __version__ = "{major}.{minor}.{patch-1}"
+  + __version__ = "{major}.{minor}.{patch}"
+  ```
+- `setup.cfg`
+  ```diff
+  - version = {major}.{minor}.{patch-1}
+  + version = {major}.{minor}.{patch}
+  ```
+
+#### 4. Commit and push these changes
+
+```shell
+git add trl/__init__.py setup.cfg
+git commit -m 'Release: {major}.{minor}.{patch}'
+git push origin v{major}.{minor}-release
+```
+
+#### 5. Wait for the CI to pass
+
+#### 6. Add a tag in git to mark the release
+
+```shell
+git tag -a v{major}.{minor}.{patch} -m 'Adds tag v{major}.{minor}.{patch} for PyPI'
+git push origin v{major}.{minor}.{patch}
+```
+
+#### 7. Create the wheels for your release
+
+These are the artifacts that will be uploaded to PyPI and installed by users via `pip install trl`.
+
+Clean previous builds:
+
+```shell
+rm -rf build dist
+```
+
+At the root of your repo, run
+
+```bash
+python -m build .
+```
+
+This will create a folders named `dist` with the new versions of your package.
+
+#### 8. Upload the package to PyPI Test
+
+> [!IMPORTANT]
+> Do not skip this step. It is important to test the package before uploading it to the main PyPI server.
+
+```shell
+twine upload dist/* -r testpypi
+```
+
+Then in a fresh environment containing all dependencies you need, try to install your new package from the PyPI test server.
+
+```bash
+pip install -i https://test.pypi.org/simple/ trl
+```
+
+You might get errors for missing dependencies since the PyPI test server does not contain all packages like PyPI does. To make sure you have everything you can do:
+
+```bash
+pip install trl
+pip uninstall trl
+```
+
+(the second line will remove trl but keep all its dependencies).
+
+Also make sure you can actually use the package! Run the following line:
+
+```bash
+python -c "from trl import *"
+```
+
+along with anything that tests:
+
+- the core feature of your package
+- the new features you’re adding in the release
+
+#### 9. Publish on PyPI
+
+> [!WARNING]
+> This can't be reverted. Make sure you have tested everything before doing this step.
+
+```shell
+twine upload dist/*
+```
+
+#### 10. Create a GitHub Release
+
+1. Go to the repo’s [releases section](https://github.com/huggingface/trl/releases) on GitHub.
+2. Click **Draft a new release**.
+3. Select the `v{major}.{minor}.{patch}` tag you just created in step 7.
+4. Add a title (`v{major}.{minor}.{patch}`) and a short description of what’s new.
+5. Click **Publish Release**.

LICENSE

@@ -186,7 +186,7 @@
       same "printed page" as the copyright notice for easier
       identification within third-party archives.
 
-   Copyright [yyyy] [name of copyright owner]
+   Copyright 2020-2025 The HuggingFace Team
 
    Licensed under the Apache License, Version 2.0 (the "License");
    you may not use this file except in compliance with the License.

MANIFEST.in

@@ -1,5 +1,6 @@
-include settings.ini
 include LICENSE
 include CONTRIBUTING.md
 include README.md
-recursive-exclude * __pycache__
+recursive-exclude * __pycache__
+include trl/templates/*.md
+include trl/accelerate_configs/*.yaml

Makefile

@@ -1,33 +1,19 @@
-.PHONY: test precommit benchmark_core benchmark_aux common_tests slow_tests test_examples tests_gpu
+.PHONY: test precommit common_tests slow_tests test_examples tests_gpu
 
 check_dirs := examples tests trl
 
 ACCELERATE_CONFIG_PATH = `pwd`/examples/accelerate_configs
 COMMAND_FILES_PATH = `pwd`/commands
 
-
-dev:
-	[ -L "$(pwd)/trl/commands/scripts" ] && unlink "$(pwd)/trl/commands/scripts" || true
-	pip install -e ".[dev]"
-	ln -s `pwd`/examples/scripts/ `pwd`/trl/commands
-
 test:
-	python -m pytest -n auto --dist=loadfile -s -v --reruns 5 --reruns-delay 1 --only-rerun '(OSError|Timeout|HTTPError.*502|HTTPError.*504||not less than or equal to 0.01)' ./tests/
+	pytest -n auto -m "not slow and not low-priority" -s -v --reruns 5 --reruns-delay 1 --only-rerun '(OSError|Timeout|HTTPError.*502|HTTPError.*504||not less than or equal to 0.01)' tests/
 
 precommit:
+	python scripts/add_copyrights.py
 	pre-commit run --all-files
 
-benchmark_core:
-	bash ./benchmark/benchmark_core.sh
-
-benchmark_aux:
-	bash ./benchmark/benchmark_aux.sh
-
-tests_gpu:
-	python -m pytest tests/test_* $(if $(IS_GITHUB_CI),--report-log "common_tests.log",)
-
 slow_tests:
-	python -m pytest tests/slow/test_* $(if $(IS_GITHUB_CI),--report-log "slow_tests.log",)
+	pytest -m "slow" tests/ $(if $(IS_GITHUB_CI),--report-log "slow_tests.log",)
 
 test_examples:
 	touch temp_results_sft_tests.txt

README.md

@@ -1,228 +1,193 @@
+# TRL - Transformer Reinforcement Learning
+
 <div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/trl_banner_dark.png">
+<img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/trl_banner_dark.png" alt="TRL Banner">
 </div>
 
-# TRL - Transformer Reinforcement Learning
-> Full stack library to fine-tune and align large language models.
+<hr> <br>
+
+<h3 align="center">
+    <p>A comprehensive library to post-train foundation models</p>
+</h3>
 
 <p align="center">
-    <a href="https://github.com/huggingface/trl/blob/main/LICENSE">
-        <img alt="License" src="https://img.shields.io/github/license/huggingface/trl.svg?color=blue">
-    </a>
-    <a href="https://huggingface.co/docs/trl/index">
-        <img alt="Documentation" src="https://img.shields.io/website/http/huggingface.co/docs/trl/index.svg?down_color=red&down_message=offline&up_message=online">
-    </a>
-    <a href="https://github.com/huggingface/trl/releases">
-        <img alt="GitHub release" src="https://img.shields.io/github/release/huggingface/trl.svg">
-    </a>
+    <a href="https://github.com/huggingface/trl/blob/main/LICENSE"><img alt="License" src="https://img.shields.io/github/license/huggingface/trl.svg?color=blue"></a>
+    <a href="https://huggingface.co/docs/trl/index"><img alt="Documentation" src="https://img.shields.io/website?label=documentation&url=https%3A%2F%2Fhuggingface.co%2Fdocs%2Ftrl%2Findex&down_color=red&down_message=offline&up_color=blue&up_message=online"></a>
+    <a href="https://github.com/huggingface/trl/releases"><img alt="GitHub release" src="https://img.shields.io/github/release/huggingface/trl.svg"></a>
+    <a href="https://huggingface.co/trl-lib"><img alt="Hugging Face Hub" src="https://img.shields.io/badge/🤗%20Hub-trl--lib-yellow"></a> 
 </p>
 
+## Overview
 
-## What is it?
-
-The `trl` library is a full stack tool to fine-tune and align transformer language and diffusion models using methods such as Supervised Fine-tuning step (SFT), Reward Modeling (RM) and the Proximal Policy Optimization (PPO) as well as Direct Preference Optimization (DPO). 
-
-The library is built on top of the [`transformers`](https://github.com/huggingface/transformers) library and thus allows to use any model architecture available there.
-
+TRL is a cutting-edge library designed for post-training foundation models using advanced techniques like Supervised Fine-Tuning (SFT), Proximal Policy Optimization (PPO), and Direct Preference Optimization (DPO). Built on top of the [🤗 Transformers](https://github.com/huggingface/transformers) ecosystem, TRL supports a variety of model architectures and modalities, and can be scaled-up across various hardware setups.
 
 ## Highlights
 
-- **`Efficient and scalable`**: 
-    - [`accelerate`](https://github.com/huggingface/accelerate) is the backbone of `trl` which allows to scale model training from a single GPU to a large scale multi-node cluster with methods such as DDP and DeepSpeed.
-    - [`PEFT`](https://github.com/huggingface/peft) is fully integrated and allows to train even the largest models on modest hardware with quantisation and methods such as LoRA or QLoRA.
-    - [`unsloth`](https://github.com/unslothai/unsloth) is also integrated and allows to significantly speed up training with dedicated kernels.
-- **`CLI`**: With the [CLI](https://huggingface.co/docs/trl/clis) you can fine-tune and chat with LLMs without writing any code using a single command and a flexible config system.
-- **`Trainers`**: The Trainer classes are an abstraction to apply many fine-tuning methods with ease such as the [`SFTTrainer`](https://huggingface.co/docs/trl/sft_trainer), [`DPOTrainer`](https://huggingface.co/docs/trl/trainer#trl.DPOTrainer), [`RewardTrainer`](https://huggingface.co/docs/trl/reward_trainer), [`PPOTrainer`](https://huggingface.co/docs/trl/trainer#trl.PPOTrainer), [`CPOTrainer`](https://huggingface.co/docs/trl/trainer#trl.CPOTrainer), and [`ORPOTrainer`](https://huggingface.co/docs/trl/trainer#trl.ORPOTrainer).
-- **`AutoModels`**: The [`AutoModelForCausalLMWithValueHead`](https://huggingface.co/docs/trl/models#trl.AutoModelForCausalLMWithValueHead) & [`AutoModelForSeq2SeqLMWithValueHead`](https://huggingface.co/docs/trl/models#trl.AutoModelForSeq2SeqLMWithValueHead) classes add an additional value head to the model which allows to train them with RL algorithms such as PPO.
-- **`Examples`**: Train GPT2 to generate positive movie reviews with a BERT sentiment classifier, full RLHF using adapters only, train GPT-j to be less toxic, [StackLlama example](https://huggingface.co/blog/stackllama), etc. following the [examples](https://github.com/huggingface/trl/tree/main/examples).
+- **Trainers**: Various fine-tuning methods are easily accessible via trainers like [`SFTTrainer`](https://huggingface.co/docs/trl/sft_trainer), [`GRPOTrainer`](https://huggingface.co/docs/trl/grpo_trainer), [`DPOTrainer`](https://huggingface.co/docs/trl/dpo_trainer), [`RewardTrainer`](https://huggingface.co/docs/trl/reward_trainer) and more.
+
+- **Efficient and scalable**: 
+    - Leverages [🤗 Accelerate](https://github.com/huggingface/accelerate) to scale from single GPU to multi-node clusters using methods like [DDP](https://pytorch.org/tutorials/intermediate/ddp_tutorial.html) and [DeepSpeed](https://github.com/deepspeedai/DeepSpeed).
+    - Full integration with [🤗 PEFT](https://github.com/huggingface/peft) enables training on large models with modest hardware via quantization and LoRA/QLoRA.
+    - Integrates [🦥 Unsloth](https://github.com/unslothai/unsloth) for accelerating training using optimized kernels.
+
+- **Command Line Interface (CLI)**: A simple interface lets you fine-tune with models without needing to write code.
 
 ## Installation
 
-### Python package
-Install the library with `pip`:
+### Python Package
+
+Install the library using `pip`:
+
 ```bash
 pip install trl
 ```
 
 ### From source
-If you want to use the latest features before an official release you can install from source:
+
+If you want to use the latest features before an official release, you can install TRL from source:
+
 ```bash
 pip install git+https://github.com/huggingface/trl.git
 ```
 
 ### Repository
+
 If you want to use the examples you can clone the repository with the following command:
+
 ```bash
 git clone https://github.com/huggingface/trl.git
 ```
 
-## Command Line Interface (CLI)
+## Quick Start
 
-You can use TRL Command Line Interface (CLI) to quickly get started with Supervised Fine-tuning (SFT), Direct Preference Optimization (DPO) and test your aligned model with the chat CLI: 
 
-**SFT:**
-
-```bash
-trl sft --model_name_or_path facebook/opt-125m --dataset_name imdb --output_dir opt-sft-imdb
-```
-
-**DPO:**
-
-```bash
-trl dpo --model_name_or_path facebook/opt-125m --dataset_name trl-internal-testing/hh-rlhf-helpful-base-trl-style --output_dir opt-sft-hh-rlhf 
-```
-
-**Chat:**
-
-```bash
-trl chat --model_name_or_path Qwen/Qwen1.5-0.5B-Chat
-```
-
-Read more about CLI in the [relevant documentation section](https://huggingface.co/docs/trl/main/en/clis) or use `--help` for more details.
-
-## How to use
-
-For more flexibility and control over the training, you can use the dedicated trainer classes to fine-tune the model in Python.
+For more flexibility and control over training, TRL provides dedicated trainer classes to post-train language models or PEFT adapters on a custom dataset. Each trainer in TRL is a light wrapper around the 🤗 Transformers trainer and natively supports distributed training methods like DDP, DeepSpeed ZeRO, and FSDP.
 
 ### `SFTTrainer`
 
-This is a basic example of how to use the `SFTTrainer` from the library. The `SFTTrainer` is a light wrapper around the `transformers` Trainer to easily fine-tune language models or adapters on a custom dataset.
+Here is a basic example of how to use the [`SFTTrainer`](https://huggingface.co/docs/trl/sft_trainer):
 
 ```python
-# imports
-from datasets import load_dataset
 from trl import SFTTrainer
+from datasets import load_dataset
 
-# get dataset
-dataset = load_dataset("imdb", split="train")
+dataset = load_dataset("trl-lib/Capybara", split="train")
 
-# get trainer
 trainer = SFTTrainer(
-    "facebook/opt-350m",
+    model="Qwen/Qwen2.5-0.5B",
     train_dataset=dataset,
-    dataset_text_field="text",
-    max_seq_length=512,
 )
+trainer.train()
+```
 
-# train
+### `GRPOTrainer`
+
+[`GRPOTrainer`](https://huggingface.co/docs/trl/grpo_trainer) implements the [Group Relative Policy Optimization (GRPO) algorithm](https://huggingface.co/papers/2402.03300) that is more memory-efficient than PPO and was used to train [Deepseek AI's R1](https://huggingface.co/deepseek-ai/DeepSeek-R1).
+
+```python
+from datasets import load_dataset
+from trl import GRPOTrainer
+
+dataset = load_dataset("trl-lib/tldr", split="train")
+
+# Dummy reward function: count the number of unique characters in the completions
+def reward_num_unique_chars(completions, **kwargs):
+    return [len(set(c)) for c in completions]
+
+trainer = GRPOTrainer(
+    model="Qwen/Qwen2-0.5B-Instruct",
+    reward_funcs=reward_num_unique_chars,
+    train_dataset=dataset,
+)
+trainer.train()
+```
+
+### `DPOTrainer`
+
+[`DPOTrainer`](https://huggingface.co/docs/trl/dpo_trainer) implements the popular [Direct Preference Optimization (DPO) algorithm](https://huggingface.co/papers/2305.18290) that was used to post-train [Llama 3](https://huggingface.co/papers/2407.21783) and many other models. Here is a basic example of how to use the `DPOTrainer`:
+
+```python
+from datasets import load_dataset
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from trl import DPOConfig, DPOTrainer
+
+model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")
+dataset = load_dataset("trl-lib/ultrafeedback_binarized", split="train")
+training_args = DPOConfig(output_dir="Qwen2.5-0.5B-DPO")
+trainer = DPOTrainer(
+    model=model,
+    args=training_args,
+    train_dataset=dataset,
+    processing_class=tokenizer
+)
 trainer.train()
 ```
 
 ### `RewardTrainer`
 
-This is a basic example of how to use the `RewardTrainer` from the library. The `RewardTrainer` is a wrapper around the `transformers` Trainer to easily fine-tune reward models or adapters on a custom preference dataset.
+Here is a basic example of how to use the [`RewardTrainer`](https://huggingface.co/docs/trl/reward_trainer):
 
 ```python
-# imports
+from trl import RewardConfig, RewardTrainer
+from datasets import load_dataset
 from transformers import AutoModelForSequenceClassification, AutoTokenizer
-from trl import RewardTrainer
 
-# load model and dataset - dataset needs to be in a specific format
-model = AutoModelForSequenceClassification.from_pretrained("gpt2", num_labels=1)
-tokenizer = AutoTokenizer.from_pretrained("gpt2")
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")
+model = AutoModelForSequenceClassification.from_pretrained(
+    "Qwen/Qwen2.5-0.5B-Instruct", num_labels=1
+)
+model.config.pad_token_id = tokenizer.pad_token_id
 
-...
+dataset = load_dataset("trl-lib/ultrafeedback_binarized", split="train")
 
-# load trainer
+training_args = RewardConfig(output_dir="Qwen2.5-0.5B-Reward", per_device_train_batch_size=2)
 trainer = RewardTrainer(
+    args=training_args,
     model=model,
-    tokenizer=tokenizer,
+    processing_class=tokenizer,
     train_dataset=dataset,
 )
-
-# train
 trainer.train()
 ```
 
-### `PPOTrainer`
+## Command Line Interface (CLI)
 
-This is a basic example of how to use the `PPOTrainer` from the library. Based on a query the language model creates a response which is then evaluated. The evaluation could be a human in the loop or another model's output.
+You can use the TRL Command Line Interface (CLI) to quickly get started with post-training methods like Supervised Fine-Tuning (SFT) or Direct Preference Optimization (DPO):
 
-```python
-# imports
-import torch
-from transformers import AutoTokenizer
-from trl import PPOTrainer, PPOConfig, AutoModelForCausalLMWithValueHead, create_reference_model
-from trl.core import respond_to_batch
+**SFT:**
 
-# get models
-model = AutoModelForCausalLMWithValueHead.from_pretrained('gpt2')
-ref_model = create_reference_model(model)
-
-tokenizer = AutoTokenizer.from_pretrained('gpt2')
-tokenizer.pad_token = tokenizer.eos_token
-
-# initialize trainer
-ppo_config = PPOConfig(batch_size=1, mini_batch_size=1)
-
-# encode a query
-query_txt = "This morning I went to the "
-query_tensor = tokenizer.encode(query_txt, return_tensors="pt")
-
-# get model response
-response_tensor  = respond_to_batch(model, query_tensor)
-
-# create a ppo trainer
-ppo_trainer = PPOTrainer(ppo_config, model, ref_model, tokenizer)
-
-# define a reward for response
-# (this could be any reward such as human feedback or output from another model)
-reward = [torch.tensor(1.0)]
-
-# train model for one step with ppo
-train_stats = ppo_trainer.step([query_tensor[0]], [response_tensor[0]], reward)
+```bash
+trl sft --model_name_or_path Qwen/Qwen2.5-0.5B \
+    --dataset_name trl-lib/Capybara \
+    --output_dir Qwen2.5-0.5B-SFT
 ```
 
-### `DPOTrainer`
+**DPO:**
 
-`DPOTrainer` is a trainer that uses [Direct Preference Optimization algorithm](https://huggingface.co/papers/2305.18290). This is a basic example of how to use the `DPOTrainer` from the library. The `DPOTrainer` is a wrapper around the `transformers` Trainer to easily fine-tune reward models or adapters on a custom preference dataset.
-
-```python
-# imports
-from transformers import AutoModelForCausalLM, AutoTokenizer
-from trl import DPOTrainer
-
-# load model and dataset - dataset needs to be in a specific format
-model = AutoModelForCausalLM.from_pretrained("gpt2")
-tokenizer = AutoTokenizer.from_pretrained("gpt2")
-
-...
-
-# load trainer
-trainer = DPOTrainer(
-    model=model,
-    tokenizer=tokenizer,
-    train_dataset=dataset,
-)
-
-# train
-trainer.train()
+```bash
+trl dpo --model_name_or_path Qwen/Qwen2.5-0.5B-Instruct \
+    --dataset_name argilla/Capybara-Preferences \
+    --output_dir Qwen2.5-0.5B-DPO 
 ```
 
+Read more about CLI in the [relevant documentation section](https://huggingface.co/docs/trl/main/en/clis) or use `--help` for more details.
+
 ## Development
 
-If you want to contribute to `trl` or customizing it to your needs make sure to read the [contribution guide](https://github.com/huggingface/trl/blob/main/CONTRIBUTING.md) and make sure you make a dev install:
+If you want to contribute to `trl` or customize it to your needs make sure to read the [contribution guide](https://github.com/huggingface/trl/blob/main/CONTRIBUTING.md) and make sure you make a dev install:
 
 ```bash
 git clone https://github.com/huggingface/trl.git
 cd trl/
-make dev
+pip install -e .[dev]
 ```
 
-## References
-
-### Proximal Policy Optimisation
-The PPO implementation largely follows the structure introduced in the paper **"Fine-Tuning Language Models from Human Preferences"** by D. Ziegler et al. \[[paper](https://huggingface.co/papers/1909.08593), [code](https://github.com/openai/lm-human-preferences)].
-
-### Direct Preference Optimization
-DPO is based on the original implementation of **"Direct Preference Optimization: Your Language Model is Secretly a Reward Model"** by E. Mitchell et al. \[[paper](https://huggingface.co/papers/2305.18290), [code](https://github.com/eric-mitchell/direct-preference-optimization)]
-
-
 ## Citation
 
 ```bibtex
 @misc{vonwerra2022trl,
-  author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang},
+  author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
   title = {TRL: Transformer Reinforcement Learning},
   year = {2020},
   publisher = {GitHub},
@@ -230,3 +195,7 @@ DPO is based on the original implementation of **"Direct Preference Optimization
   howpublished = {\url{https://github.com/huggingface/trl}}
 }
 ```
+
+## License
+
+This repository's source code is available under the [Apache-2.0 License](LICENSE).

benchmark/benchmark.py

@@ -1,150 +0,0 @@
-import argparse
-import math
-import os
-import shlex
-import subprocess
-import uuid
-from distutils.util import strtobool
-
-import requests
-
-
-def parse_args():
-    # fmt: off
-    parser = argparse.ArgumentParser()
-    parser.add_argument("--command", type=str, default="",
-        help="the command to run")
-    parser.add_argument("--num-seeds", type=int, default=3,
-        help="the number of random seeds")
-    parser.add_argument("--start-seed", type=int, default=1,
-        help="the number of the starting seed")
-    parser.add_argument("--workers", type=int, default=0,
-        help="the number of workers to run benchmark experimenets")
-    parser.add_argument("--auto-tag", type=lambda x: bool(strtobool(x)), default=True, nargs="?", const=True,
-        help="if toggled, the runs will be tagged with git tags, commit, and pull request number if possible")
-    parser.add_argument("--slurm-template-path", type=str, default=None,
-        help="the path to the slurm template file (see docs for more details)")
-    parser.add_argument("--slurm-gpus-per-task", type=int, default=1,
-        help="the number of gpus per task to use for slurm jobs")
-    parser.add_argument("--slurm-total-cpus", type=int, default=50,
-        help="the number of gpus per task to use for slurm jobs")
-    parser.add_argument("--slurm-ntasks", type=int, default=1,
-        help="the number of tasks to use for slurm jobs")
-    parser.add_argument("--slurm-nodes", type=int, default=None,
-        help="the number of nodes to use for slurm jobs")
-    args = parser.parse_args()
-    # fmt: on
-    return args
-
-
-def run_experiment(command: str):
-    command_list = shlex.split(command)
-    print(f"running {command}")
-
-    # Use subprocess.PIPE to capture the output
-    fd = subprocess.Popen(command_list, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
-    output, errors = fd.communicate()
-
-    return_code = fd.returncode
-    assert return_code == 0, f"Command failed with error: {errors.decode('utf-8')}"
-
-    # Convert bytes to string and strip leading/trailing whitespaces
-    return output.decode("utf-8").strip()
-
-
-def autotag() -> str:
-    wandb_tag = ""
-    print("autotag feature is enabled")
-    git_tag = ""
-    try:
-        git_tag = subprocess.check_output(["git", "describe", "--tags"]).decode("ascii").strip()
-        print(f"identified git tag: {git_tag}")
-    except subprocess.CalledProcessError as e:
-        print(e)
-    if len(git_tag) == 0:
-        try:
-            count = int(subprocess.check_output(["git", "rev-list", "--count", "HEAD"]).decode("ascii").strip())
-            hash = subprocess.check_output(["git", "rev-parse", "--short", "HEAD"]).decode("ascii").strip()
-            git_tag = f"no-tag-{count}-g{hash}"
-            print(f"identified git tag: {git_tag}")
-        except subprocess.CalledProcessError as e:
-            print(e)
-    wandb_tag = f"{git_tag}"
-
-    git_commit = subprocess.check_output(["git", "rev-parse", "--verify", "HEAD"]).decode("ascii").strip()
-    try:
-        # try finding the pull request number on github
-        prs = requests.get(f"https://api.github.com/search/issues?q=repo:huggingface/trl+is:pr+{git_commit}")
-        if prs.status_code == 200:
-            prs = prs.json()
-            if len(prs["items"]) > 0:
-                pr = prs["items"][0]
-                pr_number = pr["number"]
-                wandb_tag += f",pr-{pr_number}"
-        print(f"identified github pull request: {pr_number}")
-    except Exception as e:
-        print(e)
-
-    return wandb_tag
-
-
-if __name__ == "__main__":
-    args = parse_args()
-    if args.auto_tag:
-        existing_wandb_tag = os.environ.get("WANDB_TAGS", "")
-        wandb_tag = autotag()
-        if len(wandb_tag) > 0:
-            if len(existing_wandb_tag) > 0:
-                os.environ["WANDB_TAGS"] = ",".join([existing_wandb_tag, wandb_tag])
-            else:
-                os.environ["WANDB_TAGS"] = wandb_tag
-    print("WANDB_TAGS: ", os.environ.get("WANDB_TAGS", ""))
-    commands = []
-    for seed in range(0, args.num_seeds):
-        commands += [" ".join([args.command, "--seed", str(args.start_seed + seed)])]
-
-    print("======= commands to run:")
-    for command in commands:
-        print(command)
-
-    if args.workers > 0 and args.slurm_template_path is None:
-        from concurrent.futures import ThreadPoolExecutor
-
-        executor = ThreadPoolExecutor(max_workers=args.workers, thread_name_prefix="cleanrl-benchmark-worker-")
-        for command in commands:
-            executor.submit(run_experiment, command)
-        executor.shutdown(wait=True)
-    else:
-        print("not running the experiments because --workers is set to 0; just printing the commands to run")
-
-    # SLURM logic
-    if args.slurm_template_path is not None:
-        if not os.path.exists("slurm"):
-            os.makedirs("slurm")
-        if not os.path.exists("slurm/logs"):
-            os.makedirs("slurm/logs")
-        print("======= slurm commands to run:")
-        with open(args.slurm_template_path) as f:
-            slurm_template = f.read()
-        slurm_template = slurm_template.replace("{{array}}", f"0-{len(commands) - 1}%{args.workers}")
-        slurm_template = slurm_template.replace(
-            "{{seeds}}", f"({' '.join([str(args.start_seed + int(seed)) for seed in range(args.num_seeds)])})"
-        )
-        slurm_template = slurm_template.replace("{{len_seeds}}", f"{args.num_seeds}")
-        slurm_template = slurm_template.replace("{{command}}", args.command)
-        slurm_template = slurm_template.replace("{{gpus_per_task}}", f"{args.slurm_gpus_per_task}")
-        total_gpus = args.slurm_gpus_per_task * args.slurm_ntasks
-        slurm_cpus_per_gpu = math.ceil(args.slurm_total_cpus / total_gpus)
-        slurm_template = slurm_template.replace("{{cpus_per_gpu}}", f"{slurm_cpus_per_gpu}")
-        slurm_template = slurm_template.replace("{{ntasks}}", f"{args.slurm_ntasks}")
-        if args.slurm_nodes is not None:
-            slurm_template = slurm_template.replace("{{nodes}}", f"#SBATCH --nodes={args.slurm_nodes}")
-        else:
-            slurm_template = slurm_template.replace("{{nodes}}", "")
-        filename = str(uuid.uuid4())
-        open(os.path.join("slurm", f"{filename}.slurm"), "w").write(slurm_template)
-        slurm_path = os.path.join("slurm", f"{filename}.slurm")
-        print(f"saving command in {slurm_path}")
-        if args.workers > 0:
-            job_id = run_experiment(f"sbatch --parsable {slurm_path}")
-            print(f"Job ID: {job_id}")

benchmark/benchmark_and_report.sh

@@ -1,26 +0,0 @@
-export WANDB_ENTITY=huggingface
-export WANDB_PROJECT=trl
-bash $BENCHMARK_SCRIPT > output.txt
-
-# Extract Job IDs into an array
-job_ids=($(grep "Job ID:" output.txt | awk '{print $3}'))
-
-# Extract WANDB_TAGS into an array
-WANDB_TAGS=($(grep "WANDB_TAGS:" output.txt | awk '{print $2}'))
-WANDB_TAGS=($(echo $WANDB_TAGS | tr "," "\n"))
-
-# Print to verify
-echo "Job IDs: ${job_ids[@]}"
-echo "WANDB_TAGS: ${WANDB_TAGS[@]}"
-
-TAGS_STRING="?tag=${WANDB_TAGS[0]}"
-FOLDER_STRING="${WANDB_TAGS[0]}"
-for tag in "${WANDB_TAGS[@]:1}"; do
-    TAGS_STRING+="&tag=$tag"
-    FOLDER_STRING+="_$tag"
-done
-
-echo "TAGS_STRING: $TAGS_STRING"
-echo "FOLDER_STRING: $FOLDER_STRING"
-
-TAGS_STRING=$TAGS_STRING FOLDER_STRING=$FOLDER_STRING BENCHMARK_PLOT_SCRIPT=$BENCHMARK_PLOT_SCRIPT sbatch --dependency=afterany:$job_ids benchmark/post_github_comment.sbatch

benchmark/benchmark_level1.sh

@@ -1,44 +0,0 @@
-# hello world experiment
-python benchmark/benchmark.py \
-    --command "python examples/scripts/ppo.py --log_with wandb" \
-    --num-seeds 3 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 12 \
-    --slurm-template-path benchmark/trl.slurm_template
-
-python benchmark/benchmark.py \
-    --command "python examples/scripts/dpo.py --model_name_or_path=gpt2 --per_device_train_batch_size 4 --max_steps 1000 --learning_rate 1e-3 --gradient_accumulation_steps 1 --logging_steps 10 --eval_steps 500 --output_dir="dpo_anthropic_hh" --optim adamw_torch --warmup_steps 150 --report_to wandb --bf16 --logging_first_step --no_remove_unused_columns" \
-    --num-seeds 3 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 12 \
-    --slurm-template-path benchmark/trl.slurm_template
-
-python benchmark/benchmark.py \
-    --command "python examples/scripts/sft.py --model_name_or_path="facebook/opt-350m" --report_to="wandb" --learning_rate=1.41e-5 --per_device_train_batch_size=64 --gradient_accumulation_steps=16 --output_dir="sft_openassistant-guanaco" --logging_steps=1 --num_train_epochs=3 --max_steps=-1 --push_to_hub --gradient_checkpointing" \
-    --num-seeds 3 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 12 \
-    --slurm-template-path benchmark/trl.slurm_template
-
-python benchmark/benchmark.py \
-    --command "python examples/scripts/reward_modeling.py --model_name_or_path=facebook/opt-350m --output_dir="reward_modeling_anthropic_hh" --per_device_train_batch_size=64 --num_train_epochs=1 --gradient_accumulation_steps=16 --gradient_checkpointing=True --learning_rate=1.41e-5 --report_to="wandb" --remove_unused_columns=False --optim="adamw_torch" --logging_steps=10 --eval_strategy="steps" --max_length=512" \
-    --num-seeds 3 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 12 \
-    --slurm-template-path benchmark/trl.slurm_template

benchmark/benchmark_level1_plot.sh

@@ -1,50 +0,0 @@
-# pip install openrlbenchmark==0.2.1a5
-# see https://github.com/openrlbenchmark/openrlbenchmark#get-started for documentation
-echo "we deal with $TAGS_STRING"
-
-python -m openrlbenchmark.rlops_multi_metrics \
-    --filters '?we=huggingface&wpn=trl&xaxis=_step&ceik=trl_ppo_trainer_config.value.reward_model&cen=trl_ppo_trainer_config.value.exp_name&metrics=env/reward_mean&metrics=objective/kl' \
-        "ppo$TAGS_STRING" \
-    --env-ids sentiment-analysis:lvwerra/distilbert-imdb \
-    --no-check-empty-runs \
-    --pc.ncols 2 \
-    --pc.ncols-legend 1 \
-    --output-filename benchmark/trl/$FOLDER_STRING/ppo \
-    --scan-history
-
-python -m openrlbenchmark.rlops_multi_metrics \
-    --filters '?we=huggingface&wpn=trl&xaxis=_step&ceik=output_dir&cen=_name_or_path&metrics=train/rewards/accuracies&metrics=train/loss' \
-        "gpt2$TAGS_STRING" \
-    --env-ids dpo_anthropic_hh \
-    --no-check-empty-runs \
-    --pc.ncols 2 \
-    --pc.ncols-legend 1 \
-    --output-filename benchmark/trl/$FOLDER_STRING/dpo \
-    --scan-history
-
-python -m openrlbenchmark.rlops_multi_metrics \
-    --filters '?we=huggingface&wpn=trl&xaxis=_step&ceik=output_dir&cen=_name_or_path&metrics=train/loss&metrics=eval/accuracy&metrics=eval/loss' \
-        "facebook/opt-350m$TAGS_STRING" \
-    --env-ids reward_modeling_anthropic_hh \
-    --no-check-empty-runs \
-    --pc.ncols 2 \
-    --pc.ncols-legend 1 \
-    --output-filename benchmark/trl/$FOLDER_STRING/reward_modeling \
-    --scan-history
-
-python -m openrlbenchmark.rlops_multi_metrics \
-    --filters '?we=huggingface&wpn=trl&xaxis=_step&ceik=output_dir&cen=_name_or_path&metrics=train/loss' \
-        "facebook/opt-350m$TAGS_STRING" \
-    --env-ids sft_openassistant-guanaco \
-    --no-check-empty-runs \
-    --pc.ncols 2 \
-    --pc.ncols-legend 1 \
-    --output-filename benchmark/trl/$FOLDER_STRING/sft \
-    --scan-history
-
-python benchmark/upload_benchmark.py \
-    --folder_path="benchmark/trl/$FOLDER_STRING" \
-    --path_in_repo="images/benchmark/$FOLDER_STRING" \
-    --repo_id="trl-internal-testing/example-images" \
-    --repo_type="dataset"
-

benchmark/benchmark_level2.sh

@@ -1,23 +0,0 @@
-# compound experiments: gpt2xl + grad_accu
-python benchmark/benchmark.py \
-    --command "python examples/scripts/ppo.py --exp_name ppo_gpt2xl_grad_accu --model_name gpt2-xl --mini_batch_size 16 --gradient_accumulation_steps 8 --log_with wandb" \
-    --num-seeds 3 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 12 \
-    --slurm-template-path benchmark/trl.slurm_template
-
-# compound experiments: Cerebras-GPT-6.7B + deepspeed zero2 + grad_accu
-python benchmark/benchmark.py \
-    --command "accelerate launch --config_file examples/accelerate_configs/deepspeed_zero2.yaml examples/scripts/ppo.py --exp_name ppo_Cerebras-GPT-6.7B_grad_accu_deepspeed_stage2  --batch_size 32  --mini_batch_size 32 --log_with wandb --model_name cerebras/Cerebras-GPT-6.7B --reward_model sentiment-analysis:cerebras/Cerebras-GPT-6.7B" \
-    --num-seeds 3 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 8 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 90 \
-    --slurm-template-path benchmark/trl.slurm_template

benchmark/benchmark_level2_plot.sh

@@ -1,31 +0,0 @@
-# pip install openrlbenchmark==0.2.1a5
-# see https://github.com/openrlbenchmark/openrlbenchmark#get-started for documentation
-echo "we deal with $TAGS_STRING"
-
-python -m openrlbenchmark.rlops_multi_metrics \
-    --filters '?we=huggingface&wpn=trl&xaxis=_step&ceik=trl_ppo_trainer_config.value.reward_model&cen=trl_ppo_trainer_config.value.exp_name&metrics=env/reward_mean&metrics=objective/kl' \
-        "ppo$TAGS_STRING" \
-        "ppo_gpt2xl_grad_accu$TAGS_STRING" \
-    --env-ids sentiment-analysis:lvwerra/distilbert-imdb \
-    --no-check-empty-runs \
-    --pc.ncols 2 \
-    --pc.ncols-legend 1 \
-    --output-filename benchmark/trl/$FOLDER_STRING/different_models \
-    --scan-history
-
-python -m openrlbenchmark.rlops_multi_metrics \
-    --filters '?we=huggingface&wpn=trl&xaxis=_step&ceik=trl_ppo_trainer_config.value.reward_model&cen=trl_ppo_trainer_config.value.exp_name&metrics=env/reward_mean&metrics=objective/kl' \
-        "ppo_Cerebras-GPT-6.7B_grad_accu_deepspeed_stage2$TAGS_STRING" \
-    --env-ids sentiment-analysis:cerebras/Cerebras-GPT-6.7B \
-    --no-check-empty-runs \
-    --pc.ncols 2 \
-    --pc.ncols-legend 1 \
-    --output-filename benchmark/trl/$FOLDER_STRING/deepspeed \
-    --scan-history
-
-python benchmark/upload_benchmark.py \
-    --folder_path="benchmark/trl/$FOLDER_STRING" \
-    --path_in_repo="images/benchmark/$FOLDER_STRING" \
-    --repo_id="trl-internal-testing/example-images" \
-    --repo_type="dataset"
-

benchmark/benchmark_level3.sh

@@ -1,46 +0,0 @@
-## w/ and w/o gradient accumulation
-python benchmark/benchmark.py \
-    --command "python examples/scripts/ppo.py --exp_name ppo_step_grad_accu --mini_batch_size 1 --gradient_accumulation_steps 128 --log_with wandb" \
-    --num-seeds 3 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 12 \
-    --slurm-template-path benchmark/trl.slurm_template
-
-## w/ different models (gpt2, gpt2-xl, falcon, llama2)
-python benchmark/benchmark.py \
-    --command "python examples/scripts/ppo.py --exp_name ppo_gpt2 --log_with wandb" \
-    --num-seeds 3 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 12 \
-    --slurm-template-path benchmark/trl.slurm_template
-python benchmark/benchmark.py \
-    --command "python examples/scripts/ppo.py --exp_name ppo_falcon_rw_1b --model_name tiiuae/falcon-rw-1b --log_with wandb" \
-    --num-seeds 3 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 12 \
-    --slurm-template-path benchmark/trl.slurm_template
-
-
-## w/ and w/o PEFT
-python benchmark/benchmark.py \
-    --command "python examples/scripts/ppo.py --exp_name ppo_peft --use_peft --log_with wandb" \
-    --num-seeds 3 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 12 \
-    --slurm-template-path benchmark/trl.slurm_template

benchmark/plot.sh

@@ -1,56 +0,0 @@
-# pip install openrlbenchmark==0.2.1a5
-# see https://github.com/openrlbenchmark/openrlbenchmark#get-started for documentation
-BASELINE_PR_TAG=v0.4.7-55-g110e672
-BASELINE_PR_NAME=PR-662
-
-python -m openrlbenchmark.rlops_multi_metrics \
-    --filters '?we=huggingface&wpn=trl&xaxis=_step&ceik=trl_ppo_trainer_config.value.reward_model&cen=trl_ppo_trainer_config.value.exp_name&metrics=env/reward_mean&metrics=objective/kl' \
-        "sentiment_tuning?tag=$BASELINE_PR_TAG&cl=sentiment lvwerra/gpt2-imdb ($BASELINE_PR_NAME)" \
-    --env-ids sentiment-analysis:lvwerra/distilbert-imdb \
-    --no-check-empty-runs \
-    --pc.ncols 2 \
-    --pc.ncols-legend 1 \
-    --output-filename benchmark/trl/$BASELINE_PR_TAG/sentiment \
-    --scan-history
-
-python -m openrlbenchmark.rlops_multi_metrics \
-    --filters '?we=huggingface&wpn=trl&xaxis=_step&ceik=trl_ppo_trainer_config.value.reward_model&cen=trl_ppo_trainer_config.value.exp_name&metrics=env/reward_mean&metrics=objective/kl' \
-        "sentiment_tuning?tag=$BASELINE_PR_TAG&cl=sentiment lvwerra/gpt2-imdb ($BASELINE_PR_NAME)" \
-        "sentiment_tuning_step_grad_accu?tag=$BASELINE_PR_TAG&cl=sentiment lvwerra/gpt2-imdb gradient accumulation ($BASELINE_PR_NAME)" \
-    --env-ids sentiment-analysis:lvwerra/distilbert-imdb \
-    --no-check-empty-runs \
-    --pc.ncols 2 \
-    --pc.ncols-legend 1 \
-    --output-filename benchmark/trl/$BASELINE_PR_TAG/gradient_accu \
-    --scan-history
-
-python -m openrlbenchmark.rlops_multi_metrics \
-    --filters '?we=huggingface&wpn=trl&xaxis=_step&ceik=trl_ppo_trainer_config.value.reward_model&cen=trl_ppo_trainer_config.value.exp_name&metrics=env/reward_mean&metrics=objective/kl' \
-        "sentiment_tuning?tag=$BASELINE_PR_TAG&cl=sentiment lvwerra/gpt2-imdb ($BASELINE_PR_NAME)" \
-        "sentiment_tuning_gpt2?tag=$BASELINE_PR_TAG&cl=sentiment gpt2 ($BASELINE_PR_NAME)" \
-        "sentiment_tuning_falcon_rw_1b?tag=$BASELINE_PR_TAG&cl=sentiment tiiuae/falcon-rw-1b ($BASELINE_PR_NAME)" \
-        "sentiment_tuning_gpt2xl_grad_accu?tag=$BASELINE_PR_TAG&cl=sentiment gpt2xl ($BASELINE_PR_NAME)" \
-    --env-ids sentiment-analysis:lvwerra/distilbert-imdb \
-    --no-check-empty-runs \
-    --pc.ncols 2 \
-    --pc.ncols-legend 1 \
-    --output-filename benchmark/trl/$BASELINE_PR_TAG/different_models \
-    --scan-history
-
-python -m openrlbenchmark.rlops_multi_metrics \
-    --filters '?we=huggingface&wpn=trl&xaxis=_step&ceik=trl_ppo_trainer_config.value.reward_model&cen=trl_ppo_trainer_config.value.exp_name&metrics=env/reward_mean&metrics=objective/kl' \
-        "sentiment_tuning?tag=$BASELINE_PR_TAG&cl=sentiment lvwerra/gpt2-imdb ($BASELINE_PR_NAME)" \
-        "sentiment_tuning_peft?tag=$BASELINE_PR_TAG&cl=sentiment lvwerra/gpt2-imdb w/ peft ($BASELINE_PR_NAME)" \
-    --env-ids sentiment-analysis:lvwerra/distilbert-imdb \
-    --no-check-empty-runs \
-    --pc.ncols 2 \
-    --pc.ncols-legend 1 \
-    --output-filename benchmark/trl/$BASELINE_PR_TAG/peft \
-    --scan-history
-
-
-python benchmark/upload_benchmark.py \
-    --folder_path="benchmark/trl/$BASELINE_PR_TAG" \
-    --path_in_repo="images/benchmark/$BASELINE_PR_TAG" \
-    --repo_id="trl-internal-testing/example-images" \
-    --repo_type="dataset"

benchmark/post_github_comment.py

@@ -1,26 +0,0 @@
-import json
-import os
-
-from ghapi.all import GhApi
-
-
-FOLDER_STRING = os.environ.get("FOLDER_STRING", "")
-folder = f"benchmark/trl/{FOLDER_STRING}"
-host_url = f"https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/benchmark/{FOLDER_STRING}"
-
-# Create a GitHub API instance
-github_context = json.loads(os.environ["GITHUB_CONTEXT"])
-token = os.environ["PERSONAL_ACCESS_TOKEN_GITHUB"]  # this needs to refreshed every 12 months
-status_message = "**[COSTA BENCHMARK BOT]**: Here are the results"
-body = status_message
-repo = github_context["repository"]
-owner, repo = repo.split("/")
-api = GhApi(owner=owner, repo=repo, token=token)
-
-# for each `.png` file in the folder, add it to the comment
-for file in os.listdir(folder):
-    if file.endswith(".png"):
-        body += f"\n![{file}]({host_url}/{file})"
-
-# Create a comment on the issue
-api.issues.create_comment(issue_number=github_context["event"]["issue"]["number"], body=body)

benchmark/post_github_comment.sbatch

@@ -1,9 +0,0 @@
-#!/bin/bash
-#SBATCH --job-name=trl
-#SBATCH --partition=hopper-cpu
-#SBATCH --ntasks=1
-#SBATCH --output=slurm/logs/%x_%j.out
-
-sleep 2m
-bash $BENCHMARK_PLOT_SCRIPT
-srun python benchmark/post_github_comment.py

benchmark/regression_test.sh

@@ -1,3 +0,0 @@
-BENCHMARK_SCRIPT="benchmark/benchmark_level1.sh" \
-BENCHMARK_PLOT_SCRIPT="benchmark/benchmark_level1_plot.sh" \
-bash benchmark/benchmark_and_report.sh

benchmark/trl.slurm_template

@@ -1,19 +0,0 @@
-#!/bin/bash
-#SBATCH --job-name=trl
-#SBATCH --partition=hopper-prod
-#SBATCH --gpus-per-task={{gpus_per_task}}
-#SBATCH --cpus-per-gpu={{cpus_per_gpu}}
-#SBATCH --ntasks={{ntasks}}
-#SBATCH --output=slurm/logs/%x_%j.out
-#SBATCH --array={{array}}
-##SBATCH --exclude=ip-26-0-149-199
-
-module load cuda/12.1
-
-{{nodes}}
-
-seeds={{seeds}}
-seed=${seeds[$SLURM_ARRAY_TASK_ID % {{len_seeds}}]}
-
-echo "Running task $SLURM_ARRAY_TASK_ID with seed: $seed"
-srun {{command}} --seed $seed

benchmark/upload_benchmark.py

@@ -1,23 +0,0 @@
-from dataclasses import dataclass
-
-import tyro
-from huggingface_hub import HfApi
-
-
-@dataclass
-class Args:
-    folder_path: str = "benchmark/trl"
-    path_in_repo: str = "images/benchmark"
-    repo_id: str = "trl-internal-testing/example-images"
-    repo_type: str = "dataset"
-
-
-args = tyro.cli(Args)
-api = HfApi()
-
-api.upload_folder(
-    folder_path=args.folder_path,
-    path_in_repo=args.path_in_repo,
-    repo_id=args.repo_id,
-    repo_type=args.repo_type,
-)

commands/run_dpo.sh

@@ -2,7 +2,7 @@
 # This script runs an SFT example end-to-end on a tiny model using different possible configurations
 # but defaults to QLoRA + PEFT
 OUTPUT_DIR="test_dpo/"
-MODEL_NAME="trl-internal-testing/tiny-random-LlamaForCausalLM"
+MODEL_NAME="trl-internal-testing/tiny-Qwen2ForCausalLM-2.5"
 DATASET_NAME="trl-internal-testing/hh-rlhf-helpful-base-trl-style"
 MAX_STEPS=5
 BATCH_SIZE=2
@@ -35,7 +35,7 @@ CMD="""
 accelerate launch $EXTRA_ACCELERATE_ARGS \
     --num_processes $NUM_GPUS \
     --mixed_precision 'fp16' \
-    `pwd`/examples/scripts/dpo.py \
+    `pwd`/trl/scripts/dpo.py \
     --model_name_or_path $MODEL_NAME \
     --dataset_name $DATASET_NAME \
     --output_dir $OUTPUT_DIR \

commands/run_sft.sh

@@ -2,8 +2,8 @@
 # This script runs an SFT example end-to-end on a tiny model using different possible configurations
 # but defaults to QLoRA + PEFT
 OUTPUT_DIR="test_sft/"
-MODEL_NAME="trl-internal-testing/tiny-random-LlamaForCausalLM"
-DATASET_NAME="imdb"
+MODEL_NAME="trl-internal-testing/tiny-Qwen2ForCausalLM-2.5"
+DATASET_NAME="stanfordnlp/imdb"
 MAX_STEPS=5
 BATCH_SIZE=2
 SEQ_LEN=128
@@ -36,14 +36,13 @@ CMD="""
 accelerate launch $EXTRA_ACCELERATE_ARGS \
     --num_processes $NUM_GPUS \
     --mixed_precision 'fp16' \
-    `pwd`/examples/scripts/sft.py \
+    `pwd`/trl/scripts/sft.py \
     --model_name $MODEL_NAME \
     --dataset_name $DATASET_NAME \
     --output_dir $OUTPUT_DIR \
     --max_steps $MAX_STEPS \
-    --dataset_text_field 'text' \
     --per_device_train_batch_size $BATCH_SIZE \
-    --max_seq_length $SEQ_LEN \
+    --max_length $SEQ_LEN \
     $EXTRA_TRAINING_ARGS
 """
 

docs/source/_toctree.yml

@@ -1,76 +1,118 @@
 - sections:
   - local: index
     title: TRL
-  - local: quickstart
-    title: Quickstart
   - local: installation
     title: Installation
-  - local: clis
-    title: Get started with Command Line Interfaces (CLIs)
+  - local: quickstart
+    title: Quickstart
+  title: Getting started
+- sections:
+  - local: dataset_formats
+    title: Dataset Formats
+  - local: paper_index
+    title: Paper Index
   - local: how_to_train
-    title: PPO Training FAQ
-  - local: use_model
-    title: Use Trained Models
-  - local: customization
-    title: Customize the Training
+    title: Training FAQ
   - local: logging
     title: Understanding Logs
-  title: Get started
+  title: Conceptual Guides
 - sections:
-  - local: models
-    title: Model Classes
-  - local: trainer
-    title: Trainer Classes
-  - local: reward_trainer
-    title: Reward Model Training
-  - local: sft_trainer
-    title: Supervised Fine-Tuning
-  - local: ppo_trainer
-    title: PPO Trainer
-  - local: ppov2_trainer
-    title: PPOv2 Trainer
-  - local: rloo_trainer
-    title: RLOO Trainer
-  - local: best_of_n
-    title: Best of N Sampling
-  - local: dpo_trainer
-    title: DPO Trainer
-  - local: online_dpo_trainer
-    title: Online DPO Trainer
-  - local: kto_trainer
-    title: KTO Trainer
-  - local: bco_trainer
-    title: BCO Trainer
-  - local: cpo_trainer
-    title: CPO Trainer
-  - local: ddpo_trainer
-    title: Denoising Diffusion Policy Optimization
-  - local: alignprop_trainer
-    title: AlignProp Trainer
-  - local: orpo_trainer
-    title: ORPO Trainer
-  - local: iterative_sft_trainer
-    title: Iterative Supervised Fine-Tuning
-  - local: callbacks
-    title: Callback Classes
-  - local: judges
-    title: Judge Classes
-  - local: text_environments
-    title: Text Environments
-  title: API
+  - local: clis
+    title: Command Line Interface (CLI)
+  - local: customization
+    title: Customizing the Training
+  - local: reducing_memory_usage
+    title: Reducing Memory Usage
+  - local: speeding_up_training
+    title: Speeding Up Training
+  - local: distributing_training
+    title: Distributing Training
+  - local: use_model
+    title: Using Trained Models
+  title: How-to guides
+- sections:
+  - local: deepspeed_integration
+    title: DeepSpeed
+  - local: liger_kernel_integration
+    title: Liger Kernel
+  - local: peft_integration
+    title: PEFT
+  - local: unsloth_integration
+    title: Unsloth
+  - local: vllm_integration
+    title: vLLM
+  title: Integrations
 - sections:
   - local: example_overview
     title: Example Overview
+  - local: community_tutorials
+    title: Community Tutorials
   - local: sentiment_tuning
     title: Sentiment Tuning
-  - local: lora_tuning_peft
-    title: Training with PEFT
-  - local: detoxifying_a_lm
-    title: Detoxifying a Language Model
   - local: using_llama_models
     title: Training StackLlama
-  - local: learning_tools
-    title: Learning to Use Tools
+  - local: detoxifying_a_lm
+    title: Detoxifying a Language Model
   - local: multi_adapter_rl
     title: Multi Adapter RLHF
+  - local: training_vlm_sft
+    title: Fine-tuning a Multimodal Model Using SFT (Single or Multi-Image Dataset)
   title: Examples
+- sections:
+  - sections: # Sorted alphabetically
+    - local: alignprop_trainer
+      title: AlignProp
+    - local: bco_trainer
+      title: BCO
+    - local: cpo_trainer
+      title: CPO
+    - local: ddpo_trainer
+      title: DDPO
+    - local: dpo_trainer
+      title: DPO
+    - local: online_dpo_trainer
+      title: Online DPO
+    - local: gkd_trainer
+      title: GKD
+    - local: grpo_trainer
+      title: GRPO
+    - local: kto_trainer
+      title: KTO
+    - local: nash_md_trainer
+      title: Nash-MD
+    - local: orpo_trainer
+      title: ORPO
+    - local: ppo_trainer
+      title: PPO
+    - local: prm_trainer
+      title: PRM
+    - local: reward_trainer
+      title: Reward
+    - local: rloo_trainer
+      title: RLOO
+    - local: sft_trainer
+      title: SFT
+    - local: iterative_sft_trainer
+      title: Iterative SFT
+    - local: xpo_trainer
+      title: XPO
+    title: Trainers
+  - local: models
+    title: Model Classes
+  - local: model_utils
+    title: Model Utilities
+  - local: best_of_n
+    title: Best of N Sampling
+  - local: judges
+    title: Judges
+  - local: callbacks
+    title: Callbacks
+  - local: data_utils
+    title: Data Utilities
+  - local: rewards
+    title: Reward Functions
+  - local: script_utils
+    title: Script Utilities
+  - local: others
+    title: Others
+  title: API

docs/source/alignprop_trainer.md

@@ -1,11 +1,13 @@
 # Aligning Text-to-Image Diffusion Models with Reward Backpropagation
 
+[![](https://img.shields.io/badge/All_models-AlignProp-blue)](https://huggingface.co/models?other=alignprop,trl)
+
 ## The why
 
 If your reward function is differentiable, directly backpropagating gradients from the reward models to the diffusion model is significantly more sample and compute efficient (25x) than doing policy gradient algorithm like DDPO.
 AlignProp does full backpropagation through time, which allows updating the earlier steps of denoising via reward backpropagation.
 
-<div style="text-align: center"><img src="https://align-prop.github.io/reward_tuning.png"/></div>
+<div style="text-align: center"><img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/reward_tuning.png"/></div>
 
 
 ## Getting started with `examples/scripts/alignprop.py`
@@ -14,7 +16,7 @@ The `alignprop.py` script is a working example of using the `AlignProp` trainer
 
 **Note:** one A100 GPU is recommended to get this running. For lower memory setting, consider setting truncated_backprop_rand to False. With default settings this will do truncated backpropagation with K=1.
 
-Almost every configuration parameter has a default. There is only one commandline flag argument that is required of the user to get things up and running. The user is expected to have a [huggingface user access token](https://huggingface.co/docs/hub/security-tokens) that will be used to upload the model post finetuning to HuggingFace hub. The following bash command is to be entered to get things running
+Almost every configuration parameter has a default. There is only one commandline flag argument that is required of the user to get things up and running. The user is expected to have a [huggingface user access token](https://huggingface.co/docs/hub/security-tokens) that will be used to upload the model post-finetuning to HuggingFace hub. The following bash command is to be entered to get things running
 
 ```batch
 python alignprop.py --hf_user_access_token <token>
@@ -24,7 +26,7 @@ To obtain the documentation of `stable_diffusion_tuning.py`, please run `python
 
 The following are things to keep in mind (The code checks this for you as well) in general while configuring the trainer (beyond the use case of using the example script)
 
-- The configurable randomized truncation range (`--alignprop_config.truncated_rand_backprop_minmax=(0,50)`) the first number should be equal and greater to 0, while the second number should equal or less to the number of diffusion timesteps (sample_num_steps)
+- The configurable randomized truncation range (`--alignprop_config.truncated_rand_backprop_minmax=(0,50)`) the first number should be equal and greater than 0, while the second number should equal or less to the number of diffusion timesteps (sample_num_steps)
 - The configurable truncation backprop absolute step (`--alignprop_config.truncated_backprop_timestep=49`) the number should be less than the number of diffusion timesteps (sample_num_steps), it only matters when truncated_backprop_rand is set to False
 
 ## Setting up the image logging hook function

docs/source/bco_trainer.md

@@ -1,54 +1,15 @@
 # BCO Trainer
 
+[![](https://img.shields.io/badge/All_models-BCO-blue)](https://huggingface.co/models?other=bco,trl)
+
 TRL supports the Binary Classifier Optimization (BCO).
 The [BCO](https://huggingface.co/papers/2404.04656) authors train a binary classifier whose logit serves as a reward so that the classifier maps {prompt, chosen completion} pairs to 1 and {prompt, rejected completion} pairs to 0.
 For a full example have a look at  [`examples/scripts/bco.py`].
 
-## Expected dataset format
-
-The BCO trainer expects a very specific format for the dataset as it does not require pairwise preferences. Since the model will be trained to directly optimize examples that consist of a prompt, model completion, and a label to indicate whether the completion is "good" or "bad", we expect a dataset with the following columns:
-
-- `prompt`
-- `completion`
-- `label`
-
-for example:
-
-```
-bco_dataset_dict = {
-    "prompt": [
-        "Hey, hello",
-        "How are you",
-        "What is your name?",
-        "What is your name?",
-        "Which is the best programming language?",
-        "Which is the best programming language?",
-        "Which is the best programming language?",
-    ],
-    "completion": [
-        "hi nice to meet you",
-        "leave me alone",
-        "I don't have a name",
-        "My name is Mary",
-        "Python",
-        "C++",
-        "Java",
-    ],
-    "label": [
-        True,
-        False,
-        False,
-        True,
-        True,
-        False,
-        False,
-    ],
-}
-```
-
-where the `prompt` contains the context inputs, `completion` contains the corresponding responses and `label` contains the corresponding flag that indicates if the generated completion is desired (`True`) or undesired (`False`).
-A prompt can have multiple responses and this is reflected in the entries being repeated in the dictionary's value arrays. It is required that the dataset contains at least one desirable and one undesirable completion.
+## Expected dataset type
 
+The [`BCOTrainer`] requires an [unpaired preference dataset](dataset_formats#unpaired-preference).
+The [`BCOTrainer`] supports both [conversational](dataset_formats#conversational) and [standard](dataset_formats#standard) dataset format. When provided with a conversational dataset, the trainer will automatically apply the chat template to the dataset.
 
 ## Expected model format
 The BCO trainer expects a model of `AutoModelForCausalLM`, compared to PPO that expects `AutoModelForCausalLMWithValueHead` for the value function.
@@ -71,7 +32,7 @@ bco_trainer = BCOTrainer(
     model_ref,
     args=training_args,
     train_dataset=train_dataset,
-    tokenizer=tokenizer,
+    processing_class=tokenizer,
 )
 ```
 After this one can then call:
@@ -101,7 +62,7 @@ embedding_model = Accelerator().prepare_model(self.embedding_model)
 embedding_func = partial(embed_prompt, model=embedding_model)
 ```
 
-Set `prompt_sample_size` to defined how many prompts are selected to train the UDM classifier and start the training with the provided embedding function:
+Set `prompt_sample_size` to define how many prompts are selected to train the UDM classifier and start the training with the provided embedding function:
 
 ```py
 training_args = BCOConfig(
@@ -114,7 +75,7 @@ bco_trainer = BCOTrainer(
     model_ref,
     args=training_args,
     train_dataset=train_dataset,
-    tokenizer=tokenizer,
+    processing_class=tokenizer,
     embedding_func=embedding_func,
     embedding_tokenizer=self.embedding_tokenizer,
 )
@@ -136,4 +97,4 @@ To scale how much the auxiliary loss contributes to the total loss, use the hype
 
 ## BCOConfig
 
-[[autodoc]] BCOConfig
+[[autodoc]] BCOConfig

docs/source/best_of_n.md

@@ -67,6 +67,6 @@ best_of_n.generate(query_tensors, device=device)
 
 ```
 
-Furthermore, at the time of initialization you can set the seed to control repeatability of the generation process and the number of samples to generate for each query
+Furthermore, at the time of initialization you can set the seed to control the repeatability of the generation process and the number of samples to generate for each query
 
 

docs/source/callbacks.md

@@ -11,3 +11,11 @@
 ## WinRateCallback
 
 [[autodoc]] WinRateCallback
+
+## LogCompletionsCallback
+
+[[autodoc]] LogCompletionsCallback
+
+## MergeModelCallback
+
+[[autodoc]] MergeModelCallback

docs/source/clis.md

@@ -0,0 +1,272 @@
+# Command Line Interfaces (CLIs)
+
+TRL provides a powerful command-line interface (CLI) to fine-tune large language models (LLMs) using methods like Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and more. The CLI abstracts away much of the boilerplate, letting you launch training jobs quickly and reproducibly.
+
+Currently supported commands are:
+
+#### Training Commands
+
+- `trl dpo`: fine-tune a LLM with DPO
+- `trl grpo`: fine-tune a LLM with GRPO
+- `trl kto`: fine-tune a LLM with KTO
+- `trl sft`: fine-tune a LLM with SFT
+
+#### Other Commands
+
+- `trl env`: get the system information
+- `trl vllm-serve`: serve a model with vLLM
+
+## Fine-Tuning with the TRL CLI
+
+### Basic Usage
+
+You can launch training directly from the CLI by specifying required arguments like the model and dataset:
+
+<hfoptions id="command_line">
+<hfoption id="SFT">
+
+```bash
+trl sft \
+  --model_name_or_path Qwen/Qwen2.5-0.5B \
+  --dataset_name stanfordnlp/imdb
+```
+
+</hfoption>
+<hfoption id="DPO">
+
+```bash
+trl dpo \
+  --model_name_or_path Qwen/Qwen2.5-0.5B \
+  --dataset_name anthropic/hh-rlhf
+```
+
+</hfoption>
+</hfoptions>
+
+### Using Configuration Files
+
+To keep your CLI commands clean and reproducible, you can define all training arguments in a YAML configuration file:
+
+<hfoptions id="config_file">
+<hfoption id="SFT">
+
+```yaml
+# sft_config.yaml
+model_name_or_path: Qwen/Qwen2.5-0.5B
+dataset_name: stanfordnlp/imdb
+```
+
+Launch with:
+
+```bash
+trl sft --config sft_config.yaml
+```
+
+</hfoption>
+<hfoption id="DPO">
+
+```yaml
+# dpo_config.yaml
+model_name_or_path: Qwen/Qwen2.5-0.5B
+dataset_name: anthropic/hh-rlhf
+```
+
+Launch with:
+
+```bash
+trl dpo --config dpo_config.yaml
+```
+
+</hfoption>
+</hfoptions>
+
+### Scaling Up with Accelerate
+
+TRL CLI natively supports [🤗 Accelerate](https://huggingface.co/docs/accelerate), making it easy to scale training across multiple GPUs, machines, or use advanced setups like DeepSpeed — all from the same CLI.
+
+You can pass any `accelerate launch` arguments directly to `trl`, such as `--num_processes`. For more information see [Using accelerate launch](https://huggingface.co/docs/accelerate/en/basic_tutorials/launch#using-accelerate-launch).
+
+<hfoptions id="launch_args">
+<hfoption id="SFT inline">
+
+```bash
+trl sft \
+  --model_name_or_path Qwen/Qwen2.5-0.5B \
+  --dataset_name stanfordnlp/imdb \
+  --num_processes 4
+```
+
+</hfoption>
+<hfoption id="SFT w/ config file">
+
+```yaml
+# sft_config.yaml
+model_name_or_path: Qwen/Qwen2.5-0.5B
+dataset_name: stanfordnlp/imdb
+num_processes: 4
+```
+
+Launch with:
+
+```bash
+trl sft --config sft_config.yaml
+```
+
+</hfoption>
+<hfoption id="DPO inline">
+
+```bash
+trl dpo \
+  --model_name_or_path Qwen/Qwen2.5-0.5B \
+  --dataset_name anthropic/hh-rlhf \
+  --num_processes 4
+```
+
+</hfoption>
+<hfoption id="DPO w/ config file">
+
+```yaml
+# dpo_config.yaml
+model_name_or_path: Qwen/Qwen2.5-0.5B
+dataset_name: anthropic/hh-rlhf
+num_processes: 4
+```
+
+Launch with:
+
+```bash
+trl dpo --config dpo_config.yaml
+```
+</hfoption>
+</hfoptions>
+
+### Using `--accelerate_config` for Accelerate Configuration
+
+The `--accelerate_config` flag lets you easily configure distributed training with [🤗 Accelerate](https://github.com/huggingface/accelerate). This flag accepts either:
+
+* the name of a predefined config profile (built into TRL), or
+* a path to a custom Accelerate YAML config file.
+
+#### Predefined Config Profiles
+
+TRL provides several ready-to-use Accelerate configs to simplify common training setups:
+
+| Name         | Description                         |
+| ------------ | ----------------------------------- |
+| `fsdp1`      | Fully Sharded Data Parallel Stage 1 |
+| `fsdp2`      | Fully Sharded Data Parallel Stage 2 |
+| `zero1`      | DeepSpeed ZeRO Stage 1              |
+| `zero2`      | DeepSpeed ZeRO Stage 2              |
+| `zero3`      | DeepSpeed ZeRO Stage 3              |
+| `multi_gpu`  | Multi-GPU training                  |
+| `single_gpu` | Single-GPU training                 |
+
+To use one of these, just pass the name to `--accelerate_config`. TRL will automatically load the corresponding config file from `trl/accelerate_config/`.
+
+#### Example Usage
+
+<hfoptions id="accelerate_config">
+<hfoption id="SFT inline">
+
+```bash
+trl sft \
+  --model_name_or_path Qwen/Qwen2.5-0.5B \
+  --dataset_name stanfordnlp/imdb \
+  --accelerate_config zero2  # or path/to/my/accelerate/config.yaml
+```
+
+</hfoption>
+<hfoption id="SFT w/ config file">
+
+```yaml
+# sft_config.yaml
+model_name_or_path: Qwen/Qwen2.5-0.5B
+dataset_name: stanfordnlp/imdb
+accelerate_config: zero2  # or path/to/my/accelerate/config.yaml
+```
+
+Launch with:
+
+```bash
+trl sft --config sft_config.yaml
+```
+
+</hfoption>
+<hfoption id="DPO inline">
+
+```bash
+trl dpo \
+  --model_name_or_path Qwen/Qwen2.5-0.5B \
+  --dataset_name anthropic/hh-rlhf \
+  --accelerate_config zero2  # or path/to/my/accelerate/config.yaml
+```
+
+</hfoption>
+<hfoption id="DPO w/ config file">
+
+```yaml
+# dpo_config.yaml
+model_name_or_path: Qwen/Qwen2.5-0.5B
+dataset_name: anthropic/hh-rlhf
+accelerate_config: zero2  # or path/to/my/accelerate/config.yaml
+```
+
+Launch with:
+
+```bash
+trl dpo --config dpo_config.yaml
+```
+</hfoption>
+</hfoptions>
+
+## Getting the System Information
+
+You can get the system information by running the following command:
+
+```bash
+trl env
+```
+
+This will print out the system information, including the GPU information, the CUDA version, the PyTorch version, the transformers version, the TRL version, and any optional dependencies that are installed.
+
+```txt
+Copy-paste the following information when reporting an issue:
+
+- Platform: Linux-5.15.0-1048-aws-x86_64-with-glibc2.31
+- Python version: 3.11.9
+- PyTorch version: 2.4.1
+- accelerator(s): NVIDIA H100 80GB HBM3
+- Transformers version: 4.45.0.dev0
+- Accelerate version: 0.34.2
+- Accelerate config: 
+  - compute_environment: LOCAL_MACHINE
+  - distributed_type: DEEPSPEED
+  - mixed_precision: no
+  - use_cpu: False
+  - debug: False
+  - num_processes: 4
+  - machine_rank: 0
+  - num_machines: 1
+  - rdzv_backend: static
+  - same_network: True
+  - main_training_function: main
+  - enable_cpu_affinity: False
+  - deepspeed_config: {'gradient_accumulation_steps': 4, 'offload_optimizer_device': 'none', 'offload_param_device': 'none', 'zero3_init_flag': False, 'zero_stage': 2}
+  - downcast_bf16: no
+  - tpu_use_cluster: False
+  - tpu_use_sudo: False
+  - tpu_env: []
+- Datasets version: 3.0.0
+- HF Hub version: 0.24.7
+- TRL version: 0.12.0.dev0+acb4d70
+- bitsandbytes version: 0.41.1
+- DeepSpeed version: 0.15.1
+- Diffusers version: 0.30.3
+- Liger-Kernel version: 0.3.0
+- LLM-Blender version: 0.0.2
+- OpenAI version: 1.46.0
+- PEFT version: 0.12.0
+- vLLM version: not installed
+```
+
+This information is required when reporting an issue.

docs/source/clis.mdx

@@ -1,119 +0,0 @@
-# Command Line Interfaces (CLIs)
-
-You can use TRL to fine-tune your Language Model with Supervised Fine-Tuning (SFT) or Direct Policy Optimization (DPO) or even chat with your model using the TRL CLIs.
-
-Currently supported CLIs are:
-
-- `trl sft`: fine-tune a LLM on a text/instruction dataset
-- `trl dpo`: fine-tune a LLM with DPO on a preference dataset 
-- `trl chat`: quickly spin up a LLM fine-tuned for chatting
-
-## Fine-tuning with the CLI
-
-Before getting started, pick up a Language Model from Hugging Face Hub. Supported models can be found with the filter "text-generation" within models. Also make sure to pick up a relevant dataset for your task.
-
-Before using the `sft` or `dpo` commands make sure to run:
-```bash
-accelerate config
-```
-and pick up the right configuration for your training setup (single / multi-GPU, DeepSpeed, etc.). Make sure to complete all steps of `accelerate config` before running any CLI command.
-
-We also recommend you passing a YAML config file to configure your training protocol. Below is a simple example of a YAML file that you can use for training your models with `trl sft` command.
-
-```yaml
-model_name_or_path:
-  trl-internal-testing/tiny-random-LlamaForCausalLM
-dataset_name:
-  imdb
-dataset_text_field:
-  text
-report_to:
-  none
-learning_rate:
-  0.0001
-lr_scheduler_type:
-  cosine
-```
-
-Save that config in a `.yaml` and get started immediately! An example CLI config is available as `examples/cli_configs/example_config.yaml`. Note you can overwrite the arguments from the config file by explicitly passing them to the CLI, e.g. from the root folder:
-
-```bash
-trl sft --config examples/cli_configs/example_config.yaml --output_dir test-trl-cli --lr_scheduler_type cosine_with_restarts
-```
-
-Will force-use `cosine_with_restarts` for `lr_scheduler_type`.
-
-### Supported Arguments 
-
-We do support all arguments from `transformers.TrainingArguments`, for loading your model, we support all arguments from `~trl.ModelConfig`:
-
-[[autodoc]] ModelConfig
-
-You can pass any of these arguments either to the CLI or the YAML file.
-
-### Supervised Fine-tuning (SFT)
-
-Follow the basic instructions above and run `trl sft --output_dir <output_dir> <*args>`: 
-
-```bash
-trl sft --model_name_or_path facebook/opt-125m --dataset_name imdb --output_dir opt-sft-imdb
-```
-
-The SFT CLI is based on the `examples/scripts/sft.py` script.
-
-### Direct Policy Optimization (DPO)
-
-To use the DPO CLI, you need to have a dataset in the TRL format such as 
-
-* TRL's Anthropic HH dataset: https://huggingface.co/datasets/trl-internal-testing/hh-rlhf-helpful-base-trl-style
-* TRL's OpenAI TL;DR summarization dataset: https://huggingface.co/datasets/trl-internal-testing/tldr-preference-trl-style
-
-These datasets always have at least three columns `prompt, chosen, rejected`:
-
-* `prompt` is a list of strings.
-* `chosen` is the chosen response in [chat format](https://huggingface.co/docs/transformers/main/en/chat_templating)
-* `rejected` is the rejected response [chat format](https://huggingface.co/docs/transformers/main/en/chat_templating) 
-
-
-To do a quick start, you can run the following command:
-
-```bash
-trl dpo --model_name_or_path facebook/opt-125m --output_dir trl-hh-rlhf --dataset_name trl-internal-testing/hh-rlhf-helpful-base-trl-style
-```
-
-
-The DPO CLI is based on the `examples/scripts/dpo.py` script.
-
-
-#### Custom preference dataset
-
-Format the dataset into TRL format (you can adapt the `examples/datasets/anthropic_hh.py`):
-
-```bash
-python examples/datasets/anthropic_hh.py --push_to_hub --hf_entity your-hf-org
-```
-
-## Chat interface
-
-The chat CLI lets you quickly load the model and talk to it. Simply run the following:
-
-```bash
-trl chat --model_name_or_path  Qwen/Qwen1.5-0.5B-Chat 
-```
-
-> [!TIP]
-> To use the chat CLI with the developer installation, you must run `make dev` 
->
-
-Note that the chat interface relies on the tokenizer's [chat template](https://huggingface.co/docs/transformers/chat_templating) to format the inputs for the model. Make sure your tokenizer has a chat template defined.
-
-Besides talking to the model there are a few commands you can use:
-
-- **clear**: clears the current conversation and start a new one
-- **example {NAME}**: load example named `{NAME}` from the config and use it as the user input
-- **set {SETTING_NAME}={SETTING_VALUE};**: change the system prompt or generation settings (multiple settings are separated by a ';').
-- **reset**: same as clear but also resets the generation configs to defaults if they have been changed by **set**
-- **save {SAVE_NAME} (optional)**: save the current chat and settings to file by default to `./chat_history/{MODEL_NAME}/chat_{DATETIME}.yaml` or `{SAVE_NAME}` if provided
-- **exit**: closes the interface
-
-The default examples are defined in `examples/scripts/config/default_chat_config.yaml` but you can pass your own with `--config CONFIG_FILE` where you can also specify the default generation parameters.

docs/source/community_tutorials.md

@@ -0,0 +1,34 @@
+# Community Tutorials
+
+Community tutorials are made by active members of the Hugging Face community who want to share their knowledge and expertise with others. They are a great way to learn about the library and its features, and to get started with core classes and modalities.
+
+# Language Models
+
+| Task | Class | Description | Author | Tutorial | Colab |
+| --- | --- | --- | --- | --- | --- |
+| Reinforcement Learning | [`GRPOTrainer`] | Post training an LLM for reasoning with GRPO in TRL | [Sergio Paniego](https://huggingface.co/sergiopaniego) | [Link](https://huggingface.co/learn/cookbook/fine_tuning_llm_grpo_trl) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/cookbook/blob/main/notebooks/en/fine_tuning_llm_grpo_trl.ipynb) |
+| Reinforcement Learning | [`GRPOTrainer`] | Mini-R1: Reproduce Deepseek R1 „aha moment“ a RL tutorial | [Philipp Schmid](https://huggingface.co/philschmid) | [Link](https://www.philschmid.de/mini-deepseek-r1) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/philschmid/deep-learning-pytorch-huggingface/blob/main/training/mini-deepseek-r1-aha-grpo.ipynb) |
+| Reinforcement Learning | [`GRPOTrainer`] | RL on LLaMA 3.1-8B with GRPO and Unsloth optimizations | [Andrea Manzoni](https://huggingface.co/AManzoni) | [Link](https://colab.research.google.com/github/amanzoni1/fine_tuning/blob/main/RL_LLama3_1_8B_GRPO.ipynb) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/amanzoni1/fine_tuning/blob/main/RL_LLama3_1_8B_GRPO.ipynb) | 
+| Instruction tuning | [`SFTTrainer`] | Fine-tuning Google Gemma LLMs using ChatML format with QLoRA | [Philipp Schmid](https://huggingface.co/philschmid) | [Link](https://www.philschmid.de/fine-tune-google-gemma) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/philschmid/deep-learning-pytorch-huggingface/blob/main/training/gemma-lora-example.ipynb) |
+| Structured Generation | [`SFTTrainer`] | Fine-tuning Llama-2-7B to generate Persian product catalogs in JSON using QLoRA and PEFT | [Mohammadreza Esmaeilian](https://huggingface.co/Mohammadreza) | [Link](https://huggingface.co/learn/cookbook/en/fine_tuning_llm_to_generate_persian_product_catalogs_in_json_format) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/cookbook/blob/main/notebooks/en/fine_tuning_llm_to_generate_persian_product_catalogs_in_json_format.ipynb) |
+| Preference Optimization | [`DPOTrainer`] | Align Mistral-7b using Direct Preference Optimization for human preference alignment | [Maxime Labonne](https://huggingface.co/mlabonne) | [Link](https://mlabonne.github.io/blog/posts/Fine_tune_Mistral_7b_with_DPO.html) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/mlabonne/llm-course/blob/main/Fine_tune_a_Mistral_7b_model_with_DPO.ipynb) |
+| Preference Optimization | [`ORPOTrainer`] | Fine-tuning Llama 3 with ORPO combining instruction tuning and preference alignment | [Maxime Labonne](https://huggingface.co/mlabonne) | [Link](https://mlabonne.github.io/blog/posts/2024-04-19_Fine_tune_Llama_3_with_ORPO.html) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1eHNWg9gnaXErdAa8_mcvjMupbSS6rDvi) |
+| Instruction tuning | [`SFTTrainer`] | How to fine-tune open LLMs in 2025 with Hugging Face | [Philipp Schmid](https://huggingface.co/philschmid) | [Link](https://www.philschmid.de/fine-tune-llms-in-2025) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/philschmid/deep-learning-pytorch-huggingface/blob/main/training/fine-tune-llms-in-2025.ipynb) |
+
+<Youtube id="cnGyyM0vOes" />
+
+# Vision Language Models
+
+| Task | Class | Description | Author | Tutorial | Colab |
+| --- | --- | --- | --- | --- | --- |
+| Visual QA | [`SFTTrainer`] | Fine-tuning Qwen2-VL-7B for visual question answering on ChartQA dataset | [Sergio Paniego](https://huggingface.co/sergiopaniego) | [Link](https://huggingface.co/learn/cookbook/fine_tuning_vlm_trl) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/cookbook/blob/main/notebooks/en/fine_tuning_vlm_trl.ipynb) |
+| Visual QA | [`SFTTrainer`] | Fine-tuning SmolVLM with TRL on a consumer GPU | [Sergio Paniego](https://huggingface.co/sergiopaniego) | [Link](https://huggingface.co/learn/cookbook/fine_tuning_smol_vlm_sft_trl) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/cookbook/blob/main/notebooks/en/fine_tuning_smol_vlm_sft_trl.ipynb) |
+| SEO Description | [`SFTTrainer`] | Fine-tuning Qwen2-VL-7B for generating SEO-friendly descriptions from images | [Philipp Schmid](https://huggingface.co/philschmid) | [Link](https://www.philschmid.de/fine-tune-multimodal-llms-with-trl) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/philschmid/deep-learning-pytorch-huggingface/blob/main/training/fine-tune-multimodal-llms-with-trl.ipynb) |
+| Visual QA | [`DPOTrainer`] | PaliGemma 🤝 Direct Preference Optimization | [Merve Noyan](https://huggingface.co/merve) | [Link](https://github.com/merveenoyan/smol-vision/blob/main/PaliGemma_DPO.ipynb) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/merveenoyan/smol-vision/blob/main/PaliGemma_DPO.ipynb) |
+| Visual QA | [`DPOTrainer`] | Fine-tuning SmolVLM using direct preference optimization (DPO) with TRL on a consumer GPU | [Sergio Paniego](https://huggingface.co/sergiopaniego) | [Link](https://huggingface.co/learn/cookbook/fine_tuning_vlm_dpo_smolvlm_instruct) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/cookbook/blob/main/notebooks/en/fine_tuning_vlm_dpo_smolvlm_instruct.ipynb) |
+| Object Detection Grounding | [`SFTTrainer`] | Fine tuning a VLM for Object Detection Grounding using TRL | [Sergio Paniego](https://huggingface.co/sergiopaniego) | [Link](https://huggingface.co/learn/cookbook/fine_tuning_vlm_object_detection_grounding) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/cookbook/blob/main/notebooks/en/fine_tuning_vlm_object_detection_grounding.ipynb) |
+| Visual QA | [`DPOTrainer`] | Fine-Tuning a Vision Language Model with TRL using MPO | [Sergio Paniego](https://huggingface.co/sergiopaniego) | [Link](https://huggingface.co/learn/cookbook/fine_tuning_vlm_mpo) | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/cookbook/blob/main/notebooks/en/fine_tuning_vlm_mpo.ipynb) |
+
+## Contributing
+
+If you have a tutorial that you would like to add to this list, please open a PR to add it. We will review it and merge it if it is relevant to the community.

docs/source/cpo_trainer.md

@@ -0,0 +1,107 @@
+# CPO Trainer
+
+[![](https://img.shields.io/badge/All_models-CPO-blue)](https://huggingface.co/models?other=cpo,trl)
+
+## Overview
+
+Contrastive Preference Optimization (CPO) as introduced in the paper [Contrastive Preference Optimization: Pushing the Boundaries of LLM Performance in Machine Translation](https://huggingface.co/papers/2401.08417) by [Haoran Xu](https://huggingface.co/haoranxu), [Amr Sharaf](https://huggingface.co/amrsharaf), [Yunmo Chen](https://huggingface.co/yunmochen), Weiting Tan, Lingfeng Shen, Benjamin Van Durme, [Kenton Murray](https://huggingface.co/Kenton), and [Young Jin Kim](https://huggingface.co/ykim362). At a high-level, CPO trains models to avoid generating adequate, but not perfect translations in Machine Translation (MT) tasks. However, CPO is a general approximation of the DPO loss and can be applied to other domains, such as chat.
+
+CPO aims to mitigate two fundamental shortcomings of SFT. First, SFT’s methodology of minimizing the discrepancy between predicted outputs and gold-standard references inherently caps model performance at the quality level of the training data. Secondly, SFT lacks a mechanism to prevent the model from rejecting mistakes in translations. The CPO objective is derived from the DPO objective.
+
+## Quick start
+
+This example demonstrates how to train a model using the CPO method. We use the [Qwen 0.5B model](https://huggingface.co/Qwen/Qwen2-0.5B-Instruct) as the base model. We use the preference data from the [UltraFeedback dataset](https://huggingface.co/datasets/openbmb/UltraFeedback). You can view the data in the dataset here:
+
+<iframe
+  src="https://huggingface.co/datasets/trl-lib/ultrafeedback_binarized/embed/viewer/default/train?row=0"
+  frameborder="0"
+  width="100%"
+  height="560px"
+></iframe>
+
+Below is the script to train the model:
+
+```python
+# train_cpo.py
+from datasets import load_dataset
+from trl import CPOConfig, CPOTrainer
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
+train_dataset = load_dataset("trl-lib/ultrafeedback_binarized", split="train")
+
+training_args = CPOConfig(output_dir="Qwen2-0.5B-CPO")
+trainer = CPOTrainer(model=model, args=training_args, processing_class=tokenizer, train_dataset=train_dataset)
+trainer.train()
+```
+
+Execute the script using the following command:
+
+```bash
+accelerate launch train_cpo.py
+```
+
+## Expected dataset type
+
+CPO requires a [preference dataset](dataset_formats#preference). The [`CPOTrainer`] supports both [conversational](dataset_formats#conversational) and [standard](dataset_formats#standard) dataset format. When provided with a conversational dataset, the trainer will automatically apply the chat template to the dataset.
+
+## Example script
+
+We provide an example script to train a model using the CPO method. The script is available in [`examples/scripts/cpo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/cpo.py)
+
+To test the CPO script with the [Qwen2 0.5B model](https://huggingface.co/Qwen/Qwen2-0.5B-Instruct) on the [UltraFeedback dataset](https://huggingface.co/datasets/trl-lib/ultrafeedback_binarized), run the following command:
+
+```bash
+accelerate launch examples/scripts/cpo.py \
+    --model_name_or_path Qwen/Qwen2-0.5B-Instruct \
+    --dataset_name trl-lib/ultrafeedback_binarized \
+    --num_train_epochs 1 \
+    --output_dir Qwen2-0.5B-CPO
+```
+
+## Logged metrics
+
+While training and evaluating we record the following reward metrics:
+
+* `rewards/chosen`: the mean log probabilities of the policy model for the chosen responses scaled by beta
+* `rewards/rejected`: the mean log probabilities of the policy model for the rejected responses scaled by beta
+* `rewards/accuracies`: mean of how often the chosen rewards are > than the corresponding rejected rewards
+* `rewards/margins`: the mean difference between the chosen and corresponding rejected rewards
+* `nll_loss`: the mean negative log likelihood loss of the policy model for the chosen responses
+
+## CPO variants
+
+### Simple Preference Optimization (SimPO)
+
+The [SimPO](https://huggingface.co/papers/2405.14734) method is also implemented in the [`CPOTrainer`]. SimPO is an alternative loss that adds a reward margin, allows for length normalization, and does not use BC regularization. To use this loss, we can use SimPO easily by turning on `loss_type="simpo"` and `cpo_alpha=0.0` in the [`CPOConfig`].
+
+### CPO-SimPO
+
+We also offer the combined use of CPO and SimPO, which enables more stable training and improved performance. Learn more details at [CPO-SimPO GitHub](https://github.com/fe1ixxu/CPO_SIMPO). To use this method, simply enable SimPO by setting `loss_type="simpo"` and a non-zero `cpo_alpha` in the [`CPOConfig`].
+
+## Loss functions
+
+The CPO algorithm supports several loss functions. The loss function can be set using the `loss_type` parameter in the [`CPOConfig`]. The following loss functions are supported:
+
+| `loss_type=`                           | Description                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                      |
+| -------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `"sigmoid"` (default)                  | Given the preference data, we can fit a binary classifier according to the Bradley-Terry model and in fact the [DPO](https://huggingface.co/papers/2305.18290) authors propose the sigmoid loss on the normalized likelihood via the `logsigmoid` to fit a logistic regression.                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  |
+| `"hinge"`                              | The [RSO](https://huggingface.co/papers/2309.06657) authors propose to use a hinge loss on the normalized likelihood from the [SLiC](https://huggingface.co/papers/2305.10425) paper. In this case, the `beta` is the reciprocal of the margin.                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                  |
+| `"ipo"`                                | The [IPO](https://huggingface.co/papers/2310.12036) authors provide a deeper theoretical understanding of the DPO algorithms and identify an issue with overfitting and propose an alternative loss. In this case, the `beta` is the reciprocal of the gap between the log-likelihood ratios of the chosen vs the rejected completion pair and thus the smaller the `beta` the larger this gaps is. As per the paper the loss is averaged over log-likelihoods of the completion (unlike DPO which is summed only).                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                              |
+
+### For Mixture of Experts Models: Enabling the auxiliary loss
+
+MOEs are the most efficient if the load is about equally distributed between experts.  
+To ensure that we train MOEs similarly during preference-tuning, it is beneficial to add the auxiliary loss from the load balancer to the final loss.
+
+This option is enabled by setting `output_router_logits=True` in the model config (e.g. [`~transformers.MixtralConfig`]).  
+To scale how much the auxiliary loss contributes to the total loss, use the hyperparameter `router_aux_loss_coef=...` (default: `0.001`) in the model config.
+
+## CPOTrainer
+
+[[autodoc]] CPOTrainer
+
+## CPOConfig
+
+[[autodoc]] CPOConfig

docs/source/cpo_trainer.mdx

@@ -1,113 +0,0 @@
-# CPO Trainer
-
-Contrastive Preference Optimization (CPO) as introduced in the paper [Contrastive Preference Optimization: Pushing the Boundaries of LLM Performance in Machine Translation](https://huggingface.co/papers/2401.08417) by Haoran Xu, Amr Sharaf, Yunmo Chen, Weiting Tan, Lingfeng Shen, Benjamin Van Durme, Kenton Murray, and Young Jin Kim. At a high-level, CPO  trains models to
-avoid generating adequate, but not perfect translations in Machine Translation (MT) tasks. However, CPO is a general approximation to the DPO loss and can be applied to other domains like chat.
-
-CPO aims to mitigate two fundamental shortcomings of SFT. First, SFT’s methodology of minimizing the discrepancy between predicted outputs and gold-standard references inherently caps model performance at the quality level of the training data. Secondly, SFT lacks a mechanism to prevent the model from rejecting mistakes in translations. The CPO objective is derived from the DPO objective.
-
-## SimPO
-The [SimPO](https://huggingface.co/papers/2405.14734) method is also implemented in the `CPOTrainer`. SimPO is an alternative loss that adds a reward margin, allows for length normalization, and does not use BC regularization. To use this loss, we can use SimPO easily by turning on `loss_type="simpo"` and `cpo_alpha=0` in the `CPOConfig`.
-
-## CPO-SimPO
-We also offer the combined use of CPO and SimPO, which enables more stable training and improved performance. Learn more details at [CPO-SimPO Github](https://github.com/fe1ixxu/CPO_SIMPO). To use this method, simply enable SimPO by setting `loss_type="simpo"` and a non-zero `cpo_alpha` in the CPOConfig.
-
-## Expected dataset format
-
-The CPO trainer expects a format identical to the DPO trainer, which should include three entries. These entries should be named as follows:
-
-- `prompt`
-- `chosen`
-- `rejected`
-
-for example:
-
-```py
-cpo_dataset_dict = {
-    "prompt": [
-        "hello",
-        "how are you",
-        "What is your name?",
-        "What is your name?",
-        "Which is the best programming language?",
-        "Which is the best programming language?",
-        "Which is the best programming language?",
-    ],
-    "chosen": [
-        "hi nice to meet you",
-        "I am fine",
-        "My name is Mary",
-        "My name is Mary",
-        "Python",
-        "Python",
-        "Java",
-    ],
-    "rejected": [
-        "leave me alone",
-        "I am not fine",
-        "Whats it to you?",
-        "I dont have a name",
-        "Javascript",
-        "C++",
-        "C++",
-    ],
-}
-```
-where the `prompt` contains the context inputs, `chosen` contains the corresponding chosen responses and `rejected` contains the corresponding negative (rejected) responses. As can be seen a prompt can have multiple responses and this is reflected in the entries being repeated in the dictionary's value arrays.
-
-## Expected model format
-The CPO trainer expects a model of `AutoModelForCausalLM`, compared to PPO that expects `AutoModelForCausalLMWithValueHead` for the value function.
-
-## Using the `CPOTrainer`
-For a detailed example have a look at the `examples/scripts/cpo.py` script. At a high level we need to initialize the `CPOTrainer` with a `model` we wish to train. **Note that CPOTrainer eliminates the need to use the reference model, simplifying the optimization process.** The `beta` refers to the hyperparameter of the implicit reward, and the dataset contains the 3 entries listed above.
-
-```py
-cpo_config = CPOConfig(
-    beta=0.1,
-)
-
-cpo_trainer = CPOTrainer(
-    model,
-    args=cpo_config,
-    train_dataset=train_dataset,
-    tokenizer=tokenizer,
-)
-```
-After this one can then call:
-
-```py
-cpo_trainer.train()
-```
-
-## Loss functions
-
-Given the preference data, the `CPOTrainer` uses the sigmoid loss on the normalized likelihood via the `logsigmoid` to fit a logistic regression.
-
-The [RSO](https://huggingface.co/papers/2309.06657) authors propose to use a hinge loss on the normalized likelihood from the [SLiC](https://huggingface.co/papers/2305.10425) paper. The `CPOTrainer` can be switched to this loss via the `loss_type="hinge"` argument and the `beta` in this case is the reciprocal of the margin.
-
-The [IPO](https://huggingface.co/papers/2310.12036) authors provide a deeper theoretical understanding of the CPO algorithms and identify an issue with overfitting and propose an alternative loss which can be used via the `loss_type="ipo"` argument to the trainer. Note that the `beta`  parameter is the reciprocal of the gap between the log-likelihood ratios of the chosen vs the rejected completion pair and thus the smaller the `beta` the larger this gaps is. As per the paper the loss is averaged over log-likelihoods of the completion (unlike CPO which is summed only).
-
-### For Mixture of Experts Models: Enabling the auxiliary loss
-
-MOEs are the most efficient if the load is about equally distributed between experts.  
-To ensure that we train MOEs similarly during preference-tuning, it is beneficial to add the auxiliary loss from the load balancer to the final loss.  
-
-This option is enabled by setting `output_router_logits=True` in the model config (e.g. MixtralConfig).  
-To scale how much the auxiliary loss contributes to the total loss, use the hyperparameter `router_aux_loss_coef=...` (default: 0.001).
-
-## Logging
-
-While training and evaluating we record the following reward metrics:
-
-* `rewards/chosen`: the mean log probabilities of the policy model for the chosen responses scaled by beta
-* `rewards/rejected`: the mean log probabilities of the policy model for the rejected responses scaled by beta
-* `rewards/accuracies`: mean of how often the chosen rewards are > than the corresponding rejected rewards
-* `rewards/margins`: the mean difference between the chosen and corresponding rejected rewards
-* `nll_loss`: the mean negative log likelihood loss of the policy model for the chosen responses
-
-## CPOTrainer
-
-[[autodoc]] CPOTrainer
-
-## CPOConfig
-
-[[autodoc]] CPOConfig

docs/source/customization.md

@@ -0,0 +1,121 @@
+# Training customization
+
+TRL is designed with modularity in mind so that users to be able to efficiently customize the training loop for their needs. Below are some examples on how you can apply and test different techniques.  Note: Although these examples use the DPOTrainer, the customization applies to most (if not all) trainers.
+
+
+
+## Use different optimizers and schedulers
+
+By default, the `DPOTrainer` creates a `torch.optim.AdamW` optimizer. You can create and define a different optimizer and pass it to `DPOTrainer` as follows:
+
+```python
+from datasets import load_dataset
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from torch import optim
+from trl import DPOConfig, DPOTrainer
+
+model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")
+dataset = load_dataset("trl-lib/ultrafeedback_binarized", split="train")
+training_args = DPOConfig(output_dir="Qwen2.5-0.5B-DPO")
+
+optimizer = optim.SGD(model.parameters(), lr=training_args.learning_rate)
+
+trainer = DPOTrainer(
+    model=model,
+    args=training_args,
+    train_dataset=dataset,
+    tokenizer=tokenizer,
+    optimizers=(optimizer, None),
+)
+trainer.train()
+```
+
+### Add a learning rate scheduler
+
+You can also play with your training by adding learning rate schedulers.
+
+```python
+from datasets import load_dataset
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from torch import optim
+from trl import DPOConfig, DPOTrainer
+
+model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")
+dataset = load_dataset("trl-lib/ultrafeedback_binarized", split="train")
+training_args = DPOConfig(output_dir="Qwen2.5-0.5B-DPO")
+
+optimizer = optim.AdamW(model.parameters(), lr=training_args.learning_rate)
+lr_scheduler = optim.lr_scheduler.StepLR(optimizer, step_size=30, gamma=0.1)
+
+trainer = DPOTrainer(
+    model=model,
+    args=training_args,
+    train_dataset=dataset,
+    tokenizer=tokenizer,
+    optimizers=(optimizer, lr_scheduler),
+)
+trainer.train()
+```
+
+## Memory efficient fine-tuning by sharing layers
+
+Another tool you can use for more memory efficient fine-tuning is to share layers between the reference model and the model you want to train.
+
+```python
+from datasets import load_dataset
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from trl import create_reference_model, DPOConfig, DPOTrainer
+
+model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")
+ref_model = create_reference_model(model, num_shared_layers=6)
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")
+dataset = load_dataset("trl-lib/ultrafeedback_binarized", split="train[:1%]")
+training_args = DPOConfig(output_dir="Qwen2.5-0.5B-DPO")
+
+trainer = DPOTrainer(
+    model=model,
+    ref_model=ref_model,
+    args=training_args,
+    train_dataset=dataset,
+    tokenizer=tokenizer,
+)
+trainer.train()
+```
+
+## Pass 8-bit reference models 
+ 
+Since `trl` supports all keyword arguments when loading a model from `transformers` using `from_pretrained`, you can also leverage `load_in_8bit` from `transformers` for more memory efficient fine-tuning.
+
+Read more about 8-bit model loading in `transformers` [here](https://huggingface.co/docs/transformers/en/peft#load-in-8bit-or-4bit).
+
+```python
+from datasets import load_dataset
+from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
+from trl import DPOConfig, DPOTrainer
+
+model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")
+quantization_config = BitsAndBytesConfig(load_in_8bit=True)
+ref_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct", quantization_config= quantization_config)
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")
+dataset = load_dataset("trl-lib/ultrafeedback_binarized", split="train")
+training_args = DPOConfig(output_dir="Qwen2.5-0.5B-DPO")
+
+trainer = DPOTrainer(
+    model=model,
+    ref_model=ref_model,
+    args=training_args,
+    train_dataset=dataset,
+    tokenizer=tokenizer,
+)
+trainer.train()
+```
+
+## Use the accelerator cache optimizer
+
+When training large models, you should better handle the accelerator cache by iteratively clearing it. To do so, simply pass `optimize_device_cache=True` to `DPOConfig`:
+
+```python
+training_args = DPOConfig(..., optimize_device_cache=True)
+```

docs/source/customization.mdx

@@ -1,216 +0,0 @@
-# Training customization
-
-TRL is designed with modularity in mind so that users to be able to efficiently customize the training loop for their needs. Below are some examples on how you can apply and test different techniques.
-
-## Train on multiple GPUs / nodes
-
-The trainers in TRL use 🤗 Accelerate to enable distributed training across multiple GPUs or nodes. To do so, first create an 🤗 Accelerate config file by running
-
-```bash
-accelerate config
-```
-
-and answering the questions according to your multi-gpu / multi-node setup. You can then launch distributed training by running:
-
-```bash
-accelerate launch your_script.py
-```
-
-We also provide config files in the [examples folder](https://github.com/huggingface/trl/tree/main/examples/accelerate_configs) that can be used as templates. To use these templates, simply pass the path to the config file when launching a job, e.g.:
-
-```shell
-accelerate launch --config_file=examples/accelerate_configs/multi_gpu.yaml --num_processes {NUM_GPUS} path_to_script.py --all_arguments_of_the_script
-```
-
-Refer to the [examples page](https://github.com/huggingface/trl/tree/main/examples) for more details.
-
-### Distributed training with DeepSpeed
-
-All of the trainers in TRL can be run on multiple GPUs together with DeepSpeed ZeRO-{1,2,3} for efficient sharding of the optimizer states, gradients, and model weights. To do so, run:
-
-```shell
-accelerate launch --config_file=examples/accelerate_configs/deepspeed_zero{1,2,3}.yaml --num_processes {NUM_GPUS} path_to_your_script.py --all_arguments_of_the_script
-```
-
-Note that for ZeRO-3, a small tweak is needed to initialize your reward model on the correct device via the `zero3_init_context_manager()` context manager. In particular, this is needed to avoid DeepSpeed hanging after a fixed number of training steps. Here is a snippet of what is involved from the [`sentiment_tuning`](https://github.com/huggingface/trl/blob/main/examples/scripts/ppo.py) example:
-
-```python
-ds_plugin = ppo_trainer.accelerator.state.deepspeed_plugin
-if ds_plugin is not None and ds_plugin.is_zero3_init_enabled():
-    with ds_plugin.zero3_init_context_manager(enable=False):
-        sentiment_pipe = pipeline("sentiment-analysis", model="lvwerra/distilbert-imdb", device=device)
-else:
-    sentiment_pipe = pipeline("sentiment-analysis", model="lvwerra/distilbert-imdb", device=device)
-```
-
-Consult the 🤗 Accelerate [documentation](https://huggingface.co/docs/accelerate/usage_guides/deepspeed) for more information about the DeepSpeed plugin.
-
-
-## Use different optimizers
-
-By default, the `PPOTrainer` creates a `torch.optim.Adam` optimizer. You can create and define a different optimizer and pass it to `PPOTrainer`:
-```python
-import torch
-from transformers import GPT2Tokenizer
-from trl import PPOTrainer, PPOConfig, AutoModelForCausalLMWithValueHead
-
-# 1. load a pretrained model
-model = AutoModelForCausalLMWithValueHead.from_pretrained('gpt2')
-ref_model = AutoModelForCausalLMWithValueHead.from_pretrained('gpt2')
-tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
-
-# 2. define config
-ppo_config = {'batch_size': 1, 'learning_rate':1e-5}
-config = PPOConfig(**ppo_config)
-
-
-# 2. Create optimizer
-optimizer = torch.optim.SGD(model.parameters(), lr=config.learning_rate)
-
-
-# 3. initialize trainer
-ppo_trainer = PPOTrainer(config, model, ref_model, tokenizer, optimizer=optimizer)
-```
-
-For memory efficient fine-tuning, you can also pass `Adam8bit` optimizer from `bitsandbytes`:
-
-```python
-import torch
-import bitsandbytes as bnb
-
-from transformers import GPT2Tokenizer
-from trl import PPOTrainer, PPOConfig, AutoModelForCausalLMWithValueHead
-
-# 1. load a pretrained model
-model = AutoModelForCausalLMWithValueHead.from_pretrained('gpt2')
-ref_model = AutoModelForCausalLMWithValueHead.from_pretrained('gpt2')
-tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
-
-# 2. define config
-ppo_config = {'batch_size': 1, 'learning_rate':1e-5}
-config = PPOConfig(**ppo_config)
-
-
-# 2. Create optimizer
-optimizer = bnb.optim.Adam8bit(model.parameters(), lr=config.learning_rate)
-
-# 3. initialize trainer
-ppo_trainer = PPOTrainer(config, model, ref_model, tokenizer, optimizer=optimizer)
-```
-
-### Use LION optimizer
-
-You can use the new [LION optimizer from Google](https://huggingface.co/papers/2302.06675) as well, first take the source code of the optimizer definition [here](https://github.com/lucidrains/lion-pytorch/blob/main/lion_pytorch/lion_pytorch.py), and copy it so that you can import the optimizer. Make sure to initialize the optimizer by considering the trainable parameters only for a more memory efficient training:
-```python
-optimizer = Lion(filter(lambda p: p.requires_grad, self.model.parameters()), lr=self.config.learning_rate)
-
-...
-ppo_trainer = PPOTrainer(config, model, ref_model, tokenizer, optimizer=optimizer)
-```
-We advise you to use the learning rate that you would use for `Adam` divided by 3 as pointed out [here](https://github.com/lucidrains/lion-pytorch#lion---pytorch). We observed an improvement when using this optimizer compared to classic Adam (check the full logs [here](https://wandb.ai/distill-bloom/trl/runs/lj4bheke?workspace=user-younesbelkada)):
-
-<div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/trl-lion.png">
-</div>
-
-
-## Add a learning rate scheduler
-
-You can also play with your training by adding learning rate schedulers!
-```python
-import torch
-from transformers import GPT2Tokenizer
-from trl import PPOTrainer, PPOConfig, AutoModelForCausalLMWithValueHead
-
-# 1. load a pretrained model
-model = AutoModelForCausalLMWithValueHead.from_pretrained('gpt2')
-ref_model = AutoModelForCausalLMWithValueHead.from_pretrained('gpt2')
-tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
-
-# 2. define config
-ppo_config = {'batch_size': 1, 'learning_rate':1e-5}
-config = PPOConfig(**ppo_config)
-
-
-# 2. Create optimizer
-optimizer = torch.optim.SGD(model.parameters(), lr=config.learning_rate)
-lr_scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer, gamma=0.9)
-
-# 3. initialize trainer
-ppo_trainer = PPOTrainer(config, model, ref_model, tokenizer, optimizer=optimizer, lr_scheduler=lr_scheduler)
-```
-
-## Memory efficient fine-tuning by sharing layers
-
-Another tool you can use for more memory efficient fine-tuning is to share layers between the reference model and the model you want to train.
-```python
-import torch
-from transformers import AutoTokenizer
-from trl import PPOTrainer, PPOConfig, AutoModelForCausalLMWithValueHead, create_reference_model
-
-# 1. load a pretrained model
-model = AutoModelForCausalLMWithValueHead.from_pretrained('bigscience/bloom-560m')
-ref_model = create_reference_model(model, num_shared_layers=6)
-tokenizer = AutoTokenizer.from_pretrained('bigscience/bloom-560m')
-
-# 2. initialize trainer
-ppo_config = {'batch_size': 1}
-config = PPOConfig(**ppo_config)
-ppo_trainer = PPOTrainer(config, model, ref_model, tokenizer)
-```
-
-## Pass 8-bit reference models 
- 
-<div>
-
-Since `trl` supports all key word arguments when loading a model from `transformers` using `from_pretrained`, you can also leverage `load_in_8bit` from `transformers` for more memory efficient fine-tuning.
-
-Read more about 8-bit model loading in `transformers` [here](https://huggingface.co/docs/transformers/perf_infer_gpu_one#bitsandbytes-integration-for-int8-mixedprecision-matrix-decomposition).
-
-</div>
-
-```python
-# 0. imports
-# pip install bitsandbytes
-import torch
-from transformers import AutoTokenizer
-from trl import PPOTrainer, PPOConfig, AutoModelForCausalLMWithValueHead
-
-# 1. load a pretrained model
-model = AutoModelForCausalLMWithValueHead.from_pretrained('bigscience/bloom-560m')
-ref_model = AutoModelForCausalLMWithValueHead.from_pretrained('bigscience/bloom-560m', device_map="auto", load_in_8bit=True)
-tokenizer = AutoTokenizer.from_pretrained('bigscience/bloom-560m')
-
-# 2. initialize trainer
-ppo_config = {'batch_size': 1}
-config = PPOConfig(**ppo_config)
-ppo_trainer = PPOTrainer(config, model, ref_model, tokenizer)
-```
-
-## Use the CUDA cache optimizer
-
-When training large models, you should better handle the CUDA cache by iteratively clearing it. Do do so, simply pass `optimize_cuda_cache=True` to `PPOConfig`:
-
-```python
-config = PPOConfig(..., optimize_cuda_cache=True)
-```
-
-
-
-## Use score scaling/normalization/clipping
-As suggested by [Secrets of RLHF in Large Language Models Part I: PPO](https://huggingface.co/papers/2307.04964), we support score (aka reward) scaling/normalization/clipping to improve training stability via `PPOConfig`:
-```python
-from trl import PPOConfig
-
-ppo_config = {
-    use_score_scaling=True,
-    use_score_norm=True,
-    score_clip=0.5,
-}
-config = PPOConfig(**ppo_config)
-```
-
-To run `ppo.py`, you can use the following command:
-```
-python examples/scripts/ppo.py --log_with wandb --use_score_scaling --use_score_norm --score_clip 0.5
-```

docs/source/data_utils.md

@@ -0,0 +1,41 @@
+# Data Utilities
+
+## is_conversational
+
+[[autodoc]] is_conversational
+
+## apply_chat_template
+
+[[autodoc]] apply_chat_template
+
+## maybe_apply_chat_template
+
+[[autodoc]] maybe_apply_chat_template
+
+## maybe_convert_to_chatml
+    
+[[autodoc]] maybe_convert_to_chatml
+
+## extract_prompt
+
+[[autodoc]] extract_prompt
+
+## maybe_extract_prompt
+
+[[autodoc]] maybe_extract_prompt
+
+## unpair_preference_dataset
+
+[[autodoc]] unpair_preference_dataset
+
+## maybe_unpair_preference_dataset
+
+[[autodoc]] maybe_unpair_preference_dataset
+
+## pack_dataset
+
+[[autodoc]] pack_dataset
+
+## truncate_dataset
+
+[[autodoc]] truncate_dataset

docs/source/ddpo_trainer.md

@@ -1,18 +1,21 @@
 # Denoising Diffusion Policy Optimization
+
+[![](https://img.shields.io/badge/All_models-DDPO-blue)](https://huggingface.co/models?other=ddpo,trl)
+
 ## The why
 
 | Before | After DDPO finetuning |
 | --- | --- |
-| <div style="text-align: center"><img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/pre_squirrel.png"/></div> |  <div style="text-align: center"><img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/post_squirrel.png"/></div> |
-| <div style="text-align: center"><img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/pre_crab.png"/></div> |  <div style="text-align: center"><img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/post_crab.png"/></div> |
-|  <div style="text-align: center"><img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/pre_starfish.png"/></div> |  <div style="text-align: center"><img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/post_starfish.png"/></div> |
+| <div style="text-align: center"><img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/pre_squirrel.png"/></div> |  <div style="text-align: center"><img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/post_squirrel.png"/></div> |
+| <div style="text-align: center"><img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/pre_crab.png"/></div> |  <div style="text-align: center"><img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/post_crab.png"/></div> |
+| <div style="text-align: center"><img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/pre_starfish.png"/></div> |  <div style="text-align: center"><img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/post_starfish.png"/></div> |
 
 
 ## Getting started with Stable Diffusion finetuning with reinforcement learning
 
 The machinery for finetuning of Stable Diffusion models with reinforcement learning makes heavy use of HuggingFace's `diffusers`
-library. A reason for  stating this is that getting started requires a bit of familiarity with the `diffusers` library concepts, mainly two of them - pipelines and schedulers.
-Right out of the box (`diffusers` library), there isn't a `Pipeline` nor a `Scheduler` instance that is suitable for finetuning with reinforcement learning. Some adjustments need to made. 
+library. A reason for stating this is that getting started requires a bit of familiarity with the `diffusers` library concepts, mainly two of them - pipelines and schedulers.
+Right out of the box (`diffusers` library), there isn't a `Pipeline` nor a `Scheduler` instance that is suitable for finetuning with reinforcement learning. Some adjustments need to be made. 
 
 There is a pipeline interface that is provided by this library that is required to be implemented to be used with the `DDPOTrainer`, which is the main machinery for fine-tuning Stable Diffusion with reinforcement learning. **Note: Only the StableDiffusion architecture is supported at this point.**
 There is a default implementation of this interface that you can use out of the box. Assuming the default implementation is sufficient and/or to get things moving, refer to the training example alongside this guide. 
@@ -23,7 +26,7 @@ For a more detailed look into the interface and the associated default implement
 
 Note that the default implementation has a LoRA implementation path and a non-LoRA based implementation path. The LoRA flag enabled by default and this can be turned off by passing in the flag to do so. LORA based training is faster and the LORA associated model hyperparameters responsible for model convergence aren't as finicky as non-LORA based training.
 
-Also in addition, there is the expectation of providing a reward function and a prompt function. The reward function is used to evaluate the generated images  and the prompt function is used to generate the prompts that are used to generate the images.
+Also in addition, there is the expectation of providing a reward function and a prompt function. The reward function is used to evaluate the generated images and the prompt function is used to generate the prompts that are used to generate the images.
 
 ## Getting started with `examples/scripts/ddpo.py`
 
@@ -116,4 +119,13 @@ for prompt, image in zip(prompts,results.images):
 ## Credits
 
 This work is heavily influenced by the repo [here](https://github.com/kvablack/ddpo-pytorch) and the associated paper [Training Diffusion Models
-with Reinforcement Learning by Kevin Black, Michael Janner, Yilan Du, Ilya Kostrikov, Sergey Levine](https://huggingface.co/papers/2305.13301).
+with Reinforcement Learning by Kevin Black, Michael Janner, Yilan Du, Ilya Kostrikov, Sergey Levine](https://huggingface.co/papers/2305.13301).
+
+## DDPOTrainer
+
+[[autodoc]] DDPOTrainer
+
+## DDPOConfig
+
+[[autodoc]] DDPOConfig
+

docs/source/deepspeed_integration.md

@@ -0,0 +1,39 @@
+# DeepSpeed Integration
+
+<Tip warning={true}>
+
+Section under construction. Feel free to contribute!
+
+</Tip>
+
+TRL supports training with DeepSpeed, a library that implements advanced training optimization techniques. These include optimizer state partitioning, offloading, gradient partitioning, and more.
+
+DeepSpeed integrates the [Zero Redundancy Optimizer (ZeRO)](https://huggingface.co/papers/1910.02054), which allows to scale the model size proportional to the number of devices with sustained high efficiency.
+
+![ZeRO Stages](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/zero_stages.png)
+
+## Installation
+
+To use DeepSpeed with TRL, install it using the following command:
+
+```bash
+pip install deepspeed
+```
+
+## Running Training Scripts with DeepSpeed
+
+No modifications to your training script are required. Simply run it with the DeepSpeed configuration file:
+
+```bash
+accelerate launch --config_file <ACCELERATE_WITH_DEEPSPEED_CONFIG_FILE.yaml> train.py
+```
+
+We provide ready-to-use DeepSpeed configuration files in the [`examples/accelerate_configs`](https://github.com/huggingface/trl/tree/main/examples/accelerate_configs) directory. For example, to run training with ZeRO Stage 2, use the following command:
+
+```bash
+accelerate launch --config_file examples/accelerate_configs/deepspeed_zero2.yaml train.py
+```
+
+## Additional Resources
+
+Consult the 🤗 Accelerate [documentation](https://huggingface.co/docs/accelerate/usage_guides/deepspeed) for more information about the DeepSpeed plugin.

docs/source/detoxifying_a_lm.md

@@ -30,7 +30,7 @@ We selected the following models for our experiments to show that TRL can be eas
 * [`EleutherAI/gpt-neo-2.7B`](https://huggingface.co/EleutherAI/gpt-neo-2.7B) (2.7 billion parameters)
 * [`EleutherAI/gpt-j-6B`](https://huggingface.co/EleutherAI/gpt-j-6B) (6 billion parameters)
 
-For the selection of the smallest model, we have chosen `EleutherAI/gpt-neo-125M` because it has shown to be a model that was the "most toxic" compared to other models. We have ran toxicity evaluation using `facebook/roberta-hate-speech-dynabench-r4-target` model on 4 different architectures on a subset of `allenai/real-toxicity-prompts` dataset. Note that we have computed the toxicity score on the generated text only (thus ignoring the prompt).
+For the selection of the smallest model, we have chosen `EleutherAI/gpt-neo-125M` because it has shown to be a model that was the "most toxic" compared to other models. We have run toxicity evaluation using `facebook/roberta-hate-speech-dynabench-r4-target` model on 4 different architectures on a subset of `allenai/real-toxicity-prompts` dataset. Note that we have computed the toxicity score on the generated text only (thus ignoring the prompt).
 
 | Model | Mean toxicity score |
 |---|---| 
@@ -45,7 +45,7 @@ When doing PPO, it is very important to design the problem efficiently so that t
 
 ### Pre-processing the dataset
 
-The dataset consist of prompts and their continuations, and each of them has an associated `toxicity` score.
+The dataset consists of prompts and their continuations, and each of them has an associated `toxicity` score.
 
 A `prompt` example:
 ```
@@ -58,13 +58,13 @@ And its `continuation` value:
 
 We want to increase the chance for the model to generate toxic prompts so we get more learning signal. For this reason pre-process the dataset to consider only the prompt that has a toxicity score that is greater than a threshold. We can do this in a few lines of code:
 ```python
-ds = load_dataset("allenai/real-toxicity-prompts", split="train")
+train_dataset = load_dataset("allenai/real-toxicity-prompts", split="train")
 
 def filter_fn(sample):
     toxicity = sample["prompt"]["toxicity"]
     return toxicity is not None and toxicity > 0.3
 
-ds = ds.filter(filter_fn, batched=False)
+train_dataset = train_dataset.filter(filter_fn, batched=False)
 ```
 
 ### Reward function
@@ -83,12 +83,12 @@ As a compromise between the two we took for a context window of 10 to 15 tokens
 
 
 <div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/trl-long-vs-short-context.png">
+<img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/trl-long-vs-short-context.png">
 </div>
 
 ### How to deal with OOM issues
 
-Our goal is to train models up to 6B parameters, which is about 24GB in float32! Here two tricks we use to be able to train a 6B model on a single 40GB-RAM GPU:
+Our goal is to train models up to 6B parameters, which is about 24GB in float32! Here are two tricks we use to be able to train a 6B model on a single 40GB-RAM GPU:
 
 - Use `bfloat16` precision: Simply load your model in `bfloat16` when calling `from_pretrained` and you can reduce the size of the model by 2:
 
@@ -98,22 +98,18 @@ model = AutoModelForCausalLM.from_pretrained("EleutherAI/gpt-j-6B", torch_dtype=
 
 and the optimizer will take care of computing the gradients in `bfloat16` precision. Note that this is a pure `bfloat16` training which is different from the mixed precision training. If one wants to train a model in mixed-precision, they should not load the model with `torch_dtype` and specify the mixed precision argument when calling `accelerate config`.
 
-- Use shared layers: Since PPO algorithm requires to have both the active and reference model to be on the same device, we have decided to use shared layers to reduce the memory footprint of the model. This can be achieved by just speifying `num_shared_layers` argument when creating a `PPOTrainer`:
+- Use shared layers: Since PPO algorithm requires to have both the active and reference model to be on the same device, we have decided to use shared layers to reduce the memory footprint of the model. This can be achieved by specifying `num_shared_layers` argument when calling the `create_reference_model()` function. For example, if you want to share the first 6 layers of the model, you can do it like this:
 
 <div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/trl-shared-layers.png">
+<img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/trl-shared-layers.png">
 </div>
 
 ```python
-ppo_trainer = PPOTrainer(
-    model=model,
-    tokenizer=tokenizer,
-    num_shared_layers=4,
-    ...
-)
+ref_model = create_reference_model(model, num_shared_layers=6)
+trainer = PPOTrainer(..., ref_model=ref_model)
 ```
 
-In the example above this means that the model have the 4 first layers frozen (i.e. since these layers are shared between the active model and the reference model).
+In the example above this means that the model has the 4 first layers frozen (i.e. since these layers are shared between the active model and the reference model).
 
 - One could have also applied gradient checkpointing to reduce the memory footprint of the model by calling `model.pretrained_model.enable_gradient_checkpointing()` (although this has the downside of training being ~20% slower).
 
@@ -128,13 +124,13 @@ We have decided to keep 3 models in total that correspond to our best models:
 We have used different learning rates for each model, and have found out that the largest models were quite hard to train and can easily lead to collapse mode if the learning rate is not chosen correctly (i.e. if the learning rate is too high):
 
 <div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/trl-collapse-mode.png">
+<img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/trl-collapse-mode.png">
 </div>
 
 The final training run of `ybelkada/gpt-j-6b-detoxified-20shdl` looks like this:
 
 <div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/trl-gpt-j-final-run-2.png">
+<img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/trl-gpt-j-final-run-2.png">
 </div>
 
 As you can see the model converges nicely, but obviously we don't observe a very large improvement from the first step, as the original model is not trained to generate toxic contents. 
@@ -142,7 +138,7 @@ As you can see the model converges nicely, but obviously we don't observe a very
 Also we have observed that training with larger `mini_batch_size` leads to smoother convergence and better results on the test set:
 
 <div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/trl-gpt-j-mbs-run.png">
+<img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/trl-gpt-j-mbs-run.png">
 </div>
 
 ## Results
@@ -163,7 +159,7 @@ We report the toxicity score of 400 sampled examples, compute its mean and stand
 
 <div class="column" style="text-align:center">
   <figure>
-    <img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/trl-final-barplot.png" style="width:80%">
+    <img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/trl-final-barplot.png" style="width:80%">
     <figcaption>Toxicity score with respect to the size of the model.</figcaption>
   </figure>
 </div>
@@ -171,7 +167,7 @@ We report the toxicity score of 400 sampled examples, compute its mean and stand
 Below are few generation examples of `gpt-j-6b-detox` model:
 
 <div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/trl-toxicity-examples.png">
+<img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/trl-toxicity-examples.png">
 </div>
 
 The evaluation script can be found [here](https://github.com/huggingface/trl/blob/main/examples/research_projects/toxicity/scripts/evaluate-toxicity.py).
@@ -180,7 +176,7 @@ The evaluation script can be found [here](https://github.com/huggingface/trl/blo
 
 The results are quite promising, as we can see that the models are able to reduce the toxicity score of the generated text by an interesting margin. The gap is clear for `gpt-neo-2B` model but we less so for the `gpt-j-6B` model. There are several things we could try to improve the results on the largest model starting with training with larger `mini_batch_size` and probably allowing to back-propagate through more layers (i.e. use less shared layers).
 
-To sum up, in addition to human feedback this could be a useful additional signal when training large language models to ensure there outputs are less toxic as well as useful.
+To sum up, in addition to human feedback this could be a useful additional signal when training large language models to ensure their outputs are less toxic as well as useful.
 
 ### Limitations
 

docs/source/distributing_training.md

@@ -0,0 +1,60 @@
+# Distributing Training
+
+<Tip warning={true}>
+Section under construction. Feel free to contribute!
+</Tip>
+
+## Multi-GPU Training with TRL
+
+The trainers in TRL use [🤗 Accelerate](https://github.com/huggingface/accelerate) to enable distributed training across multiple GPUs or nodes. To do so, first create an [🤗 Accelerate](https://github.com/huggingface/accelerate) config file by running
+
+```bash
+accelerate config
+```
+
+and answering the questions according to your multi-GPU / multi-node setup. You can then launch distributed training by running:
+
+```bash
+accelerate launch train.py
+```
+
+We also provide config files in the [examples folder](https://github.com/huggingface/trl/tree/main/examples/accelerate_configs) that can be used as templates. To use these templates, simply pass the path to the config file when launching a job, e.g.:
+
+```shell
+accelerate launch --config_file examples/accelerate_configs/multi_gpu.yaml train.py <SCRIPT_ARGS>
+```
+
+This automatically distributes the workload across all available GPUs.
+
+Under the hood, [🤗 Accelerate](https://github.com/huggingface/accelerate) creates one model per GPU. Each process:
+- Processes its own batch of data
+- Computes the loss and gradients for that batch
+- Shares gradient updates across all GPUs
+
+![](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/multi_gpu.png)
+
+The effective batch size is calculated as:
+
+$$
+\text{Batch Size} = \text{per\_device\_train\_batch\_size} \times \text{num\_devices} \times \text{gradient\_accumulation\_steps}
+$$
+
+To maintain a consistent batch size when scaling to multiple GPUs, make sure to update `per_device_train_batch_size` and `gradient_accumulation_steps` accordingly.
+
+Example, these configurations are equivalent, and should yield the same results:
+
+| Number of GPUs | Per device batch size | Gradient accumulation steps | Comments |
+| --- | --- | --- | --- |
+| 1 | 32 | 1 | Possibly high memory usage, but faster training |
+| 1 | 4 | 8 | Lower memory usage, slower training |
+| 8 | 4 | 1 | Multi-GPU to get the best of both worlds |
+
+<Tip> 
+
+Having one model per GPU can lead to high memory usage, which may not be feasible for large models or low-memory GPUs. In such cases, you can leverage [DeepSpeed](https://github.com/deepspeedai/DeepSpeed), which provides optimizations like model sharding, Zero Redundancy Optimizer, mixed precision training, and offloading to CPU or NVMe. Check out our [DeepSpeed Integration](deepspeed_integration.md) guide for more details.
+
+</Tip>
+
+## Multi-Nodes Training
+
+We're working on a guide for multi-node training. Stay tuned! 🚀

docs/source/dpo_trainer.md

@@ -0,0 +1,295 @@
+# DPO Trainer
+
+[![](https://img.shields.io/badge/All_models-DPO-blue)](https://huggingface.co/models?other=dpo,trl) [![](https://img.shields.io/badge/smol_course-Chapter_2-yellow)](https://github.com/huggingface/smol-course/tree/main/2_preference_alignment)
+
+## Overview
+
+TRL supports the DPO Trainer for training language models from preference data, as described in the paper [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://huggingface.co/papers/2305.18290) by [Rafael Rafailov](https://huggingface.co/rmrafailov), Archit Sharma, Eric Mitchell, [Stefano Ermon](https://huggingface.co/ermonste), [Christopher D. Manning](https://huggingface.co/manning), [Chelsea Finn](https://huggingface.co/cbfinn).
+
+The abstract from the paper is the following:
+
+> While large-scale unsupervised language models (LMs) learn broad world knowledge and some reasoning skills, achieving precise control of their behavior is difficult due to the completely unsupervised nature of their training. Existing methods for gaining such steerability collect human labels of the relative quality of model generations and fine-tune the unsupervised LM to align with these preferences, often with reinforcement learning from human feedback (RLHF). However, RLHF is a complex and often unstable procedure, first fitting a reward model that reflects the human preferences, and then fine-tuning the large unsupervised LM using reinforcement learning to maximize this estimated reward without drifting too far from the original model. In this paper we introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form, allowing us to solve the standard RLHF problem with only a simple classification loss. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight, eliminating the need for sampling from the LM during fine-tuning or performing significant hyperparameter tuning. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods. Notably, fine-tuning with DPO exceeds PPO-based RLHF in ability to control sentiment of generations, and matches or improves response quality in summarization and single-turn dialogue while being substantially simpler to implement and train.
+
+The first step is to train an SFT model, to ensure the data we train on is in-distribution for the DPO algorithm.
+
+Then, fine-tuning a language model via DPO consists of two steps and is easier than [PPO](ppo_trainer):
+
+1. **Data collection**: Gather a [preference dataset](dataset_formats#preference) with positive and negative selected pairs of generation, given a prompt.
+2. **Optimization**: Maximize the log-likelihood of the DPO loss directly.
+
+This process is illustrated in the sketch below (from [Figure 1 of the DPO paper](https://huggingface.co/papers/2305.18290)):
+
+![](https://github.com/huggingface/trl/assets/49240599/9150fac6-3d88-4ca2-8ec6-2a6f3473216d)
+
+Read more about DPO algorithm in the [original paper](https://huggingface.co/papers/2305.18290).
+
+## Quick start
+
+This example demonstrates how to train a model using the DPO method. We use the [Qwen 0.5B model](https://huggingface.co/Qwen/Qwen2-0.5B-Instruct) as the base model. We use the preference data from the [UltraFeedback dataset](https://huggingface.co/datasets/openbmb/UltraFeedback). You can view the data in the dataset here:
+
+<iframe
+  src="https://huggingface.co/datasets/trl-lib/ultrafeedback_binarized/embed/viewer/default/train?row=0"
+  frameborder="0"
+  width="100%"
+  height="560px"
+></iframe>
+
+Below is the script to train the model:
+
+```python
+# train_dpo.py
+from datasets import load_dataset
+from trl import DPOConfig, DPOTrainer
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
+train_dataset = load_dataset("trl-lib/ultrafeedback_binarized", split="train")
+
+training_args = DPOConfig(output_dir="Qwen2-0.5B-DPO")
+trainer = DPOTrainer(model=model, args=training_args, processing_class=tokenizer, train_dataset=train_dataset)
+trainer.train()
+```
+
+Execute the script using the following command:
+
+```bash
+accelerate launch train_dpo.py
+```
+
+Distributed across 8 GPUs, the training takes approximately 3 minutes. You can verify the training progress by checking the reward graph. An increasing trend in the reward margin indicates that the model is improving and generating better responses over time.
+
+![](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/dpo-qwen2-reward-margin.png)
+
+To see how the [trained model](https://huggingface.co/trl-lib/Qwen2-0.5B-DPO) performs, you can use the [Transformers Chat CLI](https://huggingface.co/docs/transformers/quicktour#chat-with-text-generation-models).
+
+<pre><code>$ transformers chat trl-lib/Qwen2-0.5B-DPO
+<strong><span style="color: red;">&lt;shirin_yamani&gt;:</span></strong>
+What is Huggingface?
+
+<strong><span style="color: blue;">&lt;trl-lib/Qwen2-0.5B-DPO&gt;:</span></strong>
+Huggingface is a platform that allows users to access a variety of open-source machine learning resources such as pre-trained models and datasets Huggingface is a platform that allows users to access a variety of open-source machine learning resources such as pre-trained models and datasets for the development of machine learning models and applications. It provides a repository of over 300, 000 pre-trained models in  Huggingface is a platform that allows users to access a variety of open-source machine learning resources such as pre-trained models and datasets for the development of machine learning models and applications. It provides a repository of over 300, 000  pre-trained models in a variety of languages, enabling users to explore and utilize the latest techniques and technologies in the field of machine learning.
+</code></pre>
+
+## Expected dataset type
+
+DPO requires a [preference dataset](dataset_formats#preference). The [`DPOTrainer`] supports both [conversational](dataset_formats#conversational) and [standard](dataset_formats#standard) dataset formats. When provided with a conversational dataset, the trainer will automatically apply the chat template to the dataset.
+
+Although the [`DPOTrainer`] supports both explicit and implicit prompts, we recommend using explicit prompts. If provided with an implicit prompt dataset, the trainer will automatically extract the prompt from the `"chosen"` and `"rejected"` columns. For more information, refer to the [preference style](dataset_formats#preference) section.
+
+### Special considerations for vision-language models
+
+The [`DPOTrainer`] supports fine-tuning vision-language models (VLMs). For these models, a vision dataset is required. To learn more about the specific format for vision datasets, refer to the [Vision dataset format](dataset_formats#vision-datasets) section.
+
+Additionally, unlike standard text-based models where a `tokenizer` is used, for VLMs, you should replace the `tokenizer` with a `processor`.
+
+```diff
+- model = AutoModelForCausalLM.from_pretrained(model_id)
++ model = AutoModelForVision2Seq.from_pretrained(model_id)
+
+- tokenizer = AutoTokenizer.from_pretrained(model_id)
++ processor = AutoProcessor.from_pretrained(model_id)
+
+  trainer = DPOTrainer(
+      model,
+      args=training_args,
+      train_dataset=train_dataset,
+-     processing_class=tokenizer,
++     processing_class=processor,
+)
+```
+
+For a complete example of fine-tuning a vision-language model, refer to the script in [`examples/scripts/dpo_vlm.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/dpo_vlm.py).
+
+
+## Example script
+
+We provide an example script to train a model using the DPO method. The script is available in [`trl/scripts/dpo.py`](https://github.com/huggingface/trl/blob/main/trl/scripts/dpo.py)
+
+To test the DPO script with the [Qwen2 0.5B model](https://huggingface.co/Qwen/Qwen2-0.5B-Instruct) on the [UltraFeedback dataset](https://huggingface.co/datasets/trl-lib/ultrafeedback_binarized), run the following command:
+
+```bash
+accelerate launch trl/scripts/dpo.py \
+    --model_name_or_path Qwen/Qwen2-0.5B-Instruct \
+    --dataset_name trl-lib/ultrafeedback_binarized \
+    --num_train_epochs 1 \
+    --output_dir Qwen2-0.5B-DPO
+```
+
+## Logged metrics
+
+While training and evaluating we record the following reward metrics:
+
+- `rewards/chosen`: the mean difference between the log probabilities of the policy model and the reference model for the chosen responses scaled by beta
+- `rewards/rejected`: the mean difference between the log probabilities of the policy model and the reference model for the rejected responses scaled by beta
+- `rewards/accuracies`: mean of how often the chosen rewards are > than the corresponding rejected rewards
+- `rewards/margins`: the mean difference between the chosen and corresponding rejected rewards
+
+## Loss functions
+
+The DPO algorithm supports several loss functions. The loss function can be set using the `loss_type` parameter in the [`DPOConfig`]. The following loss functions are supported:
+
+| `loss_type=` | Description |
+| --- | --- |
+| `"sigmoid"` (default) | Given the preference data, we can fit a binary classifier according to the Bradley-Terry model and in fact the [DPO](https://huggingface.co/papers/2305.18290) authors propose the sigmoid loss on the normalized likelihood via the `logsigmoid` to fit a logistic regression. |
+| `"hinge"` | The [RSO](https://huggingface.co/papers/2309.06657) authors propose to use a hinge loss on the normalized likelihood from the [SLiC](https://huggingface.co/papers/2305.10425) paper. In this case, the `beta` is the reciprocal of the margin. |
+| `"ipo"` | The [IPO](https://huggingface.co/papers/2310.12036) authors provide a deeper theoretical understanding of the DPO algorithms and identify an issue with overfitting and propose an alternative loss. In this case, the `beta` is the reciprocal of the gap between the log-likelihood ratios of the chosen vs the rejected completion pair and thus the smaller the `beta` the larger this gaps is. As per the paper the loss is averaged over log-likelihoods of the completion (unlike DPO which is summed only). |
+| `"exo_pair"` | The [EXO](https://huggingface.co/papers/2402.00856) authors propose to minimize the reverse KL instead of the negative log-sigmoid loss of DPO which corresponds to forward KL. Setting non-zero `label_smoothing` (default `1e-3`) leads to a simplified version of EXO on pair-wise preferences (see Eqn. (16) of the [EXO paper](https://huggingface.co/papers/2402.00856)). The full version of EXO uses `K>2` completions generated by the SFT policy, which becomes an unbiased estimator of the PPO objective (up to a constant) when `K` is sufficiently large. |
+| `"nca_pair"` | The [NCA](https://huggingface.co/papers/2402.05369) authors shows that NCA optimizes the absolute likelihood for each response rather than the relative likelihood. |
+| `"robust"` | The [Robust DPO](https://huggingface.co/papers/2403.00409) authors propose an unbiased estimate of the DPO loss that is robust to preference noise in the data. Like in cDPO, it assumes that the preference labels are noisy with some probability. In this approach, the `label_smoothing` parameter in the [`DPOConfig`] is used to model the probability of existing label noise. To apply this conservative loss, set `label_smoothing` to a value greater than 0.0 (between 0.0 and 0.5; the default is 0.0) |
+| `"bco_pair"` | The [BCO](https://huggingface.co/papers/2404.04656) authors train a binary classifier whose logit serves as a reward so that the classifier maps {prompt, chosen completion} pairs to 1 and {prompt, rejected completion} pairs to 0. For unpaired data, we recommend the dedicated [`BCOTrainer`]. |
+| `"sppo_hard"` | The [SPPO](https://huggingface.co/papers/2405.00675) authors claim that SPPO is capable of solving the Nash equilibrium iteratively by pushing the chosen rewards to be as large as 1/2 and the rejected rewards to be as small as -1/2 and can alleviate data sparsity issues. The implementation approximates this algorithm by employing hard label probabilities, assigning 1 to the winner and 0 to the loser. |
+| `"aot"`  or `loss_type="aot_pair"` | The [AOT](https://huggingface.co/papers/2406.05882) authors propose to use Distributional Preference Alignment Via Optimal Transport. Traditionally, the alignment algorithms use paired preferences at a sample level, which does not ensure alignment on the distributional level. AOT, on the other hand, can align LLMs on paired or unpaired preference data by making the reward distribution of the positive samples stochastically dominant in the first order on the distribution of negative samples. Specifically, `loss_type="aot"` is appropriate for paired datasets, where each prompt has both chosen and rejected responses; `loss_type="aot_pair"` is for unpaired datasets. In a nutshell, `loss_type="aot"` ensures that the log-likelihood ratio of chosen to rejected of the aligned model has higher quantiles than that ratio for the reference model. `loss_type="aot_pair"` ensures that the chosen reward is higher on all quantiles than the rejected reward. Note that in both cases quantiles are obtained via sorting. To fully leverage the advantages of the AOT algorithm, it is important to maximize the per-GPU batch size. |
+| `"apo_zero"` or `loss_type="apo_down"` | The [APO](https://huggingface.co/papers/2408.06266) method introduces an "anchored" version of the alignment objective. There are two variants: `apo_zero` and `apo_down`. The `apo_zero` loss increases the likelihood of winning outputs while decreasing the likelihood of losing outputs, making it suitable when the model is less performant than the winning outputs. On the other hand, `apo_down` decreases the likelihood of both winning and losing outputs, but with a stronger emphasis on reducing the likelihood of losing outputs. This variant is more effective when the model is better than the winning outputs. |
+| `"discopop"` | The [DiscoPOP](https://huggingface.co/papers/2406.08414) paper uses LLMs to discover more efficient offline preference optimization losses. In the paper the proposed DiscoPOP loss (which is a log-ratio modulated loss) outperformed other optimization losses on different tasks (IMDb positive text generation, Reddit TLDR summarization, and Alpaca Eval 2.0). |
+| `"sft"` | SFT (Supervised Fine-Tuning) loss is the negative log likelihood loss, used to train the model to generate preferred responses. |
+
+### Multi-loss combinations
+
+The DPO trainer supports combining multiple loss functions with different weights, enabling more sophisticated optimization strategies. This is particularly useful for implementing algorithms like MPO (Mixed Preference Optimization). MPO is a training approach that combines multiple optimization objectives, as described in the paper [Enhancing the Reasoning Ability of Multimodal Large Language Models via Mixed Preference Optimization](https://huggingface.co/papers/2411.10442).
+
+To combine multiple losses, specify the loss types and corresponding weights as lists:
+
+```python
+# MPO: Combines DPO (sigmoid) for preference and BCO (bco_pair) for quality
+training_args = DPOConfig(
+    loss_type=["sigmoid", "bco_pair", "sft"],  # Loss types to combine
+    loss_weights=[0.8, 0.2, 1.0]  # Corresponding weights, as used in the MPO paper
+)
+```
+
+If `loss_weights` is not provided, all loss types will have equal weights (1.0 by default).
+
+### Label smoothing
+
+The [cDPO](https://ericmitchell.ai/cdpo.pdf) is a tweak on the DPO loss where we assume that the preference labels are noisy with some probability. In this approach, the `label_smoothing` parameter in the [`DPOConfig`] is used to model the probability of existing label noise. To apply this conservative loss, set `label_smoothing` to a value greater than 0.0 (between 0.0 and 0.5; the default is 0.0).
+
+### Syncing the reference model
+
+The [TR-DPO](https://huggingface.co/papers/2404.09656) paper suggests syncing the reference model weights after every `ref_model_sync_steps` steps of SGD with weight `ref_model_mixup_alpha` during DPO training. To toggle this callback use the `sync_ref_model=True` in the [`DPOConfig`].
+
+### RPO loss
+
+The [RPO](https://huggingface.co/papers/2404.19733) paper implements an iterative preference tuning algorithm using a loss related to the RPO loss in this [paper](https://huggingface.co/papers/2405.16436) that essentially consists of a weighted SFT loss on the chosen preferences together with the DPO loss. To use this loss, set the `rpo_alpha` in the [`DPOConfig`] to an appropriate value. The paper suggests setting this weight to `1.0`.
+
+### WPO loss
+
+The [WPO](https://huggingface.co/papers/2406.11827) paper adapts off-policy data to resemble on-policy data more closely by reweighting preference pairs according to their probability under the current policy. To use this method, set the `use_weighting` flag to `True` in the [`DPOConfig`].
+
+### LD-DPO loss
+
+The [LD-DPO](https://huggingface.co/papers/2409.06411) paper decomposes the portion of the response that exceeds the desired length into two components — human-like preferences and verbosity preference — based on a mixing coefficient  \\( \alpha \\). To use this method, set the `ld_alpha` in the [`DPOConfig`] to an appropriate value. The paper suggests setting this value between `0.0` and `1.0`.
+
+### For Mixture of Experts Models: Enabling the auxiliary loss
+
+MOEs are the most efficient if the load is about equally distributed between experts.  
+To ensure that we train MOEs similarly during preference-tuning, it is beneficial to add the auxiliary loss from the load balancer to the final loss.
+
+This option is enabled by setting `output_router_logits=True` in the model config (e.g. [`~transformers.MixtralConfig`]).  
+To scale how much the auxiliary loss contributes to the total loss, use the hyperparameter `router_aux_loss_coef=...` (default: `0.001`) in the model config.
+
+## Accelerate DPO fine-tuning using `unsloth`
+
+You can further accelerate QLoRA / LoRA (2x faster, 60% less memory) using the [`unsloth`](https://github.com/unslothai/unsloth) library that is fully compatible with `SFTTrainer`. Currently `unsloth` supports only Llama (Yi, TinyLlama, Qwen, Deepseek etc) and Mistral architectures. Some benchmarks for DPO listed below:
+
+| GPU      | Model     | Dataset    | 🤗   | 🤗 + Flash Attention 2 | 🦥 Unsloth | 🦥 VRAM saved |
+| -------- | --------- | ---------- | --- | --------------------- | --------- | ------------ |
+| A100 40G | Zephyr 7b | Ultra Chat | 1x  | 1.24x                 | **1.88x** | -11.6%       |
+| Tesla T4 | Zephyr 7b | Ultra Chat | 1x  | 1.09x                 | **1.55x** | -18.6%       |
+
+First install `unsloth` according to the [official documentation](https://github.com/unslothai/unsloth). Once installed, you can incorporate unsloth into your workflow in a very simple manner; instead of loading `AutoModelForCausalLM`, you just need to load a `FastLanguageModel` as follows:
+
+```diff
+  from datasets import load_dataset
+  from trl import DPOConfig, DPOTrainer
+- from transformers import AutoModelForCausalLM, AutoTokenizer
++ from unsloth import FastLanguageModel
+
+- model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
+- tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
++ model, tokenizer = FastLanguageModel.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
++ model = FastLanguageModel.get_peft_model(model)
+  train_dataset = load_dataset("trl-lib/ultrafeedback_binarized", split="train")
+
+- training_args = DPOConfig(output_dir="Qwen2-0.5B-DPO")
++ training_args = DPOConfig(output_dir="Qwen2-0.5B-DPO", bf16=True)
+  trainer = DPOTrainer(model=model, args=training_args, processing_class=tokenizer, train_dataset=train_dataset)
+  trainer.train()
+
+```
+
+The saved model is fully compatible with Hugging Face's transformers library. Learn more about unsloth in their [official repository](https://github.com/unslothai/unsloth).
+
+## Reference model considerations with PEFT
+
+You have three main options (plus several variants) for how the reference model works when using PEFT, assuming the model that you would like to further enhance with DPO was tuned using (Q)LoRA.
+
+1. Simply create two instances of the model, each loading your adapter - works fine but is very inefficient.
+2. Merge the adapter into the base model, create another adapter on top, then leave the `ref_model` param null, in which case DPOTrainer will unload the adapter for reference inference - efficient, but has potential downsides discussed below.
+3. Load the adapter twice with different names, then use `set_adapter` during training to swap between the adapter being DPO'd and the reference adapter - slightly less efficient compared to 2 (~adapter size VRAM overhead), but avoids the pitfalls.
+
+### Downsides to merging QLoRA before DPO (approach 2)
+
+As suggested by [Benjamin Marie](https://medium.com/@bnjmn_marie/dont-merge-your-lora-adapter-into-a-4-bit-llm-65b6da287997), the best option for merging QLoRA adapters is to first dequantize the base model, then merge the adapter. Something similar to [this script](https://github.com/jondurbin/qlora/blob/main/qmerge.py).
+
+However, after using this approach, you will have an unquantized base model. Therefore, to use QLoRA for DPO, you will need to re-quantize the merged model or use the unquantized merge (resulting in higher memory demand).
+
+### Using option 3 - load the adapter twice
+
+To avoid the downsides with option 2, you can load your fine-tuned adapter into the model twice, with different names, and set the model/ref adapter names in [`DPOTrainer`].
+
+For example:
+
+```python
+# Load the base model.
+bnb_config = BitsAndBytesConfig(
+    load_in_4bit=True,
+    llm_int8_threshold=6.0,
+    llm_int8_has_fp16_weight=False,
+    bnb_4bit_compute_dtype=torch.bfloat16,
+    bnb_4bit_use_double_quant=True,
+    bnb_4bit_quant_type="nf4",
+)
+model = AutoModelForCausalLM.from_pretrained(
+    "mistralai/mixtral-8x7b-v0.1",
+    load_in_4bit=True,
+    quantization_config=bnb_config,
+    attn_implementation="flash_attention_2",
+    torch_dtype=torch.bfloat16,
+    device_map="auto",
+)
+model.config.use_cache = False
+
+# Load the adapter.
+model = PeftModel.from_pretrained(
+    model,
+    "/path/to/peft",
+    is_trainable=True,
+    adapter_name="train",
+)
+# Load the adapter a second time, with a different name, which will be our reference model.
+model.load_adapter("/path/to/peft", adapter_name="reference")
+
+# Initialize the trainer, without a ref_model param.
+training_args = DPOConfig(
+    model_adapter_name="train",
+    ref_adapter_name="reference",
+)
+dpo_trainer = DPOTrainer(
+    model,
+    args=training_args,
+    ...
+)
+```
+
+## DPOTrainer
+
+[[autodoc]] DPOTrainer
+
+## DPOConfig
+
+[[autodoc]] DPOConfig
+
+## DataCollatorForPreference
+
+[[autodoc]] trainer.dpo_trainer.DataCollatorForPreference

docs/source/dpo_trainer.mdx

@@ -1,297 +0,0 @@
-# DPO Trainer
-
-TRL supports the DPO Trainer for training language models from preference data, as described in the paper [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://huggingface.co/papers/2305.18290) by Rafailov et al., 2023. For a full example have a look at  [`examples/scripts/dpo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/dpo.py).
-
-The first step as always is to train your SFT model, to ensure the data we train on is in-distribution for the DPO algorithm.
-
-## How DPO works
-
-Fine-tuning a language model via DPO consists of two steps and is easier than PPO:
-
-1. **Data collection**: Gather a preference dataset with positive and negative selected pairs of generation, given a prompt.
-2. **Optimization**: Maximize the log-likelihood of the DPO loss directly.
-
-DPO-compatible datasets can be found with [the tag `dpo` on Hugging Face Hub](https://huggingface.co/datasets?other=dpo). You can also explore the [librarian-bots/direct-preference-optimization-datasets](https://huggingface.co/collections/librarian-bots/direct-preference-optimization-datasets-66964b12835f46289b6ef2fc) Collection to identify datasets that are likely to support DPO training.
-
-This process is illustrated in the sketch below (from [figure 1 of the original paper](https://huggingface.co/papers/2305.18290)):
-
-<img width="835" alt="Screenshot 2024-03-19 at 12 39 41" src="https://github.com/huggingface/trl/assets/49240599/9150fac6-3d88-4ca2-8ec6-2a6f3473216d">
-
-Read more about DPO algorithm in the [original paper](https://huggingface.co/papers/2305.18290).
-
-
-## Expected dataset format
-
-The DPO trainer expects a very specific format for the dataset. Since the model will be trained to directly optimize the preference of which sentence is the most relevant, given two sentences. We provide an example from the [`Anthropic/hh-rlhf`](https://huggingface.co/datasets/Anthropic/hh-rlhf) dataset below:
-
-<div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/rlhf-antropic-example.png", width="50%">
-</div>
-
-Therefore the final dataset object should contain these 3 entries if you use the default [`DPODataCollatorWithPadding`] data collator. The entries should be named:
-
-- `prompt`
-- `chosen`
-- `rejected`
-
-for example:
-
-```py
-dpo_dataset_dict = {
-    "prompt": [
-        "hello",
-        "how are you",
-        "What is your name?",
-        "What is your name?",
-        "Which is the best programming language?",
-        "Which is the best programming language?",
-        "Which is the best programming language?",
-    ],
-    "chosen": [
-        "hi nice to meet you",
-        "I am fine",
-        "My name is Mary",
-        "My name is Mary",
-        "Python",
-        "Python",
-        "Java",
-    ],
-    "rejected": [
-        "leave me alone",
-        "I am not fine",
-        "Whats it to you?",
-        "I dont have a name",
-        "Javascript",
-        "C++",
-        "C++",
-    ],
-}
-```
-
-where the `prompt` contains the context inputs, `chosen` contains the corresponding chosen responses and `rejected` contains the corresponding negative (rejected) responses. As can be seen a prompt can have multiple responses and this is reflected in the entries being repeated in the dictionary's value arrays.
-
-[`DPOTrainer`] can be used to fine-tune visual language models (VLMs). In this case, the dataset must also contain the key `images`, and the trainer's `tokenizer` is the VLM's `processor`. For example, for Idefics2, the processor expects the dataset to have the following format:
-
-Note: Currently, VLM support is exclusive to Idefics2 and does not extend to other VLMs.
-
-```py
-dpo_dataset_dict = {
-    'images': [
-        [Image.open('beach.jpg')],
-        [Image.open('street.jpg')],
-    ],
-    'prompt': [
-        'The image <image> shows',
-        '<image> The image depicts',
-    ],
-    'chosen': [
-        'a sunny beach with palm trees.',
-        'a busy street with several cars and buildings.',
-    ],
-    'rejected': [
-        'a snowy mountain with skiers.',
-        'a calm countryside with green fields.',
-    ],
-}
-```
-
-## Expected model format
-
-The DPO trainer expects a model of `AutoModelForCausalLM` or `AutoModelForVision2Seq`, compared to PPO that expects `AutoModelForCausalLMWithValueHead` for the value function.
-
-## Using the `DPOTrainer`
-
-For a detailed example have a look at the `examples/scripts/dpo.py` script. At a high level we need to initialize the [`DPOTrainer`] with a `model` we wish to train, a reference `ref_model` which we will use to calculate the implicit rewards of the preferred and rejected response, the `beta` refers to the hyperparameter of the implicit reward, and the dataset contains the 3 entries listed above. Note that the `model` and `ref_model` need to have the same architecture (ie decoder only or encoder-decoder).
-
-```py
-training_args = DPOConfig(
-    beta=0.1,
-)
-dpo_trainer = DPOTrainer(
-    model,
-    ref_model,
-    args=training_args,
-    train_dataset=train_dataset,
-    tokenizer=tokenizer,  # for visual language models, use tokenizer=processor instead
-)
-```
-
-After this one can then call:
-
-```py
-dpo_trainer.train()
-```
-
-Note that the `beta` is the temperature parameter for the DPO loss, typically something in the range of `0.1` to `0.5`. We ignore the reference model as `beta` -> 0.
-
-## Loss functions
-
-Given the preference data, we can fit a binary classifier according to the Bradley-Terry model and in fact the [DPO](https://huggingface.co/papers/2305.18290) authors propose the sigmoid loss on the normalized likelihood via the `logsigmoid` to fit a logistic regression. To use this loss, set the `loss_type="sigmoid"` (default) in the [`DPOConfig`].
-
-The [RSO](https://huggingface.co/papers/2309.06657) authors propose to use a hinge loss on the normalized likelihood from the [SLiC](https://huggingface.co/papers/2305.10425) paper. To use this loss, set the `loss_type="hinge"` in the [`DPOConfig`]. In this case, the `beta` is the reciprocal of the margin.
-
-The [IPO](https://huggingface.co/papers/2310.12036) authors provide a deeper theoretical understanding of the DPO algorithms and identify an issue with overfitting and propose an alternative loss. To use the loss set the `loss_type="ipo"` in the [`DPOConfig`]. In this case, the `beta` is the reciprocal of the gap between the log-likelihood ratios of the chosen vs the rejected completion pair and thus the smaller the `beta` the larger this gaps is. As per the paper the loss is averaged over log-likelihoods of the completion (unlike DPO which is summed only). 
-
-The [cDPO](https://ericmitchell.ai/cdpo.pdf) is a tweak on the DPO loss where we assume that the preference labels are noisy with some probability. In this approach, the `label_smoothing` parameter in the [`DPOConfig`] is used to model the probability of existing label noise. To apply this conservative loss, set `label_smoothing` to a value greater than 0.0 (between 0.0 and 0.5; the default is 0.0).
-
-The [EXO](https://huggingface.co/papers/2402.00856) authors propose to minimize the reverse KL instead of the negative log-sigmoid loss of DPO which corresponds to forward KL. To use the loss set the `loss_type="exo_pair"` in the [`DPOConfig`]. Setting non-zero `label_smoothing` (default `1e-3`) leads to a simplified version of EXO on pair-wise preferences (see Eqn. (16) of the [EXO paper](https://huggingface.co/papers/2402.00856)). The full version of EXO uses `K>2` completions generated by the SFT policy, which becomes an unbiased estimator of the PPO objective (up to a constant) when `K` is sufficiently large.
-
-The [NCA](https://huggingface.co/papers/2402.05369) authors shows that NCA optimizes the absolute likelihood for each response rather than the relative likelihood. To use the loss set the `loss_type="nca_pair"` in the [`DPOConfig`].
-
-The [Robust DPO](https://huggingface.co/papers/2403.00409) authors propose an unbiased estimate of the DPO loss that is robust to preference noise in the data. Like in cDPO, it assumes that the preference labels are noisy with some probability. In this approach, the `label_smoothing` parameter in the [`DPOConfig`] is used to model the probability of existing label noise. To apply this conservative loss, set `label_smoothing` to a value greater than 0.0 (between 0.0 and 0.5; the default is 0.0) and set the `loss_type="robust"` in the [`DPOConfig`].
-
-The [BCO](https://huggingface.co/papers/2404.04656) authors train a binary classifier whose logit serves as a reward so that the classifier maps {prompt, chosen completion} pairs to 1 and {prompt, rejected completion} pairs to 0. To use this loss, set the `loss_type="bco_pair"` in the [`DPOConfig`].
-
-The [TR-DPO](https://huggingface.co/papers/2404.09656) paper suggests syncing the reference model weights after every `ref_model_sync_steps` steps of SGD with weight `ref_model_mixup_alpha` during DPO training. To toggle this callback use the `sync_ref_model=True` in the [`DPOConfig`].
-
-The [RPO](https://huggingface.co/papers/2404.19733) paper implements an iterative preference tuning algorithm using a loss related to the RPO loss in this [paper](https://huggingface.co/papers/2405.16436) that essentially consists of a weighted SFT loss on the chosen preferences together with the DPO loss. To use this loss, set the `rpo_alpha` in the [`DPOConfig`] to an appropriate value. The paper suggests setting this weight to 1.0.
-
-The [SPPO](https://huggingface.co/papers/2405.00675) authors claim that SPPO is capable of solving the Nash equilibrium iteratively by pushing the chosen rewards to be as large as 1/2 and the rejected rewards to be as small as -1/2 and can alleviate data sparsity issues. The implementation approximates this algorithm by employing hard label probabilities, assigning 1 to the winner and 0 to the loser. To use this loss, set the `loss_type="sppo_hard"` in the [`DPOConfig`].
-
-The [AOT](https://huggingface.co/papers/2406.05882) authors propose to use Distributional Preference Alignment Via Optimal Transport. Traditionally, the alignment algorithms use paired preferences at a sample level, which does not ensure alignment on the distributional level. AOT, on the other hand, can align LLMs on paired or unpaired preference data by making the reward distribution of the positive samples stochastically dominant in the first order on the distribution of negative samples. Specifically, `loss_type="aot"` is appropriate for paired datasets, where each prompt has both chosen and rejected responses; `loss_type="aot_pair"` is for unpaired datasets. In a nutshell, `loss_type="aot"` ensures that the log-likelihood ratio of chosen to rejected of the aligned model has higher quantiles than that ratio for the reference model. `loss_type="aot_pair"` ensures that the chosen reward is higher on all quantiles than the rejected reward. Note that in both cases quantiles are obtained via sorting. To fully leverage the advantages of the AOT algorithm, it is important to maximize the per-GPU batch size.
-
-The [APO](https://huggingface.co/papers/2408.06266) method introduces an "anchored" version of the alignment objective. There are two variants: `apo_zero` and `apo_down`. The `apo_zero` loss increases the likelihood of winning outputs while decreasing the likelihood of losing outputs, making it suitable when the model is less performant than the winning outputs. On the other hand, `apo_down` decreases the likelihood of both winning and losing outputs, but with a stronger emphasis on reducing the likelihood of losing outputs. This variant is more effective when the model is better than the winning outputs. To use these losses, set `loss_type="apo_zero"` or `loss_type="apo_down"` in the [`DPOConfig`].
-
-### For Mixture of Experts Models: Enabling the auxiliary loss
-
-MOEs are the most efficient if the load is about equally distributed between experts.  
-To ensure that we train MOEs similarly during preference-tuning, it is beneficial to add the auxiliary loss from the load balancer to the final loss.
-
-This option is enabled by setting `output_router_logits=True` in the model config (e.g. MixtralConfig).  
-To scale how much the auxiliary loss contributes to the total loss, use the hyperparameter `router_aux_loss_coef=...` (default: 0.001).
-
-## Logging
-
-While training and evaluating we record the following reward metrics:
-
-- `rewards/chosen`: the mean difference between the log probabilities of the policy model and the reference model for the chosen responses scaled by beta
-- `rewards/rejected`: the mean difference between the log probabilities of the policy model and the reference model for the rejected responses scaled by beta
-- `rewards/accuracies`: mean of how often the chosen rewards are > than the corresponding rejected rewards
-- `rewards/margins`: the mean difference between the chosen and corresponding rejected rewards
-
-## Accelerate DPO fine-tuning using `unsloth`
-
-You can further accelerate QLoRA / LoRA (2x faster, 60% less memory) using the [`unsloth`](https://github.com/unslothai/unsloth) library that is fully compatible with `SFTTrainer`. Currently `unsloth` supports only Llama (Yi, TinyLlama, Qwen, Deepseek etc) and Mistral architectures. Some benchmarks for DPO listed below:
-
-| GPU      | Model     | Dataset    | 🤗  | 🤗 + Flash Attention 2 | 🦥 Unsloth | 🦥 VRAM saved |
-| -------- | --------- | ---------- | --- | ---------------------- | ---------- | ------------- |
-| A100 40G | Zephyr 7b | Ultra Chat | 1x  | 1.24x                  | **1.88x**  | -11.6%        |
-| Tesla T4 | Zephyr 7b | Ultra Chat | 1x  | 1.09x                  | **1.55x**  | -18.6%        |
-
-First install `unsloth` according to the [official documentation](https://github.com/unslothai/unsloth). Once installed, you can incorporate unsloth into your workflow in a very simple manner; instead of loading `AutoModelForCausalLM`, you just need to load a `FastLanguageModel` as follows:
-
-```python
-import torch
-from trl import DPOConfig, DPOTrainer
-from unsloth import FastLanguageModel
-
-max_seq_length = 2048 # Supports automatic RoPE Scaling, so choose any number.
-
-# Load model
-model, tokenizer = FastLanguageModel.from_pretrained(
-    model_name = "unsloth/zephyr-sft",
-    max_seq_length = max_seq_length,
-    dtype = None, # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
-    load_in_4bit = True, # Use 4bit quantization to reduce memory usage. Can be False.
-    # token = "hf_...", # use one if using gated models like meta-llama/Llama-2-7b-hf
-)
-
-# Do model patching and add fast LoRA weights
-model = FastLanguageModel.get_peft_model(
-    model,
-    r = 16,
-    target_modules = ["q_proj", "k_proj", "v_proj", "o_proj",
-                      "gate_proj", "up_proj", "down_proj",],
-    lora_alpha = 16,
-    lora_dropout = 0, # Dropout = 0 is currently optimized
-    bias = "none",    # Bias = "none" is currently optimized
-    use_gradient_checkpointing = True,
-    random_state = 3407,
-)
-
-training_args = DPOConfig(
-    output_dir="./output",
-    beta=0.1,
-)
-
-dpo_trainer = DPOTrainer(
-    model,
-    ref_model=None,
-    args=training_args,
-    train_dataset=train_dataset,
-    tokenizer=tokenizer,
-)
-dpo_trainer.train()
-```
-
-The saved model is fully compatible with Hugging Face's transformers library. Learn more about unsloth in their [official repository](https://github.com/unslothai/unsloth).
-
-## Reference model considerations with PEFT
-
-You have three main options (plus several variants) for how the reference model works when using PEFT, assuming the model that you would like to further enhance with DPO was tuned using (Q)LoRA.
-
-1. Simply create two instances of the model, each loading your adapter - works fine but is very inefficient.
-2. Merge the adapter into the base model, create another adapter on top, then leave the `ref_model` param null, in which case DPOTrainer will unload the adapter for reference inference - efficient, but has potential downsides discussed below.
-3. Load the adapter twice with different names, then use `set_adapter` during training to swap between the adapter being DPO'd and the reference adapter - slightly less efficient compared to 2 (~adapter size VRAM overhead), but avoids the pitfalls.
-
-### Downsides to merging QLoRA before DPO (approach 2)
-
-As suggested by [Benjamin Marie](https://medium.com/@bnjmn_marie/dont-merge-your-lora-adapter-into-a-4-bit-llm-65b6da287997), the best option for merging QLoRA adapters is to first dequantize the base model, then merge the adapter. Something similar to [this script](https://github.com/jondurbin/qlora/blob/main/qmerge.py).
-
-However, after using this approach, you will have an unquantized base model. Therefore, to use QLoRA for DPO, you will need to re-quantize the merged model or use the unquantized merge (resulting in higher memory demand).
-
-### Using option 3 - load the adapter twice
-
-To avoid the downsides with option 2, you can load your fine-tuned adapter into the model twice, with different names, and set the model/ref adapter names in [`DPOTrainer`].
-
-For example:
-
-```python
-# Load the base model.
-bnb_config = BitsAndBytesConfig(
-    load_in_4bit=True,
-    llm_int8_threshold=6.0,
-    llm_int8_has_fp16_weight=False,
-    bnb_4bit_compute_dtype=torch.bfloat16,
-    bnb_4bit_use_double_quant=True,
-    bnb_4bit_quant_type="nf4",
-)
-model = AutoModelForCausalLM.from_pretrained(
-    "mistralai/mixtral-8x7b-v0.1",
-    load_in_4bit=True,
-    quantization_config=bnb_config,
-    attn_implementation="flash_attention_2",
-    torch_dtype=torch.bfloat16,
-    device_map="auto",
-)
-model.config.use_cache = False
-
-# Load the adapter.
-model = PeftModel.from_pretrained(
-    model,
-    "/path/to/peft",
-    is_trainable=True,
-    adapter_name="train",
-)
-# Load the adapter a second time, with a different name, which will be our reference model.
-model.load_adapter("/path/to/peft", adapter_name="reference")
-
-# Initialize the trainer, without a ref_model param.
-training_args = DPOConfig(
-    model_adapter_name="train",
-    ref_adapter_name="reference",
-)
-dpo_trainer = DPOTrainer(
-    model,
-    args=training_args,
-    ...
-)
-```
-
-## DPOTrainer
-
-[[autodoc]] DPOTrainer
-
-## DPOConfig
-
-[[autodoc]] DPOConfig

docs/source/example_overview.md

@@ -31,32 +31,39 @@ Then, it is encouraged to launch jobs with `accelerate launch`!
 
 # Maintained Examples
 
+Scripts can be used as examples of how to use TRL trainers. They are located in the [`trl/scripts`](https://github.com/huggingface/trl/blob/main/trl/scripts) directory. Additionally, we provide examples in the [`examples/scripts`](https://github.com/huggingface/trl/blob/main/examples/scripts) directory. These examples are maintained and tested regularly.
 
-
-| File                                                                                                                          | Description                                                                                                                                                                                                                                                                                                     |
-| ----------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| [`examples/scripts/alignprop.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/alignprop.py)                 | This script shows how to use the [`AlignPropTrainer`] to fine-tune a diffusion model.                                                                                                                                                                                                                           |
-| [`examples/scripts/bco.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/bco.py)                             | This script shows how to use the [`KTOTrainer`] with the BCO loss to fine-tune a model to increase instruction-following, truthfulness, honesty and helpfulness using the [openbmb/UltraFeedback](https://huggingface.co/datasets/openbmb/UltraFeedback) dataset.                                               |
-| [`examples/scripts/chat.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/chat.py)                           | This script allows you to load and use a model as a chatbot.                                                                                                                                                                                                                                                    |
-| [`examples/scripts/cpo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/cpo.py)                             | This script shows how to use the [`CPOTrainer`] to fine-tune a model to increase helpfulness and harmlessness using the [Anthropic/hh-rlhf](https://huggingface.co/datasets/Anthropic/hh-rlhf) dataset.                                                                                                         |
-| [`examples/scripts/ddpo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/ddpo.py)                           | This script shows how to use the [`DDPOTrainer`] to fine-tune a stable diffusion model using reinforcement learning.                                                                                                                                                                                            |
-| [`examples/scripts/dpo_visual.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/dpo_visual.py)               | This script shows how to use the [`DPOTrainer`] to fine-tune a Vision Language Model to reduce hallucinations using the [openbmb/RLAIF-V-Dataset](https://huggingface.co/datasets/openbmb/RLAIF-V-Dataset) dataset.                                                                                             |
-| [`examples/scripts/dpo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/dpo.py)                             | This script shows how to use the [`DPOTrainer`] to fine-tune a stable to increase helpfulness and harmlessness using the [Anthropic/hh-rlhf](https://huggingface.co/datasets/Anthropic/hh-rlhf) dataset.                                                                                                        |
-| [`examples/scripts/kto.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/kto.py)                             | This script shows how to use the [`KTOTrainer`] to fine-tune a model.                                                                                                                                                                                                                                           |
-| [`examples/scripts/orpo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/orpo.py)                           | This script shows how to use the [`ORPOTrainer`] to fine-tune a model to increase helpfulness and harmlessness using the [Anthropic/hh-rlhf](https://huggingface.co/datasets/Anthropic/hh-rlhf) dataset.                                                                                                        |
-| [`examples/scripts/ppo_multi_adapter.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/ppo_multi_adapter.py) | This script shows how to use the [`PPOTrainer`] to train a single base model with multiple adapters. Requires you to run the example script with the reward model training beforehand.                                                                                                                          |
-| [`examples/scripts/ppo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/ppo.py)                             | This script shows how to use the [`PPOTrainer`] to fine-tune a sentiment analysis model using [IMDB dataset](https://huggingface.co/datasets/stanfordnlp/imdb).                                                                                                                                                 |
-| [`examples/scripts/reward_modeling.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/reward_modeling.py)     | This script shows how to use the [`RewardTrainer`] to train a reward model on your own dataset.                                                                                                                                                                                                                 |
-| [`examples/scripts/sft.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/sft.py)                             | This script shows how to use the [`SFTTrainer`] to fine-tune a model or adapters into a target dataset.                                                                                                                                                                                                         |
-| [`examples/scripts/vsft_llava.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/vsft_llava.py)               | This script shows how to use the [`SFTTrainer`] to fine-tune a Vision Language Model in a chat setting. The script has only been tested on a [LLaVA 1.5]([llava-hf/llava-1.5-7b-hf](https://huggingface.co/llava-hf/llava-1.5-7b-hf)) model so users may see unexpected behaviour in other model architectures. |
+| File | Description |
+| --- | --- |
+| [`examples/scripts/alignprop.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/alignprop.py) | This script shows how to use the [`AlignPropTrainer`] to fine-tune a diffusion model. |
+| [`examples/scripts/bco.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/bco.py) | This script shows how to use the [`KTOTrainer`] with the BCO loss to fine-tune a model to increase instruction-following, truthfulness, honesty and helpfulness using the [openbmb/UltraFeedback](https://huggingface.co/datasets/openbmb/UltraFeedback) dataset. |
+| [`examples/scripts/cpo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/cpo.py) | This script shows how to use the [`CPOTrainer`] to fine-tune a model to increase helpfulness and harmlessness using the [Anthropic/hh-rlhf](https://huggingface.co/datasets/Anthropic/hh-rlhf) dataset. |
+| [`examples/scripts/ddpo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/ddpo.py) | This script shows how to use the [`DDPOTrainer`] to fine-tune a stable diffusion model using reinforcement learning. |
+| [`examples/scripts/dpo_online.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/dpo_online.py) | This script shows how to use the [`OnlineDPOTrainer`] to fine-tune a model. |
+| [`examples/scripts/dpo_vlm.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/dpo_vlm.py) | This script shows how to use the [`DPOTrainer`] to fine-tune a Vision Language Model to reduce hallucinations using the [openbmb/RLAIF-V-Dataset](https://huggingface.co/datasets/openbmb/RLAIF-V-Dataset) dataset. |
+| [`examples/scripts/gkd.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/gkd.py) | This script shows how to use the [`GKDTrainer`] to fine-tune a model. |
+| [`examples/scripts/nash_md.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/nash_md.py) | This script shows how to use the [`NashMDTrainer`] to fine-tune a model. |
+| [`examples/scripts/orpo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/orpo.py) | This script shows how to use the [`ORPOTrainer`] to fine-tune a model to increase helpfulness and harmlessness using the [Anthropic/hh-rlhf](https://huggingface.co/datasets/Anthropic/hh-rlhf) dataset. |
+| [`examples/scripts/ppo/ppo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/ppo/ppo.py) | This script shows how to use the [`PPOTrainer`] to fine-tune a model to improve its ability to continue text with positive sentiment or physically descriptive language |
+| [`examples/scripts/ppo/ppo_tldr.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/ppo/ppo_tldr.py) | This script shows how to use the [`PPOTrainer`] to fine-tune a model to improve its ability to generate TL;DR summaries. |
+| [`examples/scripts/prm.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/prm.py) | This script shows how to use the [`PRMTrainer`] to fine-tune a Process-supervised Reward Model (PRM). |
+| [`examples/scripts/reward_modeling.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/reward_modeling.py) | This script shows how to use the [`RewardTrainer`] to train a Outcome Reward Model (ORM) on your own dataset. |
+| [`examples/scripts/rloo/rloo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/rloo/rloo.py) | This script shows how to use the [`RLOOTrainer`] to fine-tune a model to improve its ability to continue text with positive sentiment or physically descriptive language |
+| [`examples/scripts/rloo/rloo_tldr.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/rloo/rloo_tldr.py) | This script shows how to use the [`RLOOTrainer`] to fine-tune a model to improve its ability to generate TL;DR summaries. |
+| [`examples/scripts/sft_gemma3.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/sft_gemma3.py) | This script shows how to use the [`SFTTrainer`] to fine-tune a Gemma 3 model. |
+| [`examples/scripts/sft_video_llm.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/sft_video_llm.py) | This script shows how to use the [`SFTTrainer`] to fine-tune a Video Language Model. |
+| [`examples/scripts/sft_vlm_gemma3.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/sft_vlm_gemma3.py) | This script shows how to use the [`SFTTrainer`] to fine-tune a Gemma 3 model on vision to text tasks. |
+| [`examples/scripts/sft_vlm_smol_vlm.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/sft_vlm_smol_vlm.py) | This script shows how to use the [`SFTTrainer`] to fine-tune a SmolVLM model. |
+| [`examples/scripts/sft_vlm.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/sft_vlm.py) | This script shows how to use the [`SFTTrainer`] to fine-tune a Vision Language Model in a chat setting. The script has only been tested with [LLaVA 1.5](https://huggingface.co/llava-hf/llava-1.5-7b-hf), [LLaVA 1.6](https://huggingface.co/llava-hf/llava-v1.6-mistral-7b-hf), and [Llama-3.2-11B-Vision-Instruct](https://huggingface.co/meta-llama/Llama-3.2-11B-Vision-Instruct) models so users may see unexpected behaviour in other model architectures. |
+| [`examples/scripts/xpo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/xpo.py) | This script shows how to use the [`XPOTrainer`] to fine-tune a model. |
 
 Here are also some easier-to-run colab notebooks that you can use to get started with TRL:
 
-| File                                                                                                                              | Description                                                                                                             |
-| --------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------- |
-| [`examples/notebooks/best_of_n.ipynb`](https://github.com/huggingface/trl/tree/main/examples/notebooks/best_of_n.ipynb)           | This notebook demonstrates how to use the "Best of N" sampling strategy using TRL when fine-tuning your model with PPO. |
-| [`examples/notebooks/gpt2-sentiment.ipynb`](https://github.com/huggingface/trl/tree/main/examples/notebooks/gpt2-sentiment.ipynb) | This notebook demonstrates how to reproduce the GPT2 imdb sentiment tuning example on a jupyter notebook.               |
-| [`examples/notebooks/gpt2-control.ipynb`](https://github.com/huggingface/trl/tree/main/examples/notebooks/gpt2-control.ipynb)     | This notebook demonstrates how to reproduce the GPT2 sentiment control example on a jupyter notebook.                   |
+| File | Description |
+| --- | --- |
+| [`examples/notebooks/best_of_n.ipynb`](https://github.com/huggingface/trl/tree/main/examples/notebooks/best_of_n.ipynb) | This notebook demonstrates how to use the "Best of N" sampling strategy using TRL when fine-tuning your model with PPO. |
+| [`examples/notebooks/gpt2-sentiment.ipynb`](https://github.com/huggingface/trl/tree/main/examples/notebooks/gpt2-sentiment.ipynb) | This notebook demonstrates how to reproduce the GPT2 imdb sentiment tuning example on a jupyter notebook. |
+| [`examples/notebooks/gpt2-control.ipynb`](https://github.com/huggingface/trl/tree/main/examples/notebooks/gpt2-control.ipynb) | This notebook demonstrates how to reproduce the GPT2 sentiment control example on a jupyter notebook. |
 
 
 We also have some other examples that are less maintained but can be used as a reference:

docs/source/gkd_trainer.md

@@ -0,0 +1,98 @@
+# Generalized Knowledge Distillation Trainer
+
+[![](https://img.shields.io/badge/All_models-GKD-blue)](https://huggingface.co/models?other=gkd,trl)
+
+## Overview
+
+Generalized Knowledge Distillation (GKD) was proposed in [On-Policy Distillation of Language Models: Learning from Self-Generated Mistakes](https://huggingface.co/papers/2306.13649) by Rishabh Agarwal, Nino Vieillard, Yongchao Zhou, Piotr Stanczyk, Sabela Ramos, Matthieu Geist, and Olivier Bachem. 
+
+The abstract from the paper is the following:
+
+> Knowledge distillation (KD) is widely used for compressing a teacher model to reduce its inference cost and memory footprint, by training a smaller student model. However, current KD methods for auto-regressive sequence models suffer from distribution mismatch between output sequences seen during training and those generated by the student during inference. To address this issue, we introduce Generalized Knowledge Distillation (GKD). Instead of solely relying on a fixed set of output sequences, GKD trains the student on its self-generated output sequences by leveraging feedback from the teacher on such sequences. Unlike supervised KD approaches, GKD also offers the flexibility to employ alternative loss functions between the student and teacher, which can be useful when the student lacks the expressivity to mimic the teacher's distribution. Furthermore, GKD facilitates the seamless integration of distillation with RL fine-tuning (RLHF). We demonstrate the efficacy of GKD for distilling auto-regressive language models on summarization, translation, and arithmetic reasoning tasks, and task-agnostic distillation for instruction-tuning.
+
+
+The key aspects of GKD are:
+1. It addresses the train-inference distribution mismatch in auto-regressive sequence models by training the student model on its self-generated output sequences.
+2. GKD allows flexibility in choosing different divergence measures between student and teacher models via the generalized Jensen-Shannon Divergence (JSD), which can be useful when the student lacks the capacity to fully mimic the teacher.
+
+This post-training method was contributed by [Kashif Rasul](https://huggingface.co/kashif) and [Lewis Tunstall](https://huggingface.co/lewtun).
+
+## Usage tips
+
+The [`GKDTrainer`] is a wrapper around the [`SFTTrainer`] class that takes in a teacher model argument. It needs three parameters to be set via the [`GKDConfig`] namely:
+* `lmbda`:  controls the student data fraction, i.e., the proportion of on-policy student-generated outputs. When `lmbda=0.0`, the loss reduces to supervised JSD where the student is trained with the token-level probabilities of the teacher. When `lmbda=1.0`, the loss reduces to on-policy JSD, where the student generates output sequences and token-specific feedback on these sequences from the teacher. For values in between [0, 1] it is random between the two based on the `lmbda` value for each batch.
+* `seq_kd`:  controls whether to perform Sequence-Level KD (can be viewed as supervised FT on teacher-generated out). When `seq_kd=True` and `lmbda=0.0`, the loss reduces to supervised JSD, where the teacher generates output sequences and the student receives token-specific feedback on these sequences from the teacher. 
+* `beta`: controls the interpolation in the generalized Jensen-Shannon Divergence.  When `beta=0.0` the loss approximates forward KL divergence, while for `beta=1.0` the loss approximates reverse KL divergence. For values in between [0, 1] it interpolates between the two.
+
+The authors find that on-policy data (high `lmbda`) performs better and the optimal `beta` varied depending on the task and evaluation method.
+
+> [!WARNING]
+> Make sure that `attn_implementation="flash_attention_2"` when training [Gemma models](https://huggingface.co/models?other=gemma2). Otherwise you will encounter NaNs in the logits due to the [soft capping technique](https://huggingface.co/blog/gemma2#soft-capping-and-attention-implementations) adopted by this architecture.
+
+The basic API is as follows:
+
+```python
+from datasets import Dataset
+from trl import GKDConfig, GKDTrainer
+from transformers import (
+    AutoModelForCausalLM,
+    AutoTokenizer,
+)
+
+NUM_DUMMY_SAMPLES = 100
+
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
+# The model to optimise
+model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
+# The teacher model to calculate the KL divergence against
+teacher_model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-1.5B-Instruct")
+
+train_dataset = Dataset.from_dict(
+    {
+        "messages": [
+            [
+                {"role": "user", "content": "Hi, how are you?"},
+                {"role": "assistant", "content": "I'm great thanks"},
+            ]
+        ]
+        * NUM_DUMMY_SAMPLES
+    }
+)
+eval_dataset = Dataset.from_dict(
+    {
+        "messages": [
+            [
+                {"role": "user", "content": "What colour is the sky?"},
+                {"role": "assistant", "content": "The sky is blue"},
+            ]
+        ]
+        * NUM_DUMMY_SAMPLES
+    }
+)
+
+training_args = GKDConfig(output_dir="gkd-model", per_device_train_batch_size=1)
+trainer = GKDTrainer(
+    model=model,
+    teacher_model=teacher_model,
+    args=training_args,
+    processing_class=tokenizer,
+    train_dataset=train_dataset,
+    eval_dataset=eval_dataset,
+)
+trainer.train()
+```
+
+### Expected dataset type
+
+The dataset should be formatted as a list of "messages" where each message is a list of dictionaries with the following keys:
+* `role`: either `system`, `assistant` or `user`
+* `content`: the message content
+
+
+## GKDTrainer
+
+[[autodoc]] GKDTrainer
+
+## GKDConfig
+
+[[autodoc]] GKDConfig

docs/source/grpo_trainer.md

@@ -0,0 +1,597 @@
+# GRPO Trainer
+
+[![](https://img.shields.io/badge/All_models-GRPO-blue)](https://huggingface.co/models?other=grpo,trl)
+
+## Overview
+
+TRL supports the GRPO Trainer for training language models, as described in the paper [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300) by [Zhihong Shao](https://huggingface.co/syhia), [Peiyi Wang](https://huggingface.co/peiyiwang89), [Qihao Zhu](https://huggingface.co/zqh11), Runxin Xu, [Junxiao Song](https://huggingface.co/haha-point), Mingchuan Zhang, Y. K. Li, Y. Wu, [Daya Guo](https://huggingface.co/guoday).
+
+The abstract from the paper is the following:
+
+> Mathematical reasoning poses a significant challenge for language models due to its complex and structured nature. In this paper, we introduce DeepSeekMath 7B, which continues pre-training DeepSeek-Coder-Base-v1.5 7B with 120B math-related tokens sourced from Common Crawl, together with natural language and code data. DeepSeekMath 7B has achieved an impressive score of 51.7% on the competition-level MATH benchmark without relying on external toolkits and voting techniques, approaching the performance level of Gemini-Ultra and GPT-4. Self-consistency over 64 samples from DeepSeekMath 7B achieves 60.9% on MATH. The mathematical reasoning capability of DeepSeekMath is attributed to two key factors: First, we harness the significant potential of publicly available web data through a meticulously engineered data selection pipeline. Second, we introduce Group Relative Policy Optimization (GRPO), a variant of Proximal Policy Optimization (PPO), that enhances mathematical reasoning abilities while concurrently optimizing the memory usage of PPO.
+
+This post-training method was contributed by [Quentin Gallouédec](https://huggingface.co/qgallouedec).
+
+## Quick start
+
+This example demonstrates how to train a model using the GRPO method. We train a [Qwen 0.5B Instruct model](https://huggingface.co/Qwen/Qwen2-0.5B-Instruct) with the prompts from the [TLDR dataset](https://huggingface.co/datasets/trl-lib/tldr) (completion column is ignored!). You can view the data in the dataset here:
+
+<iframe
+  src="https://huggingface.co/datasets/trl-lib/tldr/embed/viewer/default/train?row=0"
+  frameborder="0"
+  width="100%"
+  height="560px"
+></iframe>
+
+Below is the script to train the model.
+
+```python
+# train_grpo.py
+from datasets import load_dataset
+from trl import GRPOConfig, GRPOTrainer
+
+dataset = load_dataset("trl-lib/tldr", split="train")
+
+# Define the reward function, which rewards completions that are close to 20 characters
+def reward_len(completions, **kwargs):
+    return [-abs(20 - len(completion)) for completion in completions]
+
+training_args = GRPOConfig(output_dir="Qwen2-0.5B-GRPO")
+trainer = GRPOTrainer(
+    model="Qwen/Qwen2-0.5B-Instruct",
+    reward_funcs=reward_len,
+    args=training_args,
+    train_dataset=dataset,
+)
+trainer.train()
+```
+
+Execute the script using the following command:
+
+```bash
+accelerate launch train_grpo.py
+```
+
+Distributed across 8 GPUs, the training takes approximately 1 day.
+
+![](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/grpo_curves.png)
+
+## Looking deeper into the GRPO method
+
+GRPO is an online learning algorithm, meaning it improves iteratively by using the data generated by the trained model itself during training. The intuition behind GRPO objective is to maximize the advantage of the generated completions, while ensuring that the model remains close to the reference policy. To understand how GRPO works, it can be broken down into four main steps: **Generating completions**, **computing the advantage**, **estimating the KL divergence**, and **computing the loss**.
+
+![](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/grpo_visual.png)
+
+### Generating completions
+
+At each training step, we sample a batch of prompts and generate a set of  \\( G \\) completions for each prompt (denoted as  \\( o_i \\)).
+
+### Computing the advantage
+
+For each of the  \\( G \\) sequences, we compute the reward using a reward model. To align with the comparative nature of reward models—typically trained on datasets of comparisons between outputs for the same question—the advantage is calculated to reflect these relative comparisons. It is normalized as follows:
+
+$$\hat{A}_{i,t} = \frac{r_i - \text{mean}(\mathbf{r})}{\text{std}(\mathbf{r})}$$
+
+This approach gives the method its name: **Group Relative Policy Optimization (GRPO)**.
+
+<Tip>
+
+It was shown in the paper [Understanding R1-Zero-Like Training: A Critical Perspective](https://huggingface.co/papers/2503.20783) that scaling by  \\( \text{std}(\mathbf{r}) \\) may cause a question-level difficulty bias. You can disable this scaling by setting `scale_rewards=False` in [`GRPOConfig`].
+
+</Tip>
+
+### Estimating the KL divergence
+
+KL divergence is estimated using the approximator introduced by [Schulman et al. (2020)](http://joschu.net/blog/kl-approx.html). The approximator is defined as follows:
+
+$$\mathbb{D}_{\text{KL}}\left[\pi_\theta \|\pi_{\text{ref}}\right] = \frac{\pi_{\text{ref}}(o_{i,t} \mid q, o_{i,<t})}{\pi_\theta(o_{i,t} \mid q, o_{i,<t})} - \log \frac{\pi_{\text{ref}}(o_{i,t} \mid q, o_{i,<t})}{\pi_\theta(o_{i,t} \mid q, o_{i,<t})} - 1,
+$$
+
+### Computing the loss
+
+The objective is to maximize the advantage while ensuring that the model remains close to the reference policy. Consequently, the loss is defined as follows:
+
+$$
+\mathcal{L}_{\text{GRPO}}(\theta) = -\frac{1}{\sum_{i=1}^G |o_i|} \sum_{i=1}^G \sum_{t=1}^{|o_i|} \left[ \frac{\pi_\theta(o_{i,t} \mid q, o_{i,< t})}{\left[\pi_\theta(o_{i,t} \mid q, o_{i,< t})\right]_{\text{no grad}}} \hat{A}_{i,t} - \beta \mathbb{D}_{\text{KL}}\left[\pi_\theta \| \pi_{\text{ref}}\right] \right],
+$$
+
+where the first term represents the scaled advantage and the second term penalizes deviations from the reference policy through KL divergence.
+
+<Tip>
+
+Note that compared to the original formulation in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300), we don't scale by  \\( \frac{1}{|o_i|} \\) because it was shown in the paper [Understanding R1-Zero-Like Training: A Critical Perspective](https://huggingface.co/papers/2503.20783) that this introduces a response-level length bias. More details in [loss types](#loss-types).
+
+</Tip>
+
+<Tip>
+
+Note that compared to the original formulation in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300), we use  \\( \beta = 0.0 \\) by default, meaning that the KL divergence term is not used. This choice is motivated by several recent studies (e.g., [Open-Reasoner-Zero: An Open Source Approach to Scaling Up Reinforcement Learning on the Base Model](https://huggingface.co/papers/2503.24290)) which have shown that the KL divergence term is not essential for training with GRPO. As a result, it has become common practice to exclude it (e.g. [Understanding R1-Zero-Like Training: A Critical Perspective](https://huggingface.co/papers/2503.20783), [DAPO: An Open-Source LLM Reinforcement Learning System at Scale](https://huggingface.co/papers/2503.14476)). If you wish to include the KL divergence term, you can set `beta` in [`GRPOConfig`] to a non-zero value.
+
+</Tip>
+
+In the original paper, this formulation is generalized to account for multiple updates after each generation (denoted  \\( \mu \\), can be set with `num_iterations` in [`GRPOConfig`]) by leveraging the **clipped surrogate objective**:
+
+$$
+\mathcal{L}_{\text{GRPO}}(\theta) = - \frac{1}{\sum_{i=1}^G |o_i|} \sum_{i=1}^G \sum_{t=1}^{|o_i|} \left[ \min \left( \frac{\pi_\theta(o_{i,t} \mid q, o_{i,< t})}{\pi_{\theta_{\text{old}}}(o_{i,t} \mid q, o_{i,< t})} \hat{A}_{i,t}, \, \text{clip}\left( \frac{\pi_\theta(o_{i,t} \mid q, o_{i,< t})}{\pi_{\theta_{\text{old}}}(o_{i,t} \mid q, o_{i,< t})}, 1 - \epsilon, 1 + \epsilon \right) \hat{A}_{i,t} \right) - \beta \mathbb{D}_{\text{KL}}\left[\pi_\theta \| \pi_{\text{ref}}\right] \right],
+$$
+
+where  \\(\text{clip}(\cdot, 1 - \epsilon, 1 + \epsilon) \\) ensures that updates do not deviate excessively from the reference policy by bounding the policy ratio between  \\( 1 - \epsilon \\) and  \\( 1 + \epsilon \\).
+When  \\( \mu = 1 \\) (default in TRL), the clipped surrogate objective simplifies to the original objective.
+
+#### Loss Types
+
+Several formulations of the objective have been proposed in the literature. Initially, the objective of GRPO was defined as follows:
+
+$$
+\mathcal{L}_{\text{GRPO}}(\theta) = - \frac{1}{G} \sum_{i=1}^G \frac{1}{|o_i|} \sum_{t=1}^{|o_i|} l_{i,t},
+$$
+
+where
+
+$$
+l_{i,t} = \frac{\pi_\theta(o_{i,t} \mid q, o_{i,< t})}{\left[\pi_\theta(o_{i,t} \mid q, o_{i,< t})\right]_{\text{no grad}}} \hat{A}_{i,t} - \beta \mathbb{D}_{\text{KL}}\left[\pi_\theta \| \pi_{\text{ref}}\right].
+$$
+
+The [DAPO paper](https://huggingface.co/papers/2503.14476) highlights the limitations of the GRPO algorithm’s sample-level loss in long-CoT scenarios, where longer responses are under-penalized, leading to poorer quality outputs. The proposed solution is a token-level normalization, which better handles longer sequences by assigning more balanced rewards to individual tokens, regardless of response length:
+
+$$
+\mathcal{L}_{\text{DAPO}}(\theta) = - \frac{1}{\sum_{i=1}^G |o_i|} \sum_{i=1}^G \sum_{t=1}^{|o_i|} l_{i,t},
+$$
+
+
+Furthermore, it was demonstrated in the paper [Understanding R1-Zero-Like Training: A Critical Perspective](https://huggingface.co/papers/2503.20783) that the initial GRPO formulation introduces a response length bias. They show that while the DAPO formulation reduces this bias, it does not eliminate it completely. To fully remove this bias, they propose dividing by a constant instead of the sequence length, resulting in the following formulation:
+
+$$
+\mathcal{L}_{\text{Dr. GRPO}}(\theta) = - \frac{1}{LG} \sum_{i=1}^G \sum_{t=1}^{|o_i|} l_{i,t},
+$$
+
+This constant is recommended to be the maximum completion length. To use this formulation, set `loss_type="dr_grpo"` in the [`GRPOConfig`].
+
+## Logged metrics
+
+- `num_tokens`: The total number of tokens processed so far, including both prompts and completions.
+- `completions/mean_length`: The average length of generated completions.
+- `completions/min_length`: The minimum length of generated completions.
+- `completions/max_length`: The maximum length of generated completions.
+- `completions/mean_terminated_length`: The average length of generated completions that terminate with EOS.
+- `completions/min_terminated_length`: The minimum length of generated completions that terminate with EOS.
+- `completions/max_terminated_length`: The maximum length of generated completions that terminate with EOS.
+- `completions/clipped_ratio` : The ratio of truncated (clipped) completions.
+- `reward/{reward_func_name}/mean`: The average reward from a specific reward function.
+- `reward/{reward_func_name}/std`: The standard deviation of the reward from a specific reward function.
+- `reward`: The overall average reward after applying reward weights.
+- `reward_std`: The standard deviation of the overall reward within each batch after applying reward weights.
+- `frac_reward_zero_std`: The fraction of samples in the generation batch with a reward std of zero, implying there is little diversity for that prompt (all answers are correct or incorrect).
+- `entropy`: Average entropy of token predictions across generated completions. (If `mask_truncated_completions=True`, masked sequences tokens are excluded.)
+- `kl`: The average KL divergence between the model and the reference model, calculated over generated completions. Logged only if `beta` is nonzero.
+- `clip_ratio/region_mean`: The ratio of token (or sequence, if `importance_sampling_level="sequence"`) probabilities where the GRPO objective is clipped to stay within the trust region:
+$$
+\text{clip}\left( r_{i,t}(\theta), 1 - \epsilon_\mathrm{low}, 1 + \epsilon_\mathrm{high} \right)\,, \qquad r_{i,t}(\theta) = \frac{\pi_\theta(o_{i,t} \mid q, o_{i,< t})}{\pi_{\theta_{\text{old}}}(o_{i,t} \mid q, o_{i,< t})}\,.
+$$
+A higher value means more tokens are clipped, which constrains how much the policy $\pi_\theta$ can change.
+- `clip_ratio/low_mean`: The average ratio of token (or sequence, if `importance_sampling_level="sequence"`) probabilities that were clipped on the lower bound of the trust region:  \\(r_{i,t}(\theta) < 1 - \epsilon_\mathrm{low}\\)
+- `clip_ratio/low_min`: The minimum ratio of token (or sequence, if `importance_sampling_level="sequence"`) probabilities that were clipped on the lower bound of the trust region:  \\(r_{i,t}(\theta) < 1 - \epsilon_\mathrm{low}\\)
+- `clip_ratio/high_mean`: The average ratio of token (or sequence, if `importance_sampling_level="sequence"`) probabilities that were clipped on the upper bound of the trust region:  \\(r_{i,t}(\theta) > 1 + \epsilon_\mathrm{high}\\)
+- `clip_ratio/high_max`: The maximum ratio of token (or sequence, if `importance_sampling_level="sequence"`) probabilities that were clipped on the upper bound of the trust region:  \\(r_{i,t}(\theta) > 1 + \epsilon_\mathrm{high}\\).
+
+## Customization
+
+### Speed up training with vLLM-powered generation
+
+Generation is often the main bottleneck when training with online methods. To accelerate generation, you can use [vLLM](https://github.com/vllm-project/vllm), a high-throughput, low-latency inference engine for LLMs. To enable it, first install the package with
+```shell
+pip install trl[vllm]
+```
+
+We support two ways of using vLLM during training: **server mode** and **colocate mode**.
+
+#### 🔌 Option 1: Server mode
+
+In this mode, vLLM runs in a separate process (and using separate GPUs) and communicates with the trainer via HTTP. This is ideal if you have dedicated GPUs for inference.
+
+1. **Start the vLLM server**:
+   ```bash
+   trl vllm-serve --model <model_name>
+   ```
+
+2. **Enable server mode in your training script**:
+   ```python
+   from trl import GRPOConfig
+
+   training_args = GRPOConfig(
+       ...,
+       use_vllm=True,
+       vllm_mode="server",  # default value, can be omitted
+   )
+   ```
+
+<Tip warning={true}>
+
+Make sure that the server is using different GPUs than the trainer, otherwise you may run into NCCL errors. You can specify the GPUs to use with the `CUDA_VISIBLE_DEVICES` environment variable.
+
+</Tip>
+
+#### 🧩 Option 2: Colocate mode
+
+In this mode, vLLM runs inside the trainer process and shares GPU memory with the training model. This avoids launching a separate server and can improve GPU utilization, but may lead to memory contention on the training GPUs.
+
+```python
+from trl import GRPOConfig
+
+training_args = GRPOConfig(
+    ...,
+    use_vllm=True,
+    vllm_mode="colocate",
+)
+```
+
+<Tip>
+
+Depending on the model size and the overall GPU memory requirements for training, you may need to adjust the `vllm_gpu_memory_utilization` parameter in [`GRPOConfig`] to avoid underutilization or out-of-memory errors.
+
+We provide a [HF Space](https://huggingface.co/spaces/trl-lib/recommend-vllm-memory) to help estimate the recommended GPU memory utilization based on your model configuration and experiment settings. Simply use it as follows to get `vllm_gpu_memory_utilization` recommendation:
+
+<iframe
+	src="https://trl-lib-recommend-vllm-memory.hf.space"
+	frameborder="0"
+	width="850"
+	height="450"
+></iframe>
+
+If the recommended value does not work in your environment, we suggest adding a small buffer (e.g., +0.05 or +0.1) to the recommended value to ensure stability.
+
+</Tip>
+
+<Tip>
+
+By default, GRPO uses `MASTER_ADDR=localhost` and `MASTER_PORT=12345` for vLLM, but you can override these values by setting the environment variables accordingly.
+
+</Tip>
+
+For more information, see [Speeding up training with vLLM](speeding_up_training#vllm-for-fast-generation-in-online-methods).
+
+### GRPO at scale: train a 70B+ Model on multiple nodes
+
+When training large models like **Qwen2.5-72B**, you need several key optimizations to make the training efficient and scalable across multiple GPUs and nodes. These include:
+
+- **DeepSpeed ZeRO Stage 3**: ZeRO leverages data parallelism to distribute model states (weights, gradients, optimizer states) across multiple GPUs and CPUs, reducing memory and compute requirements on each device. Since large models cannot fit on a single GPU, using ZeRO Stage 3 is required for training such model. For more details, see [DeepSpeed Integration](deepspeed_integration).
+- **Accelerate**: Accelerate is a library that simplifies distributed training across multiple GPUs and nodes. It provides a simple API to launch distributed training and handles the complexities of distributed training, such as data parallelism, gradient accumulation, and distributed data loading. For more details, see [Distributing Training](distributing_training).
+- **vLLM**: See the previous section on how to use vLLM to speed up generation.
+
+Below is an example SLURM script to train a 70B model with GRPO on multiple nodes. This script trains a model on 4 nodes and uses the 5th node for vLLM-powered generation.
+
+```sh
+#!/bin/bash
+#SBATCH --nodes=5
+#SBATCH --gres=gpu:8
+
+# Get the list of allocated nodes
+NODELIST=($(scontrol show hostnames $SLURM_JOB_NODELIST))
+
+# Assign the first 4 nodes for training and the 5th node for vLLM
+TRAIN_NODES="${NODELIST[@]:0:4}"  # Nodes 0, 1, 2, 3 for training
+VLLM_NODE="${NODELIST[4]}"  # Node 4 for vLLM
+
+# Run training on the first 4 nodes (Group 1)
+srun --nodes=4 --ntasks=4 --nodelist="${NODELIST[@]:0:4}" accelerate launch \
+     --config_file examples/accelerate_configs/deepspeed_zero3.yaml \
+     --num_processes 32 \
+     --num_machines 4 \
+     --main_process_ip ${NODELIST[0]} \
+     --machine_rank $SLURM_PROCID \
+     --rdzv_backend c10d \
+     train_grpo.py \
+     --server_ip $VLLM_NODE &
+
+# Run vLLM server on the 5th node (Group 2)
+srun --nodes=1 --ntasks=1 --nodelist="${NODELIST[4]}" trl vllm-serve --model Qwen/Qwen2.5-72B --tensor_parallel_size 8 &
+
+wait
+```
+
+```python
+import argparse
+
+from datasets import load_dataset
+from trl import GRPOTrainer, GRPOConfig
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--vllm_server_host", type=str, default="", help="The server IP")
+    args = parser.parse_args()
+
+    # Example dataset from TLDR
+    dataset = load_dataset("trl-lib/tldr", split="train")
+
+    # Dummy reward function: count the number of unique characters in the completions
+    def reward_num_unique_chars(completions, **kwargs):
+        return [len(set(c)) for c in completions]
+
+    training_args = GRPOConfig(
+        output_dir="Qwen2.5-72B-GRPO",
+        per_device_train_batch_size=4,
+        bf16=True,
+        gradient_checkpointing=True,
+        use_vllm=True,
+        vllm_server_host=args.vllm_server_host.replace("ip-", "").replace("-", "."),  # from ip-X-X-X-X to X.X.X.X
+    )
+
+    trainer = GRPOTrainer(model="Qwen/Qwen2.5-72B", args=training_args, reward_funcs=reward_num_unique_chars, train_dataset=dataset)
+    trainer.train()
+
+if __name__=="__main__":
+    main()
+```
+
+### Using a custom reward function
+
+The [`GRPOTrainer`] supports using custom reward functions instead of dense reward models. To ensure compatibility, your reward function must satisfy the following requirements:
+
+1. **Input arguments**:
+   - The function must accept the following as keyword arguments:
+     - `prompts` (contains the prompts),
+     - `completions` (contains the generated completions),
+     - `completions_ids` (contains the tokenized completions),
+     - `trainer_state` ([`~transformers.TrainerState`]): The current state of the trainer. This can be used to implement dynamic reward functions, such as curriculum learning, where the reward is adjusted based on the training progress.
+     - All columns names (but `prompt`) that the dataset may have. For example, if the dataset contains a column named `ground_truth`, the function will be called with `ground_truth` as a keyword argument.
+
+     The easiest way to comply with this requirement is to use `**kwargs` in the function signature.
+   - Depending on the dataset format, the input will vary:
+     - For [standard format](dataset_formats#standard), `prompts` and `completions` will be lists of strings.
+     - For [conversational format](dataset_formats#conversational), `prompts` and `completions` will be lists of message dictionaries.
+
+2. **Return value**: The function must return a list of floats. Each float represents the reward corresponding to a single completion.
+
+#### Example 1: Reward longer completions
+
+Below is an example of a reward function for a standard format that rewards longer completions:
+
+```python
+def reward_func(completions_ids, **kwargs):
+    """Reward function that assigns higher scores to longer completions (in terms of token count)."""
+    return [float(len(ids)) for ids in completions_ids]
+```
+
+You can test it as follows:
+
+```python
+>>> prompts = ["The sky is", "The sun is"]  # not used in the reward function, but the trainer will pass it
+>>> completions = [" blue.", " in the sky."]  # not used in the reward function, but the trainer will pass it
+>>> completions_ids = [[6303, 13], [304, 279, 12884, 13]]
+>>> reward_func(prompts=prompts, completions=completions, completions_ids=completions_ids)
+[2.0, 4.0]
+```
+
+#### Example 1.1: Reward longer completions (based in the number of characters)
+
+Same as the previous example, but this time the reward function is based on the number of characters instead of tokens.
+
+```python
+def reward_func(completions, **kwargs):
+    """Reward function that assigns higher scores to longer completions (in terms of character count)."""
+    return [float(len(completion)) for completion in completions]
+```
+
+You can test it as follows:
+
+```python
+>>> prompts = ["The sky is", "The sun is"]
+>>> completions = [" blue.", " in the sky."]
+>>> completions_ids = [[6303, 13], [304, 279, 12884, 13]]  # not used in the reward function, but the trainer will pass it
+>>> reward_func(prompts=prompts, completions=completions, completions_ids=completions_ids)
+[6.0, 12.0]
+```
+
+#### Example 2: Reward completions with specific format
+
+Below is an example of a reward function that checks if the completion has a specific format. This example is inspired by the _format reward_ function used in the paper [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://huggingface.co/papers/2501.12948).
+It is designed for conversational format, where prompts and completions consist of structured messages.
+
+```python
+import re
+
+def format_reward_func(completions, **kwargs):
+    """Reward function that checks if the completion has a specific format."""
+    pattern = r"^<think>.*?</think><answer>.*?</answer>$"
+    completion_contents = [completion[0]["content"] for completion in completions]
+    matches = [re.match(pattern, content) for content in completion_contents]
+    return [1.0 if match else 0.0 for match in matches]
+```
+
+You can test this function as follows:
+
+```python
+>>> prompts = [
+...     [{"role": "assistant", "content": "What is the result of (1 + 2) * 4?"}],
+...     [{"role": "assistant", "content": "What is the result of (3 + 1) * 2?"}],
+... ]
+>>> completions = [
+...     [{"role": "assistant", "content": "<think>The sum of 1 and 2 is 3, which we multiply by 4 to get 12.</think><answer>(1 + 2) * 4 = 12</answer>"}],
+...     [{"role": "assistant", "content": "The sum of 3 and 1 is 4, which we multiply by 2 to get 8. So (3 + 1) * 2 = 8."}],
+... ]
+>>> format_reward_func(prompts=prompts, completions=completions)
+[1.0, 0.0]
+```
+
+#### Example 3: Reward completions based on a reference
+
+Below is an example of a reward function that checks if the completion is correct. This example is inspired by the _accuracy reward_ function used in the paper [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://huggingface.co/papers/2501.12948).
+This example is designed for [standard format](dataset_formats#standard), where the dataset contains a column named `ground_truth`.
+
+```python
+import re
+
+def reward_func(completions, ground_truth, **kwargs):
+    # Regular expression to capture content inside \boxed{}
+    matches = [re.search(r"\\boxed\{(.*?)\}", completion) for completion in completions]
+    contents = [match.group(1) if match else "" for match in matches]
+    # Reward 1 if the content is the same as the ground truth, 0 otherwise
+    return [1.0 if c == gt else 0.0 for c, gt in zip(contents, ground_truth)]
+```
+
+You can test this function as follows:
+
+```python
+>>> prompts = ["Problem: Solve the equation $2x + 3 = 7$. Solution:", "Problem: Solve the equation $3x - 5 = 10$."]
+>>> completions = [r" The solution is \boxed{2}.", r" The solution is \boxed{6}."]
+>>> ground_truth = ["2", "5"]
+>>> reward_func(prompts=prompts, completions=completions, ground_truth=ground_truth)
+[1.0, 0.0]
+```
+#### Example 4: Multi-task reward functions
+
+Below is an example of using multiple reward functions in the [`GRPOTrainer`]. In this example, we define two task-specific reward functions: `math_reward_func` and `coding_reward_func`. The `math_reward_func` rewards math problems based on their correctness, while the `coding_reward_func` rewards coding problems based on whether the solution works.
+
+```python
+from datasets import Dataset
+from trl import GRPOTrainer
+
+# Define a dataset that contains both math and coding problems
+dataset = Dataset.from_list(
+    [
+        {"prompt": "What is 2+2?", "task": "math"},
+        {"prompt": "Write a function that returns the sum of two numbers.", "task": "code"},
+        {"prompt": "What is 3*4?", "task": "math"},
+        {"prompt": "Write a function that returns the product of two numbers.", "task": "code"},
+    ]
+)
+
+# Math-specific reward function
+def math_reward_func(prompts, completions, task, **kwargs):
+    rewards = []
+    for prompt, completion, t in zip(prompts, completions, task):
+        if t == "math":
+            # Calculate math-specific reward
+            correct = check_math_solution(prompt, completion)
+            reward = 1.0 if correct else -1.0
+            rewards.append(reward)
+        else:
+            # Return None for non-math tasks
+            rewards.append(None)
+    return rewards
+
+# Coding-specific reward function
+def coding_reward_func(prompts, completions, task, **kwargs):
+    rewards = []
+    for prompt, completion, t in zip(prompts, completions, task):
+        if t == "coding":
+            # Calculate coding-specific reward
+            works = test_code_solution(prompt, completion)
+            reward = 1.0 if works else -1.0
+            rewards.append(reward)
+        else:
+            # Return None for non-coding tasks
+            rewards.append(None)
+    return rewards
+
+# Use both task-specific reward functions
+trainer = GRPOTrainer(
+    model="Qwen/Qwen2-0.5B-Instruct",
+    reward_funcs=[math_reward_func, coding_reward_func],
+    train_dataset=dataset,
+)
+
+trainer.train()
+```
+
+In this example, the `math_reward_func` and `coding_reward_func` are designed to work with a mixed dataset that contains both math and coding problems. The `task` column in the dataset is used to determine which reward function to apply to each problem. If there is no relevant reward function for a sample in the dataset, the reward function will return `None` and the [`GRPOTrainer`] will continue with the valid functions and tasks. This allows the [`GRPOTrainer`] to handle multiple reward functions with different applicability.
+
+Note that the [`GRPOTrainer`] will ignore the `None` rewards returned by the reward functions and only consider the rewards returned by the relevant functions. This ensures that the model is trained on the relevant tasks and ignores the tasks for which there is no relevant reward function.
+
+
+
+#### Passing the reward function to the trainer
+
+To use your custom reward function, pass it to the [`GRPOTrainer`] as follows:
+
+```python
+from trl import GRPOTrainer
+
+trainer = GRPOTrainer(
+    reward_funcs=reward_func,
+    ...,
+)
+```
+
+If you have multiple reward functions, you can pass them as a list:
+
+```python
+from trl import GRPOTrainer
+
+trainer = GRPOTrainer(
+    reward_funcs=[reward_func1, reward_func2],
+    ...,
+)
+```
+
+and the reward will be computed as the sum of the rewards from each function, or the weighted sum if `reward_weights` is provided in the config.
+
+Note that [`GRPOTrainer`] supports multiple reward functions of different types. See the parameters documentation for more details.
+
+## Vision-Language Model (VLM) Training
+
+GRPO supports training Vision-Language Models (VLMs) on multimodal datasets containing both text and images.
+
+### Supported Models
+
+Tested with:
+
+- **Gemma3** — e.g., `google/gemma-3-4b-it`
+- **LLaVA-NeXT** — e.g., `llava-hf/llava-v1.6-mistral-7b-hf`
+- **Qwen2-VL** — e.g., `Qwen/Qwen2-VL-2B-Instruct`
+- **Qwen2.5-VL** — e.g., `Qwen/Qwen2.5-VL-3B-Instruct`
+- **SmolVLM2** — e.g., `HuggingFaceTB/SmolVLM2-2.2B-Instruct`
+  
+<Tip>
+Compatibility with all VLMs is not guaranteed. If you believe a model should be supported, feel free to open an issue on GitHub — or better yet, submit a pull request with the required changes.
+</Tip>
+
+### Quick Start
+
+Use [grpo\_vlm.py](https://github.com/huggingface/trl/blob/main/examples/scripts/grpo_vlm.py) to fine-tune a VLM. Example command for training on [`lmms-lab/multimodal-open-r1-8k-verified`](https://huggingface.co/datasets/lmms-lab/multimodal-open-r1-8k-verified):
+
+```bash
+accelerate launch \
+  --config_file=examples/accelerate_configs/deepspeed_zero3.yaml \
+  examples/scripts/grpo_vlm.py \
+  --model_name_or_path Qwen/Qwen2.5-VL-3B-Instruct \
+  --output_dir grpo-Qwen2.5-VL-3B-Instruct \
+  --learning_rate 1e-5 \
+  --gradient_checkpointing \
+  --torch_dtype bfloat16 \
+  --max_prompt_length 2048 \
+  --max_completion_length 1024 \
+  --use_vllm \
+  --vllm_mode colocate \
+  --use_peft \
+  --lora_target_modules "q_proj", "v_proj" \
+  --log_completions
+```
+
+### Configuration Tips
+
+<Tip warning={true}>
+VLM training may fail if image tokens are truncated. We highly recommend to disable truncation by setting `max_prompt_length` to `None`.
+</Tip>
+
+- Use LoRA on vision-language projection layers
+- Enable 4-bit quantization to reduce memory usage
+- VLMs are memory-intensive — start with smaller batch sizes
+- Most models are compatible with vLLM (`server` and `colocate` modes)
+
+### Dataset Format
+
+Each training sample should include:
+
+- `prompt`: Text formatted via the processor's chat template
+- `image`: A single image (PIL or NumPy array)
+
+The trainer automatically handles image-to-tensor conversion via the model’s image processor.
+
+## GRPOTrainer
+
+[[autodoc]] GRPOTrainer
+
+## GRPOConfig
+
+[[autodoc]] GRPOConfig

docs/source/how_to_train.md

@@ -18,7 +18,7 @@ When training RL models, optimizing solely for reward may lead to unexpected beh
 However, the RL model being optimized against the reward model may learn patterns that yield high reward but do not represent good language. This can result in extreme cases where the model generates texts with excessive exclamation marks or emojis to maximize the reward. In some worst-case scenarios, the model may generate patterns completely unrelated to natural language yet receive high rewards, similar to adversarial attacks.
 
 <div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/kl-example.png">
+<img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/kl-example.png">
 <p style="text-align: center;"> <b>Figure:</b> Samples without a KL penalty from <a href="https://huggingface.co/papers/1909.08593">https://huggingface.co/papers/1909.08593</a>. </p>
 </div>
 

docs/source/index.md

@@ -0,0 +1,82 @@
+<div style="text-align: center">
+<img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/trl_banner_dark.png">
+</div>
+
+# TRL - Transformer Reinforcement Learning
+
+TRL is a full stack library where we provide a set of tools to train transformer language models with methods like Supervised Fine-Tuning (SFT), Group Relative Policy Optimization (GRPO), Direct Preference Optimization (DPO), Reward Modeling, and more.
+The library is integrated with 🤗 [transformers](https://github.com/huggingface/transformers).
+
+You can also explore TRL-related models, datasets, and demos in the [TRL Hugging Face organization](https://huggingface.co/trl-lib).
+
+## Learn
+
+Learn post-training with TRL and other libraries in 🤗 [smol course](https://github.com/huggingface/smol-course).
+
+## Contents
+
+The documentation is organized into the following sections:
+
+- **Getting Started**: installation and quickstart guide.
+- **Conceptual Guides**: dataset formats, training FAQ, and understanding logs.
+- **How-to Guides**: reducing memory usage, speeding up training, distributing training, etc.
+- **Integrations**: DeepSpeed, Liger Kernel, PEFT, etc.
+- **Examples**: example overview, community tutorials, etc.
+- **API**: trainers, utils, etc.
+
+## Blog posts
+
+<div class="mt-10">
+  <div class="w-full flex flex-col space-y-4 md:space-y-0 md:grid md:grid-cols-2 md:gap-y-4 md:gap-x-5">
+    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/vllm-colocate">
+      <img src="https://raw.githubusercontent.com/huggingface/blog/main/assets/vllm-colocate/thumbnail.png" alt="thumbnail" class="mt-0">
+      <p class="text-gray-500 text-sm">Published on June 3, 2025</p>
+      <p class="text-gray-700">NO GPU left behind: Unlocking Efficiency with Co-located vLLM in TRL</p>
+    </a>
+    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/liger-grpo">
+      <img src="https://raw.githubusercontent.com/huggingface/blog/main/assets/liger-grpo/thumbnail.png" alt="thumbnail" class="mt-0">
+      <p class="text-gray-500 text-sm">Published on May 25, 2025</p>
+      <p class="text-gray-700">🐯 Liger GRPO meets TRL</p>
+    </a>
+    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/open-r1">
+      <img src="https://raw.githubusercontent.com/huggingface/blog/main/assets/open-r1/thumbnails.png" alt="thumbnail" class="mt-0">
+      <p class="text-gray-500 text-sm">Published on January 28, 2025</p>
+      <p class="text-gray-700">Open-R1: a fully open reproduction of DeepSeek-R1</p>
+    </a>
+    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/dpo_vlm">
+      <img src="https://raw.githubusercontent.com/huggingface/blog/main/assets/dpo_vlm/thumbnail.png" alt="thumbnail" class="mt-0">
+      <p class="text-gray-500 text-sm">Published on July 10, 2024</p>
+      <p class="text-gray-700">Preference Optimization for Vision Language Models with TRL</p>
+    </a>
+    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/putting_rl_back_in_rlhf_with_rloo">
+      <img src="https://raw.githubusercontent.com/huggingface/blog/main/assets/putting_rl_back_in_rlhf_with_rloo/thumbnail.png" alt="thumbnail" class="mt-0">
+      <p class="text-gray-500 text-sm">Published on June 12, 2024</p>
+      <p class="text-gray-700">Putting RL back in RLHF</p>
+    </a>
+    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/trl-ddpo">
+      <img src="https://raw.githubusercontent.com/huggingface/blog/main/assets/166_trl_ddpo/thumbnail.png" alt="thumbnail" class="mt-0">
+      <p class="text-gray-500 text-sm">Published on September 29, 2023</p>
+      <p class="text-gray-700">Finetune Stable Diffusion Models with DDPO via TRL</p>
+    </a>
+    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/dpo-trl">
+      <img src="https://raw.githubusercontent.com/huggingface/blog/main/assets/157_dpo_trl/dpo_thumbnail.png" alt="thumbnail" class="mt-0">
+      <p class="text-gray-500 text-sm">Published on August 8, 2023</p>
+      <p class="text-gray-700">Fine-tune Llama 2 with DPO</p>
+    </a>
+    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/stackllama">
+      <img src="https://raw.githubusercontent.com/huggingface/blog/main/assets/138_stackllama/thumbnail.png" alt="thumbnail" class="mt-0">
+      <p class="text-gray-500 text-sm">Published on April 5, 2023</p>
+      <p class="text-gray-700">StackLLaMA: A hands-on guide to train LLaMA with RLHF</p>
+   </a>
+    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/trl-peft">
+      <img src="https://raw.githubusercontent.com/huggingface/blog/main/assets/133_trl_peft/thumbnail.png" alt="thumbnail" class="mt-0">
+      <p class="text-gray-500 text-sm">Published on March 9, 2023</p>
+      <p class="text-gray-700">Fine-tuning 20B LLMs with RLHF on a 24GB consumer GPU</p>
+    </a>
+    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/rlhf">
+      <img src="https://raw.githubusercontent.com/huggingface/blog/main/assets/120_rlhf/thumbnail.png" alt="thumbnail" class="mt-0">
+      <p class="text-gray-500 text-sm">Published on December 9, 2022</p>
+      <p class="text-gray-700">Illustrating Reinforcement Learning from Human Feedback</p>
+    </a>
+  </div>
+</div>

docs/source/index.mdx

@@ -1,65 +0,0 @@
-<div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/trl_banner_dark.png">
-</div>
-
-# TRL - Transformer Reinforcement Learning
-
-TRL is a full stack library where we provide a set of tools to train transformer language models with Reinforcement Learning, from the Supervised Fine-tuning step (SFT), Reward Modeling step (RM) to the Proximal Policy Optimization (PPO) step. 
-The library is integrated with 🤗 [transformers](https://github.com/huggingface/transformers).
-
-<div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/TRL-readme.png">
-</div>
-
-Check the appropriate sections of the documentation depending on your needs:
-
-## API documentation
-
-- [Model Classes](models): *A brief overview of what each public model class does.*
-- [`SFTTrainer`](sft_trainer): *Supervise Fine-tune your model easily with `SFTTrainer`*
-- [`RewardTrainer`](reward_trainer): *Train easily your reward model using `RewardTrainer`.*
-- [`PPOTrainer`](ppo_trainer): *Further fine-tune the supervised fine-tuned model using PPO algorithm*
-- [Best-of-N Sampling](best-of-n): *Use best of n sampling as an alternative way to sample predictions from your active model*
-- [`DPOTrainer`](dpo_trainer): *Direct Preference Optimization training using `DPOTrainer`.*
-- [`TextEnvironment`](text_environments): *Text environment to train your model using tools with RL.*
-
-## Examples
-
-- [Sentiment Tuning](sentiment_tuning): *Fine tune your model to generate positive movie contents*
-- [Training with PEFT](lora_tuning_peft): *Memory efficient RLHF training using adapters with PEFT*
-- [Detoxifying LLMs](detoxifying_a_lm): *Detoxify your language model through RLHF*
-- [StackLlama](using_llama_models): *End-to-end RLHF training of a Llama model on Stack exchange dataset*
-- [Learning with Tools](learning_tools): *Walkthrough of using `TextEnvironments`*
-- [Multi-Adapter Training](multi_adapter_rl): *Use a single base model and multiple adapters for memory efficient end-to-end training*
-
-
-## Blog posts
-
-<div class="mt-10">
-  <div class="w-full flex flex-col space-y-4 md:space-y-0 md:grid md:grid-cols-2 md:gap-y-4 md:gap-x-5">
-    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/dpo_vlm">
-      <img src="https://raw.githubusercontent.com/huggingface/blog/main/assets/dpo_vlm/thumbnail.png" alt="thumbnail">
-      <p class="text-gray-700">Preference Optimization for Vision Language Models with TRL</p>
-    </a>
-    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/rlhf">
-      <img src="https://raw.githubusercontent.com/huggingface/blog/main/assets/120_rlhf/thumbnail.png" alt="thumbnail">
-      <p class="text-gray-700">Illustrating Reinforcement Learning from Human Feedback</p>
-    </a>
-    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/trl-peft">
-      <img src="https://github.com/huggingface/blog/blob/main/assets/133_trl_peft/thumbnail.png?raw=true" alt="thumbnail">
-      <p class="text-gray-700">Fine-tuning 20B LLMs with RLHF on a 24GB consumer GPU</p>
-    </a>
-    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/stackllama">
-      <img src="https://github.com/huggingface/blog/blob/main/assets/138_stackllama/thumbnail.png?raw=true" alt="thumbnail">
-      <p class="text-gray-700">StackLLaMA: A hands-on guide to train LLaMA with RLHF</p>
-   </a>
-    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/dpo-trl">
-      <img src="https://github.com/huggingface/blog/blob/main/assets/157_dpo_trl/dpo_thumbnail.png?raw=true" alt="thumbnail">
-      <p class="text-gray-700">Fine-tune Llama 2 with DPO</p>
-    </a>
-    <a class="!no-underline border dark:border-gray-700 p-5 rounded-lg shadow hover:shadow-lg" href="https://huggingface.co/blog/trl-ddpo">
-      <img src="https://github.com/huggingface/blog/blob/main/assets/166_trl_ddpo/thumbnail.png?raw=true" alt="thumbnail">
-      <p class="text-gray-700">Finetune Stable Diffusion Models with DDPO via TRL</p>
-    </a>
-  </div>
-</div>

docs/source/installation.md

@@ -0,0 +1,39 @@
+# Installation
+You can install TRL either from PyPI or from source:
+
+## PyPI
+Install the library with pip or [uv](https://docs.astral.sh/uv/):
+
+<hfoptions id="install">
+<hfoption id="uv">
+
+uv is a fast Rust-based Python package and project manager. Refer to [Installation](https://docs.astral.sh/uv/getting-started/installation/) for installation instructions), .
+
+```bash
+uv pip install trl
+```
+
+</hfoption>
+<hfoption id="pip">
+
+```bash
+pip install trl
+```
+
+</hfoption>
+</hfoptions>
+
+## Source
+You can also install the latest version from source. First clone the repo and then run the installation with `pip`:
+
+```bash
+git clone https://github.com/huggingface/trl.git
+cd trl/
+pip install -e .
+```
+
+If you want the development install you can replace the pip install with the following:
+
+```bash
+pip install -e ".[dev]"
+```

docs/source/installation.mdx

@@ -1,24 +0,0 @@
-# Installation
-You can install TRL either from pypi or from source:
-
-## pypi
-Install the library with pip:
-
-```bash
-pip install trl
-```
-
-### Source
-You can also install the latest version from source. First clone the repo and then run the installation with `pip`:
-
-```bash
-git clone https://github.com/huggingface/trl.git
-cd trl/
-pip install -e .
-```
-
-If you want the development install you can replace the pip install with the following:
-
-```bash
-pip install -e ".[dev]"
-```

docs/source/iterative_sft_trainer.md

@@ -0,0 +1,138 @@
+# Iterative Trainer
+
+[![](https://img.shields.io/badge/All_models-Iterative_SFT-blue)](https://huggingface.co/models?other=iterative-sft,trl)
+
+Iterative fine-tuning is a training method that enables to perform custom actions (generation and filtering for example) between optimization steps. In TRL we provide an easy-to-use API to fine-tune your models in an iterative way in just a few lines of code.
+
+## Quickstart
+
+To get started quickly, you can either pass a model identifier or a pre-instantiated model to the trainer:
+
+```python
+from trl import IterativeSFTConfig, IterativeSFTTrainer
+
+# Using a model identifier
+trainer = IterativeSFTTrainer(
+    "facebook/opt-350m",
+    args=IterativeSFTConfig(
+        max_length=512,
+        output_dir="./output",
+    ),
+)
+
+# Or using a pre-instantiated model
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model = AutoModelForCausalLM.from_pretrained("facebook/opt-350m")
+tokenizer = AutoTokenizer.from_pretrained("facebook/opt-350m")
+
+trainer = IterativeSFTTrainer(
+    model,
+    args=IterativeSFTConfig(
+        max_length=512,
+        output_dir="./output",
+    ),
+    processing_class=tokenizer,
+)
+```
+
+## Usage
+
+The [`IterativeSFTTrainer`] supports two ways of providing input data to the `step` function:
+
+### Using a list of tensors as input:
+
+```python
+inputs = {
+    "input_ids": input_ids,
+    "attention_mask": attention_mask,
+}
+
+trainer.step(**inputs)
+```
+
+### Using a list of strings as input:
+
+```python
+inputs = {
+    "texts": texts,
+    "texts_labels": texts_labels,  # Optional, defaults to texts
+}
+
+trainer.step(**inputs)
+```
+
+For causal language models, labels will automatically be created from `input_ids` or from `texts`. When using sequence to sequence models you will have to provide your own labels or `text_labels`.
+
+## Configuration
+
+The [`IterativeSFTConfig`] class provides several parameters to customize the training:
+
+```python
+from trl import IterativeSFTConfig
+
+config = IterativeSFTConfig(
+    # Model initialization parameters
+    model_init_kwargs={"torch_dtype": "bfloat16"},
+
+    # Data preprocessing parameters
+    max_length=512,
+    truncation_mode="keep_end",
+
+    # Training parameters
+    output_dir="./output",
+    learning_rate=2e-5,
+    per_device_train_batch_size=4,
+    gradient_accumulation_steps=4,
+    max_steps=1000,
+    save_steps=100,
+    optim="adamw_torch",
+    report_to="wandb",
+)
+```
+
+### Model Initialization
+
+You can control how the model is initialized by passing keyword arguments to `model_init_kwargs`:
+
+```python
+config = IterativeSFTConfig(
+    model_init_kwargs={
+        "torch_dtype": "bfloat16",
+        "device_map": "auto",
+        "trust_remote_code": True,
+    }
+)
+```
+
+### Data Preprocessing
+
+The trainer supports two truncation modes:
+
+- `keep_end`: Truncates from the start of the sequence
+- `keep_start`: Truncates from the end of the sequence
+
+```python
+config = IterativeSFTConfig(
+    max_length=512,
+    truncation_mode="keep_end",  # or "keep_start"
+)
+```
+
+### Training Optimization
+
+You can optimize CUDA cache usage for more memory-efficient training:
+
+```python
+config = IterativeSFTConfig(
+    optimize_device_cache=True,
+)
+```
+
+## IterativeSFTTrainer
+
+[[autodoc]] IterativeSFTTrainer
+
+## IterativeSFTConfig
+
+[[autodoc]] IterativeSFTConfig

docs/source/iterative_sft_trainer.mdx

@@ -1,54 +0,0 @@
-# Iterative Trainer
-
-Iterative fine-tuning is a training method that enables to perform custom actions (generation and filtering for example) between optimization steps. In TRL we provide an easy-to-use API to fine-tune your models in an iterative way in just a few lines of code.
-
-## Usage
-
-To get started quickly, instantiate an instance a model, and a tokenizer.
-
-```python
-
-model = AutoModelForCausalLM.from_pretrained(model_name)
-tokenizer = AutoTokenizer.from_pretrained(model_name)
-if tokenizer.pad_token is None:
-    tokenizer.pad_token = tokenizer.eos_token
-
-trainer = IterativeSFTTrainer(
-    model,
-    tokenizer
-)
-
-```
-
-You have the choice to either provide a list of strings or a list of tensors to the step function. 
-
-#### Using a list of tensors as input:
-
-```python
-
-inputs = {
-    "input_ids": input_ids,
-    "attention_mask": attention_mask
-}
-
-trainer.step(**inputs)
-
-```
-
-#### Using a list of strings as input:
-
-```python
-
-inputs = {
-    "texts": texts
-}
-
-trainer.step(**inputs)
-
-```
-
-For causal language models, labels will automatically be created from input_ids or from texts. When using sequence to sequence models you will have to provide your own labels or text_labels.
-
-## IterativeTrainer
-
-[[autodoc]] IterativeSFTTrainer

docs/source/judges.md

@@ -1,11 +1,17 @@
 # Judges
 
+<Tip warning={true}>
+
+TRL Judges is an experimental API which is subject to change at any time.
+
+</Tip>
+
 TRL provides judges to easily compare two completions.
 
 Make sure to have installed the required dependencies by running:
 
 ```bash
-pip install trl[llm_judge]
+pip install trl[judges]
 ```
 
 ## Using the provided judges
@@ -46,34 +52,38 @@ judge.judge(
 )  # Outputs: [0, 1]
 ```
 
-## BaseJudge
+## Provided judges
 
-[[autodoc]] BaseJudge
-
-## BaseRankJudge
-
-[[autodoc]] BaseRankJudge
-
-## BasePairwiseJudge
-
-[[autodoc]] BasePairwiseJudge
-
-## RandomRankJudge
-
-[[autodoc]] RandomRankJudge
-
-## RandomPairwiseJudge
-
-[[autodoc]] RandomPairwiseJudge
-
-## PairRMJudge
+### PairRMJudge
 
 [[autodoc]] PairRMJudge
 
-## HfPairwiseJudge
+### HfPairwiseJudge
 
 [[autodoc]] HfPairwiseJudge
 
-## OpenAIPairwiseJudge
+### OpenAIPairwiseJudge
 
 [[autodoc]] OpenAIPairwiseJudge
+
+### AllTrueJudge
+
+[[autodoc]] AllTrueJudge
+
+## Base classes
+
+### BaseJudge
+
+[[autodoc]] BaseJudge
+
+### BaseBinaryJudge
+
+[[autodoc]] BaseBinaryJudge
+
+### BaseRankJudge
+
+[[autodoc]] BaseRankJudge
+
+### BasePairwiseJudge
+
+[[autodoc]] BasePairwiseJudge

docs/source/kto_trainer.md

@@ -0,0 +1,138 @@
+# KTO Trainer
+
+[![](https://img.shields.io/badge/All_models-KTO-blue)](https://huggingface.co/models?other=kto,trl)
+
+## Overview
+
+Kahneman-Tversky Optimization (KTO) was introduced in [KTO: Model Alignment as Prospect Theoretic Optimization](https://huggingface.co/papers/2402.01306) by [Kawin Ethayarajh](https://huggingface.co/kawine), [Winnie Xu](https://huggingface.co/xwinxu), [Niklas Muennighoff](https://huggingface.co/Muennighoff), Dan Jurafsky, [Douwe Kiela](https://huggingface.co/douwekiela).
+
+
+The abstract from the paper is the following:
+
+> Kahneman & Tversky's prospect theory tells us that humans perceive random variables in a biased but well-defined manner; for example, humans are famously loss-averse. We show that objectives for aligning LLMs with human feedback implicitly incorporate many of these biases -- the success of these objectives (e.g., DPO) over cross-entropy minimization can partly be ascribed to them being human-aware loss functions (HALOs). However, the utility functions these methods attribute to humans still differ from those in the prospect theory literature. Using a Kahneman-Tversky model of human utility, we propose a HALO that directly maximizes the utility of generations instead of maximizing the log-likelihood of preferences, as current methods do. We call this approach Kahneman-Tversky Optimization (KTO), and it matches or exceeds the performance of preference-based methods at scales from 1B to 30B. Crucially, KTO does not need preferences -- only a binary signal of whether an output is desirable or undesirable for a given input. This makes it far easier to use in the real world, where preference data is scarce and expensive.
+
+The official code can be found in [ContextualAI/HALOs](https://github.com/ContextualAI/HALOs).
+
+This post-training method was contributed by [Kashif Rasul](https://huggingface.co/kashif), [Younes Belkada](https://huggingface.co/ybelkada), [Lewis Tunstall](https://huggingface.co/lewtun) and Pablo Vicente.
+
+## Quick start
+
+This example demonstrates how to train a model using the KTO method. We use the [Qwen 0.5B model](https://huggingface.co/Qwen/Qwen2-0.5B-Instruct) as the base model. We use the preference data from the [KTO Mix 14k](https://huggingface.co/datasets/trl-lib/kto-mix-14k). You can view the data in the dataset here:
+
+<iframe
+  src="https://huggingface.co/datasets/trl-lib/kto-mix-14k/embed/viewer/default/train?row=0"
+  frameborder="0"
+  width="100%"
+  height="560px"
+></iframe>
+
+Below is the script to train the model:
+
+```python
+# train_kto.py
+from datasets import load_dataset
+from trl import KTOConfig, KTOTrainer
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
+train_dataset = load_dataset("trl-lib/kto-mix-14k", split="train")
+
+training_args = KTOConfig(output_dir="Qwen2-0.5B-KTO")
+trainer = KTOTrainer(model=model, args=training_args, processing_class=tokenizer, train_dataset=train_dataset)
+trainer.train()
+```
+
+Execute the script using the following command:
+
+```bash
+accelerate launch train_kto.py
+```
+
+Distributed across 8 x H100 GPUs, the training takes approximately 30 minutes. You can verify the training progress by checking the reward graph. An increasing trend in the reward margin indicates that the model is improving and generating better responses over time.
+
+![](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/kto-qwen2-reward-margin.png)
+
+To see how the [trained model](https://huggingface.co/trl-lib/Qwen2-0.5B-KTO) performs, you can use the [Transformers Chat CLI](https://huggingface.co/docs/transformers/quicktour#chat-with-text-generation-models).
+
+<pre><code>$ transformers chat trl-lib/Qwen2-0.5B-KTO
+<strong><span style="color: red;">&lt;quentin_gallouedec&gt;:</span></strong>
+What is the best programming language?
+
+<strong><span style="color: blue;">&lt;trl-lib/Qwen2-0.5B-KTO&gt;:</span></strong>
+The best programming language can vary depending on individual preferences, industry-specific requirements, technical skills, and familiarity with the specific use case or task. Here are some widely-used programming languages that have been noted as popular and widely used:                                                                                  
+
+Here are some other factors to consider when choosing a programming language for a project:
+
+ <strong><span style="color: green;">1</span> JavaScript</strong>: JavaScript is at the heart of the web and can be used for building web applications, APIs, and interactive front-end applications like frameworks like React and Angular. It's similar to C, C++, and F# in syntax structure and is accessible and easy to learn, making it a popular choice for beginners and professionals alike.                                                                   
+ <strong><span style="color: green;">2</span> Java</strong>: Known for its object-oriented programming (OOP) and support for Java 8 and .NET, Java is used for developing enterprise-level software applications, high-performance games, as well as mobile apps, game development, and desktop applications.                                                                                                                                                            
+ <strong><span style="color: green;">3</span> C++</strong>: Known for its flexibility and scalability, C++ offers comprehensive object-oriented programming and is a popular choice for high-performance computing and other technical fields. It's a powerful platform for building real-world applications and games at scale.                                                                                                                                         
+ <strong><span style="color: green;">4</span> Python</strong>: Developed by Guido van Rossum in 1991, Python is a high-level, interpreted, and dynamically typed language known for its simplicity, readability, and versatility.   
+</code></pre>
+
+## Expected dataset format
+
+KTO requires an [unpaired preference dataset](dataset_formats#unpaired-preference). Alternatively, you can provide a *paired* preference dataset (also known simply as a *preference dataset*). In this case, the trainer will automatically convert it to an unpaired format by separating the chosen and rejected responses, assigning `label = True` to the chosen completions and `label = False` to the rejected ones.
+
+The [`KTOTrainer`] supports both [conversational](dataset_formats#conversational) and [standard](dataset_formats#standard) dataset format. When provided with a conversational dataset, the trainer will automatically apply the chat template to the dataset.
+
+In theory, the dataset should contain at least one chosen and one rejected completion. However, some users have successfully run KTO using *only* chosen or only rejected data. If using only rejected data, it is advisable to adopt a conservative learning rate.
+
+## Example script
+
+We provide an example script to train a model using the KTO method. The script is available in [`trl/scripts/kto.py`](https://github.com/huggingface/trl/blob/main/trl/scripts/kto.py)
+
+To test the KTO script with the [Qwen2 0.5B model](https://huggingface.co/Qwen/Qwen2-0.5B-Instruct) on the [UltraFeedback dataset](https://huggingface.co/datasets/trl-lib/kto-mix-14k), run the following command:
+
+```bash
+accelerate launch trl/scripts/kto.py \
+    --model_name_or_path Qwen/Qwen2-0.5B-Instruct \
+    --dataset_name trl-lib/kto-mix-14k \
+    --num_train_epochs 1 \
+    --output_dir Qwen2-0.5B-KTO
+```
+
+## Usage tips
+
+### For Mixture of Experts Models: Enabling the auxiliary loss
+
+MOEs are the most efficient if the load is about equally distributed between experts.  
+To ensure that we train MOEs similarly during preference-tuning, it is beneficial to add the auxiliary loss from the load balancer to the final loss.
+
+This option is enabled by setting `output_router_logits=True` in the model config (e.g. [`~transformers.MixtralConfig`]).  
+To scale how much the auxiliary loss contributes to the total loss, use the hyperparameter `router_aux_loss_coef=...` (default: `0.001`) in the model config.
+
+
+### Batch size recommendations
+
+Use a per-step batch size that is at least 4, and an effective batch size between 16 and 128. Even if your effective batch size is large, if your per-step batch size is poor, then the KL estimate in KTO will be poor.
+
+### Learning rate recommendations
+
+Each choice of `beta` has a maximum learning rate it can tolerate before learning performance degrades. For the default setting of `beta = 0.1`, the learning rate should typically not exceed `1e-6` for most models. As `beta` decreases, the learning rate should also be reduced accordingly. In general, we strongly recommend keeping the learning rate between `5e-7` and `5e-6`. Even with small datasets, we advise against using a learning rate outside this range. Instead, opt for more epochs to achieve better results.
+
+### Imbalanced data
+
+The `desirable_weight` and `undesirable_weight` of the [`KTOConfig`] refer to the weights placed on the losses for desirable/positive and undesirable/negative examples.
+By default, they are both 1. However, if you have more of one or the other, then you should upweight the less common type such that the ratio of (`desirable_weight`  \\(\times\\) number of positives) to (`undesirable_weight`  \\(\times\\) number of negatives) is in the range 1:1 to 4:3.
+
+## Logged metrics
+
+While training and evaluating we record the following reward metrics:
+
+- `rewards/chosen_sum`: the sum of log probabilities of the policy model for the chosen responses scaled by beta
+- `rewards/rejected_sum`: the sum of log probabilities of the policy model for the rejected responses scaled by beta
+- `logps/chosen_sum`: the sum of log probabilities of the chosen completions
+- `logps/rejected_sum`: the sum of log probabilities of the rejected completions
+- `logits/chosen_sum`: the sum of logits of the chosen completions
+- `logits/rejected_sum`: the sum of logits of the rejected completions
+- `count/chosen`: the count of chosen samples in a batch
+- `count/rejected`: the count of rejected samples in a batch
+
+## KTOTrainer
+
+[[autodoc]] KTOTrainer
+
+## KTOConfig
+
+[[autodoc]] KTOConfig

docs/source/kto_trainer.mdx

@@ -1,102 +0,0 @@
-# KTO Trainer
-
-TRL supports the Kahneman-Tversky Optimization (KTO) Trainer for aligning language models with binary feedback data (e.g., upvote/downvote), as described in the [paper](https://huggingface.co/papers/2402.01306) by Kawin Ethayarajh, Winnie Xu, Niklas Muennighoff, Dan Jurafsky, and Douwe Kiela.
-For a full example have a look at  [`examples/scripts/kto.py`].
-
-Depending on how good your base model is, you may or may not need to do SFT before KTO.
-This is different from standard RLHF and DPO, which always require SFT.
-
-## Expected dataset format
-
-The KTO trainer expects a very specific format for the dataset as it does not require pairwise preferences. Since the model will be trained to directly optimize examples that consist of a prompt, model completion, and a label to indicate whether the completion is "good" or "bad", we expect a dataset with the following columns:
-
-- `prompt`
-- `completion`
-- `label`
-
-for example:
-
-```
-kto_dataset_dict = {
-    "prompt": [
-        "Hey, hello",
-        "How are you",
-        "What is your name?",
-        "What is your name?",
-        "Which is the best programming language?",
-        "Which is the best programming language?",
-        "Which is the best programming language?",
-    ],
-    "completion": [
-        "hi nice to meet you",
-        "leave me alone",
-        "I don't have a name",
-        "My name is Mary",
-        "Python",
-        "C++",
-        "Java",
-    ],
-    "label": [
-        True,
-        False,
-        False,
-        True,
-        True,
-        False,
-        False,
-    ],
-}
-```
-
-where the `prompt` contains the context inputs, `completion` contains the corresponding responses and `label` contains the corresponding flag that indicates if the generated completion is desired (`True`) or undesired (`False`).
-A prompt can have multiple responses and this is reflected in the entries being repeated in the dictionary's value arrays. It is required that the dataset contains at least one desirable and one undesirable completion.
-
-
-## Expected model format
-The KTO trainer expects a model of `AutoModelForCausalLM`, compared to PPO that expects `AutoModelForCausalLMWithValueHead` for the value function.
-
-## Using the `KTOTrainer`
-
-For a detailed example have a look at the `examples/scripts/kto.py` script. At a high level we need to initialize the `KTOTrainer` with a `model` we wish to train and a reference `ref_model` which we will use to calculate the implicit rewards of the preferred and rejected response. 
-
-The `beta` refers to the hyperparameter of the implicit reward, and the dataset contains the 3 entries listed above. Note that the `model` and `ref_model` need to have the same architecture (ie decoder only or encoder-decoder).
-
-The `desirable_weight` and `undesirable_weight` refer to the weights placed on the losses for desirable/positive and undesirable/negative examples.
-By default, they are both 1. However, if you have more of one or the other, then you should upweight the less common type such that the ratio of (`desirable_weight` * number of positives) to (`undesirable_weight` * number of negatives) is in the range 1:1 to 4:3.
-
-```py
-training_args = KTOConfig(
-    beta=0.1,
-    desirable_weight=1.0,
-    undesirable_weight=1.0,
-)
-
-kto_trainer = KTOTrainer(
-    model,
-    ref_model,
-    args=training_args,
-    train_dataset=train_dataset,
-    tokenizer=tokenizer,
-)
-```
-After this one can then call:
-
-```py
-kto_trainer.train()
-```
-
-### For Mixture of Experts Models: Enabling the auxiliary loss
-
-MOEs are the most efficient if the load is about equally distributed between experts.  
-To ensure that we train MOEs similarly during preference-tuning, it is beneficial to add the auxiliary loss from the load balancer to the final loss.  
-
-This option is enabled by setting `output_router_logits=True` in the model config (e.g. MixtralConfig).  
-To scale how much the auxiliary loss contributes to the total loss, use the hyperparameter `router_aux_loss_coef=...` (default: 0.001).
-
-## KTOTrainer
-
-[[autodoc]] KTOTrainer
-
-## KTOConfig
-
-[[autodoc]] KTOConfig

docs/source/learning_tools.mdx

@@ -1,232 +0,0 @@
-# Learning Tools (Experimental 🧪)
-
-Using Large Language Models (LLMs) with tools has been a popular topic recently with awesome works such as [ToolFormer](https://huggingface.co/papers/2302.04761) and [ToolBench](https://huggingface.co/papers/2305.16504). In TRL, we provide a simple example of how to teach LLM to use tools with reinforcement learning. 
-
-
-Here's an overview of the scripts in the [trl repository](https://github.com/lvwerra/trl/tree/main/examples/research_projects/tools):
-
-| File | Description | 
-|---|---| 
-| [`calculator.py`](https://github.com/lvwerra/trl/blob/main/examples/research_projects/tools/calculator.py) | Script to train LLM to use a calculator with reinforcement learning. |
-| [`triviaqa.py`](https://github.com/lvwerra/trl/blob/main/examples/research_projects/tools/triviaqa.py) | Script to train LLM to use a wiki tool to answer questions. |
-| [`python_interpreter.py`](https://github.com/lvwerra/trl/blob/main/examples/research_projects/tools/python_interpreter.py) | Script to train LLM to use python interpreter to solve math puzzles. |
-
-<Tip warning={true}>
-
-Note that the scripts above rely heavily on the `TextEnvironment` API which is still under active development. The API may change in the future. Please see [`TextEnvironment`](text_environment) for the related docs.
-</Tip>
-
-
-## Learning to Use a Calculator
-
-
-The rough idea is as follows:
-
-1. Load a tool such as [ybelkada/simple-calculator](https://huggingface.co/spaces/ybelkada/simple-calculator) that parse a text calculation like `"14 + 34"` and return the calulated number:
-    ```python
-    from transformers import AutoTokenizer, load_tool
-    tool = load_tool("ybelkada/simple-calculator")
-    tool_fn = lambda text: str(round(float(tool(text)), 2))  # rounding to 2 decimal places
-    ```
-1. Define a reward function that returns a positive reward if the tool returns the correct answer. In the script we create a dummy reward function like `reward_fn = lambda x: 1`, but we override the rewards directly later.
-1. Create a prompt on how to use the tools
-    ```python
-    # system prompt
-    prompt = """\
-    What is 13.1-3?
-
-    <request><SimpleCalculatorTool>13.1-3<call>10.1<response>
-
-    Result=10.1<submit>
-
-    What is 4*3?
-
-    <request><SimpleCalculatorTool>4*3<call>12<response>
-
-    Result=12<submit>
-
-    What is 12.1+1?
-
-    <request><SimpleCalculatorTool>12.1+1<call>13.1<response>
-
-    Result=13.1<submit>
-
-    What is 12.1-20?
-
-    <request><SimpleCalculatorTool>12.1-20<call>-7.9<response>
-
-    Result=-7.9<submit>"""
-    ```
-3. Create a `trl.TextEnvironment` with the model 
-    ```python
-    env = TextEnvironment(
-        model,
-        tokenizer,
-        {"SimpleCalculatorTool": tool_fn},
-        reward_fn,
-        prompt,
-        generation_kwargs=generation_kwargs,
-    )
-    ```
-4. Then generate some data such as `tasks = ["\n\nWhat is 13.1-3?", "\n\nWhat is 4*3?"]` and run the environment with `queries, responses, masks, rewards, histories = env.run(tasks)`. The environment will look for the `<call>` token in the prompt and append the tool output to the response; it will also return the mask associated with the response. You can further use the `histories` to visualize the interaction between the model and the tool; `histories[0].show_text()` will show the text with color-coded tool output and `histories[0].show_tokens(tokenizer)` will show visualize the tokens.
-    ![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/learning_tools.png)
-1. Finally, we can train the model with `train_stats = ppo_trainer.step(queries, responses, rewards, masks)`. The trainer will use the mask to ignore the tool output when computing the loss, make sure to pass that argument to `step`.
-
-## Experiment results
-
-We trained a model with the above script for 10 random seeds. You can reproduce the run with the following command. Feel free to remove the `--slurm-*` arguments if you don't have access to a slurm cluster.
-
-```
-WANDB_TAGS="calculator_final" python benchmark/benchmark.py \
-    --command "python examples/research_projects/tools/calculator.py" \
-    --num-seeds 10 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 8 \
-    --slurm-template-path benchmark/trl.slurm_template
-```
-
-We can then use [`openrlbenchmark`](https://github.com/openrlbenchmark/openrlbenchmark) which generates the following plot.
-```
-python -m openrlbenchmark.rlops_multi_metrics \
-    --filters '?we=openrlbenchmark&wpn=trl&xaxis=_step&ceik=trl_ppo_trainer_config.value.tracker_project_name&cen=trl_ppo_trainer_config.value.log_with&metrics=env/reward_mean&metrics=objective/kl' \
-        'wandb?tag=calculator_final&cl=calculator_mask' \
-    --env-ids trl \
-    --check-empty-runs \
-    --pc.ncols 2 \
-    --pc.ncols-legend 1 \
-    --output-filename static/0compare \
-    --scan-history
-```
-
-![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/learning_tools_chart.png)
-
-As we can see, while 1-2 experiments crashed for some reason, most of the runs obtained near perfect proficiency in the calculator task.
-
-
-## (Early Experiments 🧪): learning to use a wiki tool for question answering
-
-In the [ToolFormer](https://huggingface.co/papers/2302.04761) paper, it shows an interesting use case that utilizes a Wikipedia Search tool to help answer questions. In this section, we attempt to perform similar experiments but uses RL instead to teach the model to use a wiki tool on the [TriviaQA](https://nlp.cs.washington.edu/triviaqa/) dataset.
-
-
-<Tip warning={true}>
-
-**Note that many settings are different so the results are not directly comparable.**
-</Tip>
-
-
-
-
-### Building a search index
-
-Since [ToolFormer](https://huggingface.co/papers/2302.04761) did not open source, we needed to first replicate the search index. It is mentioned in their paper that the authors built the search index using a BM25 retriever that indexes the Wikipedia dump from [KILT](https://github.com/facebookresearch/KILT)
-
-Fortunately, [`pyserini`](https://github.com/castorini/pyserini) already implements the BM25 retriever and provides a prebuilt index for the KILT Wikipedia dump. We can use the following code to search the index.
-
-```python
-from pyserini.search.lucene import LuceneSearcher
-import json
-searcher = LuceneSearcher.from_prebuilt_index('wikipedia-kilt-doc')
-def search(query):
-    hits = searcher.search(query, k=1)
-    hit = hits[0]
-    contents = json.loads(hit.raw)['contents']
-    return contents
-print(search("tennis racket"))
-```
-```
-Racket (sports equipment)
-A racket or racquet is a sports implement consisting of a handled frame with an open hoop across which a network of strings or catgut is stretched tightly. It is used for striking a ball or shuttlecock in games such as squash, tennis, racquetball, and badminton. Collectively, these games are known as racket sports. Racket design and manufacturing has changed considerably over the centuries.
-
-The frame of rackets for all sports was traditionally made of solid wood (later laminated wood) and the strings of animal intestine known as catgut. The traditional racket size was limited by the strength and weight of the wooden frame which had to be strong enough to hold the strings and stiff enough to hit the ball or shuttle. Manufacturers started adding non-wood laminates to wood rackets to improve stiffness. Non-wood rackets were made first of steel, then of aluminum, and then carbon fiber composites. Wood is still used for real tennis, rackets, and xare. Most rackets are now made of composite materials including carbon fiber or fiberglass, metals such as titanium alloys, or ceramics.
-...
-```
-
-We then basically deployed this snippet as a Hugging Face space [here](https://huggingface.co/spaces/vwxyzjn/pyserini-wikipedia-kilt-doc), so that we can use the space as a `transformers.Tool` later.
-
-![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/pyserini.png)
-
-### Experiment settings
-
-We use the following settings:
-
-* use the `bigcode/starcoderbase` model as the base model
-* use the `pyserini-wikipedia-kilt-doc` space as the wiki tool and only uses the first paragrahs of the search result, allowing the `TextEnvironment` to obtain at most `max_tool_reponse=400` response tokens from the tool.
-* test if the response contain the answer string, if so, give a reward of 1, otherwise, give a reward of 0.
-    * notice this is a simplified evaluation criteria. In [ToolFormer](https://huggingface.co/papers/2302.04761), the authors checks if the first 20 words of the response contain the correct answer.
-* used the following prompt that demonstrates the usage of the wiki tool.
-```python
-prompt = """\
-Answer the following question:
-
-Q: In which branch of the arts is Patricia Neary famous?
-A: Ballets
-A2: <request><Wiki>Patricia Neary<call>Patricia Neary (born October 27, 1942) is an American ballerina, choreographer and ballet director, who has been particularly active in Switzerland. She has also been a highly successful ambassador for the Balanchine Trust, bringing George Balanchine's ballets to 60 cities around the globe.<response>
-Result=Ballets<submit>
-
-Q: Who won Super Bowl XX?
-A: Chicago Bears
-A2: <request><Wiki>Super Bowl XX<call>Super Bowl XX was an American football game between the National Football Conference (NFC) champion Chicago Bears and the American Football Conference (AFC) champion New England Patriots to decide the National Football League (NFL) champion for the 1985 season. The Bears defeated the Patriots by the score of 46–10, capturing their first NFL championship (and Chicago's first overall sports victory) since 1963, three years prior to the birth of the Super Bowl. Super Bowl XX was played on January 26, 1986 at the Louisiana Superdome in New Orleans.<response>
-Result=Chicago Bears<submit>
-
-Q: """
-```
-
-
-### Result and Discussion
-
-
-Our experiments show that the agent can learn to use the wiki tool to answer questions. The learning curves would go up mostly, but one of the experiment did crash.
-
-![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/triviaqa_learning_curves.png)
-
-Wandb report is [here](https://wandb.ai/costa-huang/cleanRL/reports/TriviaQA-Final-Experiments--Vmlldzo1MjY0ODk5) for further inspection.
-
-
-Note that the correct rate of the trained model is on the low end, which could be due to the following reasons:
-
-* **incorrect searches:** When given the question `"What is Bruce Willis' real first name?"` if the model searches for `Bruce Willis`, our wiki tool returns "Patrick Poivey (born 18 February 1948) is a French actor. He is especially known for his voice: he is the French dub voice of Bruce Willis since 1988.` But a correct search should be `Walter Bruce Willis (born March 19, 1955) is an American former actor. He achieved fame with a leading role on the comedy-drama series Moonlighting (1985–1989) and appeared in over a hundred films, gaining recognition as an action hero after his portrayal of John McClane in the Die Hard franchise (1988–2013) and other roles.[1][2]"
-
-
-    ![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/real_first_name.png)
-
-* **unnecessarily long response**: The wiki tool by default sometimes output very long sequences. E.g., when the wiki tool searches for "Brown Act"
-    * Our wiki tool returns "The Ralph M. Brown Act, located at California Government Code 54950 "et seq.", is an act of the California State Legislature, authored by Assemblymember Ralph M. Brown and passed in 1953, that guarantees the public's right to attend and participate in meetings of local legislative bodies."
-    * [ToolFormer](https://huggingface.co/papers/2302.04761)'s wiki tool returns "The Ralph M. Brown Act is an act of the California State Legislature that guarantees the public's right to attend and participate in meetings of local legislative bodies." which is more succinct.
-
-    ![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/brown_act.png)
-
-
-## (Early Experiments 🧪): solving math puzzles with python interpreter
-
-In this section, we attempt to teach the model to use a python interpreter to solve math puzzles. The rough idea is to give the agent a prompt like the following:
-
-```python
-prompt = """\
-Example of using a Python API to solve math questions.
-
-Q: Olivia has $23. She bought five bagels for $3 each. How much money does she have left?
-
-<request><PythonInterpreter>
-def solution():
-    money_initial = 23
-    bagels = 5
-    bagel_cost = 3
-    money_spent = bagels * bagel_cost
-    money_left = money_initial - money_spent
-    result = money_left
-    return result
-print(solution())
-<call>72<response>
-
-Result = 72 <submit>
-
-Q: """
-```
-
-
-Training experiment can be found at https://wandb.ai/lvwerra/trl-gsm8k/runs/a5odv01y
-
-![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/gms8k_learning_curve.png)

docs/source/liger_kernel_integration.md

@@ -0,0 +1,32 @@
+# Liger Kernel Integration
+
+<Tip warning={true}>
+
+Section under construction. Feel free to contribute!
+
+</Tip>
+
+[Liger Kernel](https://github.com/linkedin/Liger-Kernel) is a collection of Triton kernels designed specifically for LLM training. It can effectively increase multi-GPU training throughput by 20% and reduce memory usage by 60%. That way, we can **4x** our context length, as described in the benchmark below. They have implemented Hugging Face compatible `RMSNorm`, `RoPE`, `SwiGLU`, `CrossEntropy`, `FusedLinearCrossEntropy`, with more to come. The kernel works out of the box with [Flash Attention](https://github.com/Dao-AILab/flash-attention), [PyTorch FSDP](https://pytorch.org/tutorials/intermediate/FSDP_tutorial.html), and [Microsoft DeepSpeed](https://github.com/microsoft/DeepSpeed).
+
+With this memory reduction, you can potentially turn off `cpu_offloading` or gradient checkpointing to further boost the performance.
+
+| Speed Up                 | Memory Reduction        |
+|--------------------------|-------------------------|
+| ![Speed up](https://raw.githubusercontent.com/linkedin/Liger-Kernel/main/docs/images/e2e-tps.png) | ![Memory](https://raw.githubusercontent.com/linkedin/Liger-Kernel/main/docs/images/e2e-memory.png) |
+
+1. To use Liger-Kernel in [`SFTTrainer`], first install it by:
+
+```bash
+pip install liger-kernel
+```
+
+2. Once installed, set `use_liger_kernel` in [`SFTConfig`]. No other changes are needed!
+
+```python
+training_args = SFTConfig(
+    use_liger_kernel=True,
+    ...
+)
+```
+
+To learn more about Liger-Kernel, visit their [official repository](https://github.com/linkedin/Liger-Kernel/).

docs/source/logging.md

@@ -0,0 +1,102 @@
+# Logging
+
+As reinforcement learning algorithms are historically challenging to debug, it's important to pay careful attention to logging.
+By default, TRL trainers like [`PPOTrainer`] and [`GRPOTrainer`] save a lot of relevant information to supported experiment trackers like Weights & Biases (wandb) or TensorBoard.
+
+Upon initialization, pass the `report_to` argument to the respective configuration object (e.g., [`PPOConfig`] for `PPOTrainer`, or [`GRPOConfig`] for `GRPOTrainer`):
+
+```python
+# For PPOTrainer
+ppo_config = PPOConfig(
+    # ...,
+    report_to="wandb"  # or "tensorboard"
+)
+
+# For GRPOTrainer
+grpc_config = GRPOConfig(
+    # ...,
+    report_to="wandb"  # or "tensorboard"
+)
+```
+
+If you want to log with TensorBoard, you might also need to specify logging directories, for example, by adding `logging_dir=PATH_TO_LOGS` to the configuration object (e.g., `PPOConfig` or `GRPOConfig`).
+
+## PPO Logging
+
+Here's a brief explanation for the logged metrics provided in the data:
+
+* `eps`: Tracks the number of episodes per second.
+* `objective/kl`: The mean Kullback-Leibler (KL) divergence between the current policy and reference policy.
+* `objective/entropy`: The mean entropy of the policy, indicating the randomness of the actions chosen by the policy.
+* `objective/non_score_reward`: The mean reward from non-score-related sources, basically `beta * kl.sum(1)`, where `beta` is the KL penalty coefficient and `kl` is the per-token KL divergence.
+* `objective/rlhf_reward`: The mean RLHF reward, which is `score - non_score_reward`.
+* `objective/scores`: The mean scores returned by the reward model / environment.
+* `policy/approxkl_avg`: The average approximate KL divergence between consecutive PPO policies. Note that this is not the same as `objective/kl`.
+* `policy/clipfrac_avg`: The average fraction of policy updates that are clipped, indicating how often the policy updates are constrained to prevent large changes.
+* `loss/policy_avg`: The average policy loss, indicating how well the policy is performing.
+* `loss/value_avg`: The average value loss, indicating the difference between the predicted value and the actual reward.
+* `val/clipfrac_avg`: The average fraction of value function updates that are clipped, similar to `policy/clipfrac_avg` but for the value function.
+* `policy/entropy_avg`: The average entropy of the policy during training, indicating how diverse the policy's actions are.
+* `val/ratio`: The mean ratio of the current policy probability to the old policy probability, providing a measure of how much the policy has changed.
+* `val/ratio_var`: The variance of the `val/ratio`, indicating the variability in policy changes.
+* `val/num_eos_tokens`: The number of end-of-sequence (EOS) tokens generated, which can indicate the number of complete responses.
+* `lr`: The current learning rate used by the optimizer.
+* `episode`: The current episode count in the training process.
+
+### Crucial values
+During training, many values are logged, here are the most important ones:
+
+1. `objective/scores`: The mean scores returned by the reward model / environment.
+1. `objective/rlhf_reward`: The mean RLHF reward. This is the ultimate objective of the RLHF training. If training works as intended, this metric should keep going up.
+1. `objective/non_score_reward`: The mean reward from non-score-related sources (e.g., KL penalty).
+
+Here are some parameters that are useful to monitor for stability (when these diverge or collapse to 0, try tuning variables):
+
+1. `loss/value_avg`: The average value loss. It will spike / NaN when not going well.
+1. `val/ratio`: The mean ratio of the current policy probability to the old policy probability. This number should float around 1.0. If this `ratio` is too high (e.g., 2.0 or 1000.0) or too small (e.g., 0.1), it means the updates between consecutive policies are too drastic.
+1. `policy/clipfrac_avg` and `policy/approxkl_avg`: If `val/ratio` is too high, the `ratio` is going to get clipped, resulting in high `policy/clipfrac_avg` and high `policy/approxkl_avg` as well.
+1. `objective/kl`: The mean KL divergence. It should stay positive and ideally not too large, so that the policy is not too far away from the reference policy.
+
+## GRPO Logging
+
+Here's a brief explanation for the logged metrics provided in the data for the GRPO trainer:
+
+* `num_tokens`: Total number of input tokens processed during training so far.
+
+**Completions:**
+* `completions/mean_length`: Mean length of all generated completions (including those not ending with an EOS token).
+* `completions/min_length`: Minimum length among all generated completions.
+* `completions/max_length`: Maximum length among all generated completions.
+* `completions/clipped_ratio`: The ratio of completions that did not end with an EOS token before reaching the maximum generation length (i.e., they were truncated).
+* `completions/mean_terminated_length`: Mean length of only those completions that successfully ended with an EOS token.
+* `completions/min_terminated_length`: Minimum length among completions that ended with an EOS token.
+* `completions/max_terminated_length`: Maximum length among completions that ended with an EOS token.
+
+**Rewards:**
+* `rewards/{reward_func_name}/mean`: The mean reward obtained from a specific, named reward function (e.g., `rewards/my_custom_reward/mean`). This is logged for each reward function used.
+* `rewards/{reward_func_name}/std`: The standard deviation of rewards from a specific, named reward function.
+* `reward`: The overall mean of the (potentially weighted and, if `args.scale_rewards` is true, normalized) rewards, after group-wise normalization (advantages).
+* `reward_std`: The standard deviation of the (potentially weighted) rewards *before* group-wise normalization for advantages.
+
+**Policy and Loss Metrics:**
+* `kl`: The mean Kullback-Leibler (KL) divergence between the current policy and the reference policy. This is logged only if `beta` (the KL coefficient in `GRPOConfig`) is non-zero.
+* `entropy`: Average entropy of token predictions across generated completions.
+* If Liger GRPOLoss is used (`use_liger_loss: True` in `GRPOConfig`):
+    *   `clip_ratio`: The fraction of policy updates where the probability ratio was clipped according to the GRPO loss's epsilon bounds.
+* If standard GRPOLoss is used (`use_liger_loss: False`):
+    *   `clip_ratio/low_mean`: The mean fraction of instances where the probability ratio `r_t(θ)` was clipped at the lower bound `1 - epsilon_low` (occurs when advantage is negative and ratio is below the bound).
+    *   `clip_ratio/low_min`: The minimum observed fraction for `clip_ratio/low_mean` across batches/processes.
+    *   `clip_ratio/high_mean`: The mean fraction of instances where the probability ratio `r_t(θ)` was clipped at the upper bound `1 + epsilon_high` (occurs when advantage is positive and ratio is above the bound).
+    *   `clip_ratio/high_max`: The maximum observed fraction for `clip_ratio/high_mean` across batches/processes.
+    *   `clip_ratio/region_mean`: The mean fraction of instances where the probability ratio was clipped at either the lower or upper bound.
+
+### Crucial GRPO values
+During GRPO training, monitor these values for insights into performance and stability:
+
+1.  `reward`: This is the primary objective. It reflects the (group-wise normalized) rewards the policy is achieving. It should generally increase during successful training.
+1.  `kl`: If `beta > 0`, this tracks the divergence from the reference model. Keep an eye on it to ensure the policy doesn't stray too far, which can lead to instability.
+1.  `clip_ratio/*` (either `clip_ratio` for Liger loss or the more detailed `clip_ratio/...` metrics for standard loss): These indicate how often the policy updates are being constrained by the GRPO clipping mechanism. Very high values might suggest that the policy is trying to change too drastically (potentially due to large advantages or a learning rate that's too high) or that the epsilon clipping range is too restrictive.
+1.  `completions/clipped_ratio`: A high ratio here indicates that the model is frequently generating completions that are cut off by `max_completion_length` rather than naturally ending with an EOS token. This might suggest issues with learning sequence termination or that `max_completion_length` is too short.
+1. `rewards/{reward_func_name}/mean`: Monitoring the mean of individual reward functions can help diagnose which aspects of the desired behavior the model is learning or struggling with, especially when using multiple reward sources.
+1. `entropy`: Measures how uncertain the policy is in its action choices, higher entropy suggests more exploration. A collapse in entropy means the policy is becoming overconfident and deterministic, often too early. This can stall learning by reducing exploration and making updates overly biased. Stable but non-zero entropy is usually a sign that the policy retains flexibility and continues to explore.
+

docs/source/logging.mdx

@@ -1,75 +0,0 @@
-# Logging
-
-As reinforcement learning algorithms are historically challenging to debug, it's important to pay careful attention to logging.
-By default, the TRL [`PPOTrainer`] saves a lot of relevant information to `wandb` or `tensorboard`.
-
-Upon initialization, pass one of these two options to the [`PPOConfig`]:
-```
-config = PPOConfig(
-    model_name=args.model_name,
-    log_with=`wandb`, # or `tensorboard`
-)
-```
-If you want to log with tensorboard, add the kwarg `project_kwargs={"logging_dir": PATH_TO_LOGS}` to the PPOConfig.
-
-## PPO Logging
-
-Here's a brief explanation for the logged metrics provided in the data:
-
-Key metrics to monitor. We want to maximize the reward, maintain a low KL divergence, and maximize entropy:
-1. `env/reward_mean`: The average reward obtained from the environment. Alias `ppo/mean_scores`, which is sed to specifically monitor the reward model.
-1. `env/reward_std`: The standard deviation of the reward obtained from the environment. Alias ``ppo/std_scores`, which is sed to specifically monitor the reward model.
-1. `env/reward_dist`: The histogram distribution of the reward obtained from the environment.
-1. `objective/kl`: The mean Kullback-Leibler (KL) divergence between the old and new policies. It measures how much the new policy deviates from the old policy. The KL divergence is used to compute the KL penalty in the objective function.
-1. `objective/kl_dist`: The histogram distribution of the `objective/kl`.
-1. `objective/kl_coef`: The coefficient for Kullback-Leibler (KL) divergence in the objective function. 
-1. `ppo/mean_non_score_reward`: The **KL penalty** calculated by `objective/kl * objective/kl_coef` as the total reward for optimization to prevent the new policy from deviating too far from the old policy.
-1. `objective/entropy`: The entropy of the model's policy, calculated by `-logprobs.sum(-1).mean()`. High entropy means the model's actions are more random, which can be beneficial for exploration.
-
-Training stats:
-1. `ppo/learning_rate`: The learning rate for the PPO algorithm.
-1. `ppo/policy/entropy`: The entropy of the model's policy, calculated by `pd = torch.nn.functional.softmax(logits, dim=-1); entropy = torch.logsumexp(logits, dim=-1) - torch.sum(pd * logits, dim=-1)`. It measures the randomness of the policy.
-1. `ppo/policy/clipfrac`: The fraction of probability ratios (old policy / new policy) that fell outside the clipping range in the PPO objective. This can be used to monitor the optimization process.
-1. `ppo/policy/approxkl`: The approximate KL divergence between the old and new policies, measured by `0.5 * masked_mean((logprobs - old_logprobs) ** 2, mask)`, corresponding to the `k2` estimator in http://joschu.net/blog/kl-approx.html
-1. `ppo/policy/policykl`: Similar to `ppo/policy/approxkl`, but measured by `masked_mean(old_logprobs - logprobs, mask)`, corresponding to the `k1` estimator in http://joschu.net/blog/kl-approx.html
-1. `ppo/policy/ratio`:  The histogram distribution of the ratio between the new and old policies, used to compute the PPO objective.
-1. `ppo/policy/advantages_mean`: The average of the GAE (Generalized Advantage Estimation) advantage estimates. The advantage function measures how much better an action is compared to the average action at a state.
-1. `ppo/policy/advantages`: The histogram distribution of `ppo/policy/advantages_mean`.
-1. `ppo/returns/mean`: The mean of the TD(λ) returns, calculated by `returns = advantage + values`, another indicator of model performance. See https://iclr-blog-track.github.io/2022/03/25/ppo-implementation-details/ for more details.
-1. `ppo/returns/var`: The variance of the TD(λ) returns, calculated by `returns = advantage + values`, another indicator of model performance.
-1. `ppo/val/mean`: The mean of the values, used to monitor the value function's performance.
-1. `ppo/val/var` : The variance of the values, used to monitor the value function's performance.
-1. `ppo/val/var_explained`: The explained variance for the value function, used to monitor the value function's performance.
-1. `ppo/val/clipfrac`: The fraction of the value function's predicted values that are clipped.
-1. `ppo/val/vpred`: The predicted values from the value function.
-1. `ppo/val/error`: The mean squared error between the `ppo/val/vpred` and returns, used to monitor the value function's performance.
-1. `ppo/loss/policy`: The policy loss for the Proximal Policy Optimization (PPO) algorithm.
-1. `ppo/loss/value`: The loss for the value function in the PPO algorithm. This value quantifies how well the function estimates the expected future rewards.
-1. `ppo/loss/total`: The total loss for the PPO algorithm. It is the sum of the policy loss and the value function loss.
-
-
-Stats on queries, responses, and logprobs:
-1. `tokens/queries_len_mean`: The average length of the queries tokens.
-1. `tokens/queries_len_std`: The standard deviation of the length of the queries tokens.
-1. `tokens/queries_dist`: The histogram distribution of the length of the queries tokens.
-1. `tokens/responses_len_mean`: The average length of the responses tokens.
-1. `tokens/responses_len_std`: The standard deviation of the length of the responses tokens.
-1. `tokens/responses_dist`: The histogram distribution of the length of the responses tokens. (Costa: inconsistent naming, should be `tokens/responses_len_dist`)
-1. `objective/logprobs`: The histogram distribution of the log probabilities of the actions taken by the model.
-1. `objective/ref_logprobs`: The histogram distribution of the log probabilities of the actions taken by the reference model.
-
-
-
-### Crucial values
-During training, many values are logged, here are the most important ones:
-
-1. `env/reward_mean`,`env/reward_std`, `env/reward_dist`: the properties of the reward distribution from the "environment" /  reward model
-1. `ppo/mean_non_score_reward`: The mean negated KL penalty during training (shows the delta between the reference model and the new policy over the batch in the step)
-
-Here are some parameters that are useful to monitor for stability (when these diverge or collapse to 0, try tuning variables):
-
-1. `ppo/loss/value`: it will spike / NaN when not going well.
-1. `ppo/policy/ratio`: `ratio` being 1 is a baseline value, meaning that the probability of sampling a token is the same under the new and old policy. If the ratio is too high like 200, it means the probability of sampling a token is 200 times higher under the new policy than the old policy. This is a sign that the new policy is too different from the old policy, which will likely cause overoptimization and collapse training later on.
-1. `ppo/policy/clipfrac` and `ppo/policy/approxkl`: if `ratio` is too high, the `ratio` is going to get clipped, resulting in high `clipfrac` and high `approxkl` as well.
-1. `objective/kl`: it should stay positive so that the policy is not too far away from the reference policy.
-1. `objective/kl_coef`: The target coefficient with [`AdaptiveKLController`]. Often increases before numerical instabilities.

docs/source/model_utils.md

@@ -0,0 +1,9 @@
+# Model Utilities
+
+## clone_chat_template
+
+[[autodoc]] clone_chat_template
+
+## get_act_offloading_ctx_manager
+
+[[autodoc]] models.get_act_offloading_ctx_manager

docs/source/multi_adapter_rl.md

@@ -10,7 +10,7 @@ You just need to install `peft` and optionally install `bitsandbytes` as well if
 
 You need to address this approach in three stages that we summarize as follows:
 
-1- Train a base model on the target domain (e.g. `imdb` dataset) - this is the Supervised Fine Tuning stage - it can leverage the `SFTTrainer` from TRL.
+1- Train a base model on the target domain (e.g. [IMDB dataset](https://huggingface.co/datasets/stanfordnlp/imdb)) - this is the Supervised Fine Tuning stage - it can leverage the `SFTTrainer` from TRL.
 2- Train a reward model using `peft`. This is required in order to re-use the adapter during the RL optimisation process (step 3 below). We show an example of leveraging the `RewardTrainer` from TRL in [this example](https://github.com/huggingface/trl/tree/main/examples/scripts/reward_modeling.py)
 3- Fine tune new adapters on the base model using PPO and the reward adapter. ("0 abstraction RL")
 

docs/source/nash_md_trainer.md

@@ -0,0 +1,158 @@
+# Nash-MD Trainer
+
+[![](https://img.shields.io/badge/All_models-Nash--MD-blue)](https://huggingface.co/models?other=nash-md,trl)
+
+## Overview
+
+Nash-MD was proposed in the paper [Nash Learning from Human Feedback](https://huggingface.co/papers/2312.00886) by Rémi Munos, [Michal Valko](https://huggingface.co/misovalko), Daniele Calandriello, Mohammad Gheshlaghi Azar, Mark Rowland, Daniel Guo, Yunhao Tang, Matthieu Geist, Thomas Mésnard, and Andrea Michi. 
+
+The abstract from the paper is the following:
+
+> Reinforcement learning from human feedback (RLHF) has emerged as the main paradigm for aligning large language models (LLMs) with human preferences. Typically, RLHF involves the initial step of learning a reward model from human feedback, often expressed as preferences between pairs of text generations produced by a pre-trained LLM. Subsequently, the LLM's policy is fine-tuned by optimizing it to maximize the reward model through a reinforcement learning algorithm. However, an inherent limitation of current reward models is their inability to fully represent the richness of human preferences and their dependency on the sampling distribution. In this study, we introduce an alternative pipeline for the fine-tuning of LLMs using pairwise human feedback. Our approach entails the initial learning of a preference model, which is conditioned on two inputs given a prompt, followed by the pursuit of a policy that consistently generates responses preferred over those generated by any competing policy, thus defining the Nash equilibrium of this preference model. We term this approach Nash learning from human feedback (NLHF). In the context of a tabular policy representation, we present a novel algorithmic solution, Nash-MD, founded on the principles of mirror descent. This algorithm produces a sequence of policies, with the last iteration converging to the regularized Nash equilibrium. Additionally, we explore parametric representations of policies and introduce gradient descent algorithms for deep-learning architectures. To demonstrate the effectiveness of our approach, we present experimental results involving the fine-tuning of a LLM for a text summarization task. We believe NLHF offers a compelling avenue for preference learning and policy optimization with the potential of advancing the field of aligning LLMs with human preferences.
+
+This post-training method was contributed by [Kashif Rasul](https://huggingface.co/kashif) and [Daniil Tiapkin](https://huggingface.co/dtiapkin), [Pierre Ménard](https://huggingface.co/menardprr), Daniele Calandriello and [Quentin Gallouédec](https://huggingface.co/qgallouedec). 
+
+## Quick start
+
+This example demonstrates how to train a model using the Nash-MD method. We use the [Qwen 0.5B model](https://huggingface.co/Qwen/Qwen2-0.5B-Instruct) as the base model and [`PairRMJudge`] as a judge. We use the prompts from the [UltraFeedback dataset](https://huggingface.co/datasets/openbmb/UltraFeedback). You can view the prompts in the dataset here:
+
+<iframe
+  src="https://huggingface.co/datasets/trl-lib/ultrafeedback-prompt/embed/viewer/default/train?row=0"
+  frameborder="0"
+  width="100%"
+  height="560px"
+></iframe>
+
+Below is the script to train the model:
+
+```python
+# train_nash_md.py
+from datasets import load_dataset
+from trl import NashMDConfig, NashMDTrainer, PairRMJudge
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
+judge = PairRMJudge()
+train_dataset = load_dataset("trl-lib/ultrafeedback-prompt", split="train")
+
+training_args = NashMDConfig(output_dir="Qwen2-0.5B-NashMD")
+trainer = NashMDTrainer(
+    model=model, judge=judge, args=training_args, processing_class=tokenizer, train_dataset=train_dataset
+)
+trainer.train()
+```
+
+Execute the script using the following command:
+
+```bash
+accelerate launch train_nash_md.py
+```
+
+Distributed across 8 GPUs, the training takes approximately 3 hours.
+
+To see how the [trained model](https://huggingface.co/trl-lib/Qwen2-0.5B-NashMD) performs, you can use the [Transformers Chat CLI](https://huggingface.co/docs/transformers/quicktour#chat-with-text-generation-models).
+
+<pre><code>$ transformers chat trl-lib/Qwen2-0.5B-NashMD
+<strong><span style="color: red;">&lt;quentin_gallouedec&gt;:</span></strong>
+What is the best programming language?
+
+<strong><span style="color: blue;">&lt;trl-lib/Qwen2-0.5B-NashMD&gt;:</span></strong>
+The best programming language depends on personal preference, the complexity of the project, and the specific requirements of the task. Some programming languages that are often recommended include Python, Java, and JavaScript, and there are many other languages to choose from depending on individual needs.
+</code></pre>
+
+## Expected dataset type
+
+Nash-MD requires a [prompt-only dataset](dataset_formats#prompt-only). The [`NashMDTrainer`] supports both [conversational](dataset_formats#conversational) and [standard](dataset_formats#standard) dataset format. When provided with a conversational dataset, the trainer will automatically apply the chat template to the dataset.
+
+## Usage tips
+
+### Use a reward model
+
+Instead of a judge, you can chose to use a reward model -- see [Reward Bench](https://huggingface.co/spaces/allenai/reward-bench) for a leaderboard of public models you can use. Below is a code example showing how to replace a judge with the [trl-lib/Qwen2-0.5B-Reward](https://huggingface.co/trl-lib/Qwen2-0.5B-Reward) model:
+
+```diff
+- from trl import PairRMJudge
++ from transformers import AutoModelForSequenceClassification
+
+- judge = PairRMJudge()
++ reward_model = AutoModelForSequenceClassification.from_pretrained("trl-lib/Qwen2-0.5B-Reward", num_labels=1)
+
+  trainer = NashMDTrainer(
+      ...
+-     judge=judge,
++     reward_model=reward_model,
+  )
+```
+
+<Tip warning={true}>
+
+Make sure that the SFT model and reward model use the _same_ chat template and the same tokenizer. Otherwise, you may find the model completions are scored incorrectly during training.
+
+</Tip>
+
+### Encourage EOS token generation
+
+We may want the model to generate completions within a given length. During training, the model will generate completions up to the maximum length specified in the `max_new_tokens` argument of [`NashMDConfig`]. If you want to penalize the model for not generating an EOS token before reaching the maximum length, you can use the `missing_eos_penalty` argument of [`NashMDConfig`]:
+
+```python
+training_args = NashMDConfig(..., max_new_tokens=128, missing_eos_penalty=1.0)
+```
+
+### Logging Completions
+
+To better understand your model’s behavior during training, you can log sample completions periodically using the [`LogCompletionsCallback`].
+
+```python
+trainer = NashMDTrainer(..., eval_dataset=eval_dataset)
+completions_callback = LogCompletionsCallback(trainer, num_prompts=8)
+trainer.add_callback(completions_callback)
+```
+
+This callback logs the model's generated completions directly to Weights & Biases.
+
+![Logged Completions](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/wandb_completions.png)
+
+## Example script
+
+We provide an example script to train a model using the Nash-MD method. The script is available in [`examples/scripts/nash_md.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/nash_md.py)
+
+To test the online DPO script with the [Qwen2.5 0.5B model](https://huggingface.co/trl-lib/Qwen/Qwen2.5-0.5B-Instruct) on the [UltraFeedback dataset](https://huggingface.co/datasets/openbmb/UltraFeedback), run the following command:
+
+```bash
+python examples/scripts/nash_md.py \
+    --model_name_or_path Qwen/Qwen2.5-0.5B-Instruct \
+    --judge pair_rm \
+    --dataset_name trl-lib/ultrafeedback-prompt \
+    --learning_rate 5.0e-7 \
+    --output_dir Qwen2.5-0.5B-NashMD-PairRM \
+    --warmup_ratio 0.1 \
+    --push_to_hub
+```
+
+## Logged metrics
+
+The logged metrics are as follows:
+
+* `loss/kl`: The mean KL divergence between the model and reference data.
+* `objective/entropy`: The mean entropy of the model and reference data.
+* `loss/score`: The mean reinforce score loss.
+* `rewards/chosen`: The mean scores (according to the reward model) of the model completions.
+* `rewards/rejected`: The mean scores (according to the reward model) of the mixture completions.
+* `rewards/probabilities`: The mean probability (according to the reward model or judge) of the model completions chosen vs the mixture completion.
+* `rewards/accuracies`: The accuracies of the Nash-MD's implicit reward model.
+* `rewards/margins`: The mean reward margin (according to reward model) between the chosen and mixture completions.
+* `logps/chosen`: The mean log probabilities of the chosen completions.
+* `logps/rejected`: The mean log probabilities of the reference completions.
+* `val/model_contain_eos_token`: The amount of times the model's output contains the eos token.
+* `val/ref_contain_eos_token`: The amount of times the mixture's output contains the eos token.
+* `beta`: The parameter that controls the weight of the loss term representing the deviation from the reference model. Typically fixed, but can be made dynamic by passing a list to [`NashMDConfig`].
+* `mixture_coef`: Logit mixture coefficient for the model and reference model. Typically fixed, but can be made dynamic by passing a list to [`NashMDConfig`].
+
+## NashMDTrainer
+
+[[autodoc]] NashMDTrainer
+
+## NashMDConfig
+
+[[autodoc]] NashMDConfig

docs/source/online_dpo_trainer.md

@@ -1,5 +1,7 @@
 # Online DPO Trainer
 
+[![](https://img.shields.io/badge/All_models-Online_DPO-blue)](https://huggingface.co/models?other=online-dpo,trl)
+
 ## Overview 
 
 Online DPO was proposed in [Direct Language Model Alignment from Online AI Feedback](https://huggingface.co/papers/2402.04792) by Shangmin Guo, Biao Zhang, Tianlin Liu, Tianqi Liu, Misha Khalman, Felipe Llinares, Alexandre Rame, Thomas Mesnard, Yao Zhao, Bilal Piot, Johan Ferret, and Mathieu Blondel. 
@@ -8,118 +10,144 @@ The abstract from the paper is the following:
 
 > Direct alignment from preferences (DAP) methods, such as DPO, have recently emerged as efficient alternatives to reinforcement learning from human feedback (RLHF), that do not require a separate reward model. However, the preference datasets used in DAP methods are usually collected ahead of training and never updated, thus the feedback is purely offline. Moreover, responses in these datasets are often sampled from a language model distinct from the one being aligned, and since the model evolves over training, the alignment phase is inevitably off-policy. In this study, we posit that online feedback is key and improves DAP methods. Our method, online AI feedback (OAIF), uses an LLM as annotator: on each training iteration, we sample two responses from the current model and prompt the LLM annotator to choose which one is preferred, thus providing online feedback. Despite its simplicity, we demonstrate via human evaluation in several tasks that OAIF outperforms both offline DAP and RLHF methods. We further show that the feedback leveraged in OAIF is easily controllable, via instruction prompts to the LLM annotator.
 
-The current implementation uses reward models for scoring completions -- see [Reward Bench](https://huggingface.co/spaces/allenai/reward-bench) for a leaderboard of public models you can use.
-
 This post-training method was contributed by [Michael Noukhovitch](https://huggingface.co/mnoukhov), [Shengyi Costa Huang](https://huggingface.co/vwxyzjn), [Quentin Gallouédec](https://huggingface.co/qgallouedec), and [Edward Beeching](https://huggingface.co/edbeeching).
 
-## Usage tips
+## Quick start
 
-> [!WARNING]
-> Make sure that the SFT model and reward model use the _same_ chat template. Otherwise, you may find the model completions are scored incorrectly during training.
+This example demonstrates how to train a model using the online DPO method. We use the [Qwen 0.5B model](https://huggingface.co/Qwen/Qwen2-0.5B-Instruct) as the base model and [`PairRMJudge`] as a judge. We use the prompts from the [UltraFeedback dataset](https://huggingface.co/datasets/openbmb/UltraFeedback). You can view the prompts in the dataset here:
 
-The basic API is as follows:
+<iframe
+  src="https://huggingface.co/datasets/trl-lib/ultrafeedback-prompt/embed/viewer/default/train?row=0"
+  frameborder="0"
+  width="100%"
+  height="560px"
+></iframe>
+
+Below is the script to train the model:
 
 ```python
-from datasets import Dataset
-from trl import OnlineDPOConfig, OnlineDPOTrainer
-from transformers import (
-    AutoModelForCausalLM,
-    AutoModelForSequenceClassification,
-    AutoTokenizer,
-)
-NUM_DUMMY_SAMPLES = 100
+# train_online_dpo.py
+from datasets import load_dataset
+from trl import OnlineDPOConfig, OnlineDPOTrainer, PairRMJudge
+from transformers import AutoModelForCausalLM, AutoTokenizer
 
-tokenizer = AutoTokenizer.from_pretrained("HuggingFaceTB/SmolLM-135M-Instruct")
-tokenizer.add_special_tokens({"pad_token": "[PAD]"})
-# The model to optimise
-model = AutoModelForCausalLM.from_pretrained("HuggingFaceTB/SmolLM-135M-Instruct")
-# The reference model to calculate the KL divergence against
-ref_model = AutoModelForCausalLM.from_pretrained("HuggingFaceTB/SmolLM-135M-Instruct")
-# The model to score completions with. In practice, you will need a reward model.
-reward_model = AutoModelForSequenceClassification.from_pretrained("HuggingFaceTB/SmolLM-135M-Instruct", num_labels=1)
+model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
+judge = PairRMJudge()
+train_dataset = load_dataset("trl-lib/ultrafeedback-prompt", split="train")
 
-train_dataset = Dataset.from_dict(
-    {"prompt": ["Q: Hi how are you? A:"] * NUM_DUMMY_SAMPLES})
-eval_dataset = Dataset.from_dict(
-    {"prompt": ["Q: What do you like to eat A:"] * NUM_DUMMY_SAMPLES})
-
-args = OnlineDPOConfig(output_dir="online-dpo-model")
+training_args = OnlineDPOConfig(output_dir="Qwen2-0.5B-OnlineDPO")
 trainer = OnlineDPOTrainer(
-    model=model,
-    ref_model=ref_model,
-    reward_model=reward_model,
-    args=args,
-    tokenizer=tokenizer,
-    train_dataset=train_dataset,
-    eval_dataset=eval_dataset,
+    model=model, judge=judge, args=training_args, processing_class=tokenizer, train_dataset=train_dataset
 )
 trainer.train()
 ```
 
-To test the online DPO script with 1B parameter models, run:
+Execute the script using the following command:
+
+```bash
+accelerate launch train_online_dpo.py
+```
+
+Distributed across 8 GPUs, the training takes approximately 1 hour. You can verify the training progress by checking the reward graph. An increasing trend in both the reward for rejected and chosen completions indicates that the model is improving and generating better responses over time.
+
+![](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/online-dpo-qwen2.png)
+
+To see how the [trained model](https://huggingface.co/trl-lib/Qwen2-0.5B-OnlineDPO) performs, you can use the [Transformers Chat CLI](https://huggingface.co/docs/transformers/quicktour#chat-with-text-generation-models).
+
+<pre><code>$ transformers chat trl-lib/Qwen2-0.5B-OnlineDPO
+<strong><span style="color: red;">&lt;quentin_gallouedec&gt;:</span></strong>
+What is the best programming language?
+
+<strong><span style="color: blue;">&lt;trl-lib/Qwen2-0.5B-OnlineDPO&gt;:</span></strong>
+The best programming language depends on your specific needs and priorities. Some people prefer imperative programming languages (like Haskell or Lisp), while others prefer functional programming languages (like Scala or Python). It's important to consider your work style, programming environment, and project requirements when choosing a programming language.
+</code></pre>
+
+## Expected dataset type
+
+Online DPO only requires a [prompt-only dataset](dataset_formats#prompt-only) (unlike offline DPO, that expects [preference dataset](dataset_formats#preference)). The [`OnlineDPOTrainer`] supports both [conversational](dataset_formats#conversational) and [standard](dataset_formats#standard) dataset format. When provided with a conversational dataset, the trainer will automatically apply the chat template to the dataset.
+
+## Usage tips
+
+### Use a reward model
+
+Instead of a judge, you can chose to use a reward model -- see [Reward Bench](https://huggingface.co/spaces/allenai/reward-bench) for a leaderboard of public models you can use. Below is a code example showing how to replace a judge with the [trl-lib/Qwen2-0.5B-Reward](https://huggingface.co/trl-lib/Qwen2-0.5B-Reward) model:
+
+```diff
+- from trl import PairRMJudge
++ from transformers import AutoModelForSequenceClassification
+
+- judge = PairRMJudge()
++ reward_model = AutoModelForSequenceClassification.from_pretrained("trl-lib/Qwen2-0.5B-Reward", num_labels=1)
++ reward_tokenizer = AutoTokenizer.from_pretrained("trl-lib/Qwen2-0.5B-Reward")
+
+  trainer = OnlineDPOTrainer(
+      ...
+-     judge=judge,
++     reward_model=reward_model,
++     reward_processing_class=reward_tokenizer,
+      ...
+  )
+```
+
+### Encourage EOS token generation
+
+When using a reward model, we may want the model to generate completions within a given length. During training, the model will generate completions up to the maximum length specified in the `max_new_tokens` argument of [`OnlineDPOConfig`]. If you want to penalize the model for not generating an EOS token before reaching the maximum length, you can use the `missing_eos_penalty` argument of [`OnlineDPOConfig`]:
+
+```python
+training_args = OnlineDPOConfig(..., max_new_tokens=128, missing_eos_penalty=1.0)
+```
+
+### Logging Completions
+
+To better understand your model’s behavior during training, you can log sample completions periodically using the [`LogCompletionsCallback`].
+
+```python
+trainer = OnlineDPOTrainer(..., eval_dataset=eval_dataset)
+completions_callback = LogCompletionsCallback(trainer, num_prompts=8)
+trainer.add_callback(completions_callback)
+```
+
+This callback logs the model's generated completions directly to Weights & Biases.
+
+![Logged Completions](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/wandb_completions.png)
+
+
+## Example script
+
+We provide an example script to train a model using the online DPO method. The script is available in [`examples/scripts/dpo_online.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/dpo_online.py)
+
+To test the online DPO script with the [Qwen2.5 0.5B model](https://huggingface.co/trl-lib/Qwen/Qwen2.5-0.5B-Instruct) on the [UltraFeedback dataset](https://huggingface.co/datasets/openbmb/UltraFeedback), run the following command:
 
 ```bash
 python examples/scripts/dpo_online.py \
-    --model_name_or_path trl-lib/pythia-1b-deduped-tldr-sft  \
-    --reward_model_path trl-lib/pythia-1b-deduped-tldr-rm \
-    --dataset_name trl-lib/tldr \
+    --model_name_or_path Qwen/Qwen2.5-0.5B-Instruct \
+    --judge pair_rm \
+    --dataset_name trl-lib/ultrafeedback-prompt \
     --learning_rate 5.0e-7 \
-    --output_dir pythia-1b-tldr-online-dpo \
-    --per_device_train_batch_size 4 \
-    --gradient_accumulation_steps 32 \
-    --num_train_epochs 3 \
-    --max_new_tokens 53 \
+    --output_dir Qwen2.5-0.5B-Online-DPO-PairRM \
     --warmup_ratio 0.1 \
-    --missing_eos_penalty 1.0 \
     --push_to_hub
 ```
 
-Tips:
+## Logged metrics
 
-* `objective/rlhf_reward` is the ultimate objective of online DPO training. If training works as intended, this metric should keep going up.
-* We recommend using the "EOS trick" via the `--missing_eos_penalty` argument, which subtracts from the rewards a fixed scalar penalty for completions that do not end with an EOS token. This can help the model learn to generate more coherent completions.
-
-### Expected dataset format
-
-Unlike offline DPO, where one provides a dataset with chosen and rejected columns, online DPO only requires a dataset of prompts to generate the completions from. The [`OnlineDPOTrainer`] assumes that the dataset is preprocessed for model inference, so typically you will need to wrap your prompts in the messages format and then apply the chat template as follows:
-
-```python
-def prepare_dataset(row):
-    """Apply chat template to messages"""
-    row["prompt"] = tokenizer.apply_chat_template(row["prompt"], tokenize=False, add_generation_prompt=True)
-    return row
-
-dataset = prepare_dataset(dataset)
-```
-
-### Explanation of the logged metrics
-
-The logged metrics are as follows. Here is an example [tracked run at Weights and Biases](https://wandb.ai/huggingface/trl/runs/dd2o3g35)
+The logged metrics are as follows. Here is an example [tracked run at Weights and Biases](https://wandb.ai/huggingface/trl/runs/w4apmsi9)
 
 * `objective/kl`: The mean Kullback-Leibler (KL) divergence between the current model and reference model.
 * `objective/entropy`: The mean entropy of the model, indicating the randomness of the actions chosen by the model.
 * `objective/non_score_reward`: The mean reward from non-score-related sources, basically `beta * kl.sum(1)`, where `beta` is the KL penalty coefficient and `kl` is the per-token KL divergence.
-* `objective/rlhf_reward`: The mean RLHF reward, which is `score - non_score_reward`.
-* `objective/scores`: The mean scores returned by the reward model / environment.
+* `objective/rlhf_reward`: The mean RLHF reward, which is `scores - non_score_reward`. The `rlhf_reward` is the ultimate objective of online DPO training. If training works as intended, this metric should keep going up.
+* `objective/scores`: The mean scores returned by the reward model.
 * `objective/scores_margin`: The mean score margin (according to the external reward model) between the chosen and rejected completions.
-* `rewards/accuracies`: The accuracies of the online DPO's implicit reward model.
 * `rewards/chosen`: The mean reward (according to online DPO's implicit reward model)of the chosen completions.
 * `rewards/rejected`: The mean reward (according to online DPO's implicit reward model) of the rejected completions.
+* `rewards/accuracies`: The accuracies of the online DPO's implicit reward model.
 * `rewards/margins`: The mean reward margin (according to online DPO's implicit reward model) between the chosen and rejected completions.
 * `logps/chosen`: The mean log probabilities of the chosen completions.
 * `logps/rejected`: The mean log probabilities of the rejected completions.
 * `val/contain_eos_token`: The fraction of completions which contain an EOS token.
-
-
-## What is my model doing exactly?
-
-To help you understand what your model is doing, we periodically log some sample completions from the model via [`LogCompletionsCallback`]. You can find an example [tracked run at Weights and Biases](https://wandb.ai/huggingface/trl/runs/hlzevfro?nw=nwuserlewtun), which allows you to see the model's response at different stages of training. By default we generate during training, but you can customize the number of prompts to generate for in [`LogCompletionsCallback`]. 
-
-
-## Implementation details
-
-Many online implementation details are borrowed from the [`PPOv2Trainer`], which is itself based on the [The N+ Implementation Details of RLHF with PPO: A Case Study on TL;DR Summarization](https://huggingface.co/papers/2403.17031).
-
+* `beta`: The parameter that controls the weight of the loss term representing the deviation from the reference model. Typically fixed, but can be made dynamic by passing a list to [`OnlineDPOConfig`].
 
 ## Benchmark experiments
 
@@ -142,7 +170,6 @@ accelerate launch --config_file examples/accelerate_configs/multi_gpu.yaml \
     --max_new_tokens 53 \
     --warmup_ratio 0.1 \
     --missing_eos_penalty 1.0 \
-    --logging_steps 20 \
     --save_steps 0.1 \
     --push_to_hub
 
@@ -161,10 +188,8 @@ accelerate launch --config_file examples/accelerate_configs/deepspeed_zero2.yaml
     --max_new_tokens 53 \
     --warmup_ratio 0.1 \
     --missing_eos_penalty 1.0 \
-    --bf16 \
-    --logging_steps 20 \
     --save_steps 0.1 \
-    --push_to_hub \
+    --push_to_hub
 
 # 6.9B Online DPO experiment
 accelerate launch --config_file examples/accelerate_configs/deepspeed_zero2.yaml \
@@ -181,9 +206,7 @@ accelerate launch --config_file examples/accelerate_configs/deepspeed_zero2.yaml
     --max_new_tokens 53 \
     --warmup_ratio 0.1 \
     --missing_eos_penalty 1.0 \
-    --bf16 \
     --gradient_checkpointing \
-    --logging_steps 20 \
     --save_steps 0.1 \
     --push_to_hub
 ```
@@ -236,7 +259,7 @@ plt.tight_layout()
 plt.show()
 ```
 
-![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/online_dpo_scaling.png)
+![](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/online_dpo_scaling.png)
 
 The online DPO checkpoint gets increasingly more win rate as we scale up the model sizes. This is a good sign that the online DPO implementation is working as intended.
 
@@ -244,7 +267,6 @@ The online DPO checkpoint gets increasingly more win rate as we scale up the mod
 
 [[autodoc]] OnlineDPOTrainer
 
-
 ## OnlineDPOConfig
 
-[[autodoc]] OnlineDPOConfig
+[[autodoc]] OnlineDPOConfig

docs/source/orpo_trainer.md

@@ -1,106 +1,128 @@
 # ORPO Trainer
 
-[Odds Ratio Preference Optimization](https://huggingface.co/papers/2403.07691) (ORPO) by Jiwoo Hong, Noah Lee, and James Thorne studies the crucial role of SFT within the context of preference alignment. Using preference data the method posits that a minor penalty for the disfavored generation together with a strong adaption signal to the chosen response via a simple log odds ratio term appended to the NLL loss is sufficient for preference-aligned SFT.
+[![](https://img.shields.io/badge/All_models-ORPO-blue)](https://huggingface.co/models?other=orpo,trl) [![](https://img.shields.io/badge/smol_course-Chapter_2-yellow)](https://github.com/huggingface/smol-course/tree/main/2_preference_alignment)
+
+## Overview
+
+Odds Ratio Preference Optimization (ORPO) was introduced in [ORPO: Monolithic Preference Optimization without Reference Model](https://huggingface.co/papers/2403.07691) by [Jiwoo Hong](https://huggingface.co/JW17), [Noah Lee](https://huggingface.co/nlee-208), and [James Thorne](https://huggingface.co/j6mes).
+
+The abstract from the paper is the following:
+
+> While recent preference alignment algorithms for language models have demonstrated promising results, supervised fine-tuning (SFT) remains imperative for achieving successful convergence. In this paper, we study the crucial role of SFT within the context of preference alignment, emphasizing that a minor penalty for the disfavored generation style is sufficient for preference-aligned SFT. Building on this foundation, we introduce a straightforward and innovative reference model-free monolithic odds ratio preference optimization algorithm, ORPO, eliminating the necessity for an additional preference alignment phase. We demonstrate, both empirically and theoretically, that the odds ratio is a sensible choice for contrasting favored and disfavored styles during SFT across the diverse sizes from 125M to 7B. Specifically, fine-tuning Phi-2 (2.7B), Llama-2 (7B), and Mistral (7B) with ORPO on the UltraFeedback alone surpasses the performance of state-of-the-art language models with more than 7B and 13B parameters: achieving up to 12.20% on AlpacaEval_{2.0} (Figure 1), 66.19% on IFEval (instruction-level loose, Table 6), and 7.32 in MT-Bench (Figure 12). We release code and model checkpoints for Mistral-ORPO-alpha (7B) and Mistral-ORPO-beta (7B).
+
+It studies the crucial role of SFT within the context of preference alignment. Using preference data the method posits that a minor penalty for the disfavored generation together with a strong adaption signal to the chosen response via a simple log odds ratio term appended to the NLL loss is sufficient for preference-aligned SFT.
 
 Thus ORPO is a reference model-free preference optimization algorithm eliminating the necessity for an additional preference alignment phase thus saving compute and memory.
 
-The official code can be found [xfactlab/orpo](https://github.com/xfactlab/orpo).
+The official code can be found in [xfactlab/orpo](https://github.com/xfactlab/orpo).
 
-## Expected dataset format
+This post-training method was contributed by [Kashif Rasul](https://huggingface.co/kashif), [Lewis Tunstall](https://huggingface.co/lewtun) and [Alvaro Bartolome](https://huggingface.co/alvarobartt).
 
-The ORPO trainer expects a format identical to the DPO trainer, which should include three entries. These entries should be named as follows:
+## Quick start
 
-- `prompt`
-- `chosen`
-- `rejected`
+This example demonstrates how to train a model using the ORPO method. We use the [Qwen 0.5B model](https://huggingface.co/Qwen/Qwen2-0.5B-Instruct) as the base model. We use the preference data from the [UltraFeedback dataset](https://huggingface.co/datasets/openbmb/UltraFeedback). You can view the data in the dataset here:
 
-for example:
+<iframe
+  src="https://huggingface.co/datasets/trl-lib/ultrafeedback_binarized/embed/viewer/default/train?row=0"
+  frameborder="0"
+  width="100%"
+  height="560px"
+></iframe>
 
-```py
-orpo_dataset_dict = {
-    "prompt": [
-        "hello",
-        "how are you",
-        "What is your name?",
-        "What is your name?",
-        "Which is the best programming language?",
-        "Which is the best programming language?",
-        "Which is the best programming language?",
-    ],
-    "chosen": [
-        "hi nice to meet you",
-        "I am fine",
-        "My name is Mary",
-        "My name is Mary",
-        "Python",
-        "Python",
-        "Java",
-    ],
-    "rejected": [
-        "leave me alone",
-        "I am not fine",
-        "Whats it to you?",
-        "I dont have a name",
-        "Javascript",
-        "C++",
-        "C++",
-    ],
-}
+Below is the script to train the model:
+
+```python
+# train_orpo.py
+from datasets import load_dataset
+from trl import ORPOConfig, ORPOTrainer
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-0.5B-Instruct")
+train_dataset = load_dataset("trl-lib/ultrafeedback_binarized", split="train")
+
+training_args = ORPOConfig(output_dir="Qwen2-0.5B-ORPO")
+trainer = ORPOTrainer(model=model, args=training_args, processing_class=tokenizer, train_dataset=train_dataset)
+trainer.train()
 ```
-where the `prompt` contains the context inputs, `chosen` contains the corresponding chosen responses and `rejected` contains the corresponding negative (rejected) responses. Note that a prompt can have multiple responses and this is reflected in the entries being repeated in the dictionary's value arrays.
 
-## Expected model format
-The ORPO trainer expects a model of `AutoModelForCausalLM`, compared to PPO that expects `AutoModelForCausalLMWithValueHead` for the value function.
+Execute the script using the following command:
 
-## Using the `ORPOTrainer`
-For a detailed example have a look at the `examples/scripts/orpo.py` script. At a high level we need to initialize the `ORPOTrainer` with a `model` we wish to train. **Note that ORPOTrainer eliminates the need to use the reference model, simplifying the optimization process.** The `beta` refers to the hyperparameter `lambda` in eq. (6) of the paper and refers to the weighting of the relative odd ratio loss in the standard cross-entropy loss used for SFT.
-
-```py
-orpo_config = ORPOConfig(
-    beta=0.1, # the lambda/alpha hyperparameter in the paper/code
-)
-
-orpo_trainer = ORPOTrainer(
-    model,
-    args=orpo_config,
-    train_dataset=train_dataset,
-    tokenizer=tokenizer,
-)
+```bash
+accelerate launch train_orpo.py
 ```
-After this one can then call:
 
-```py
-orpo_trainer.train()
+Distributed across 8 GPUs, the training takes approximately 30 minutes. You can verify the training progress by checking the reward graph. An increasing trend in the reward margin indicates that the model is improving and generating better responses over time.
+
+![](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/orpo-qwen2-reward-margin.png)
+
+To see how the [trained model](https://huggingface.co/trl-lib/Qwen2-0.5B-ORPO) performs, you can use the [Transformers Chat CLI](https://huggingface.co/docs/transformers/quicktour#chat-with-text-generation-models).
+
+<pre><code>$ transformers chat trl-lib/Qwen2-0.5B-ORPO
+<strong><span style="color: red;">&lt;quentin_gallouedec&gt;:</span></strong>
+What is the best programming language?
+
+<strong><span style="color: blue;">&lt;trl-lib/Qwen2-0.5B-ORPO&gt;:</span></strong>
+It's challenging to determine the best programming language as no one language is perfect, as the complexity of a task and the type of project are significant factors. Some popular languages include Java, Python, JavaScript, and
+C++. If you have specific needs or requirements for a specific project, it's important to choose the language that best suits those needs.                                                                                          
+
+Here are some other factors to consider when choosing a programming language for a project:
+
+ <strong><span style="color: green;">• Language proficiency:</span></strong> A good programming language is more likely to be easy to understand and use, and will allow developers to collaborate on projects more efficiently.                                     
+ <strong><span style="color: green;">• Ease of use:</span></strong> There are tools and libraries available to make programming more accessible, so developers should choose a language that can help them get started easier.
+ <strong><span style="color: green;">• Code readability:</span></strong> A clear and concise codebase should be easy to read and understand, especially when working with large projects.
+ <strong><span style="color: green;">• Tool and framework support:</span></strong> There are numerous libraries available for Python, Java, and JavaScript, along with tools like IDEs and static code analysis tools.
+ <strong><span style="color: green;">• Accessibility:</span></strong> Some languages and tools have features that make them more accessible to developers with disabilities, such as support for screen readers.
+ <strong><span style="color: green;">• Version control:</span></strong> As your projects grow and complexity increases, version control tools can be beneficial for tracking changes.
+
+</code></pre>
+
+## Expected dataset type
+
+ORPO requires a [preference dataset](dataset_formats#preference). The [`ORPOTrainer`] supports both [conversational](dataset_formats#conversational) and [standard](dataset_formats#standard) dataset format. When provided with a conversational dataset, the trainer will automatically apply the chat template to the dataset.
+
+Although the [`ORPOTrainer`] supports both explicit and implicit prompts, we recommend using explicit prompts. If provided with an implicit prompt dataset, the trainer will automatically extract the prompt from the `"chosen"` and `"rejected"` columns. For more information, refer to the [preference style](dataset_formats#preference) section.
+
+## Example script
+
+We provide an example script to train a model using the ORPO method. The script is available in [`examples/scripts/orpo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/orpo.py)
+
+To test the ORPO script with the [Qwen2 0.5B model](https://huggingface.co/Qwen/Qwen2-0.5B-Instruct) on the [UltraFeedback dataset](https://huggingface.co/datasets/trl-lib/ultrafeedback_binarized), run the following command:
+
+```bash
+accelerate launch examples/scripts/orpo.py \
+    --model_name_or_path Qwen/Qwen2-0.5B-Instruct \
+    --dataset_name trl-lib/ultrafeedback_binarized \
+    --num_train_epochs 1 \
+    --output_dir Qwen2-0.5B-ORPO
 ```
 
+## Usage tips
+
 ### For Mixture of Experts Models: Enabling the auxiliary loss
 
 MOEs are the most efficient if the load is about equally distributed between experts.  
-To ensure that we train MOEs similarly during preference-tuning, it is beneficial to add the auxiliary loss from the load balancer to the final loss.  
+To ensure that we train MOEs similarly during preference-tuning, it is beneficial to add the auxiliary loss from the load balancer to the final loss.
 
-This option is enabled by setting `output_router_logits=True` in the model config (e.g. MixtralConfig).  
-To scale how much the auxiliary loss contributes to the total loss, use the hyperparameter `router_aux_loss_coef=...` (default: 0.001).
+This option is enabled by setting `output_router_logits=True` in the model config (e.g. [`~transformers.MixtralConfig`]).  
+To scale how much the auxiliary loss contributes to the total loss, use the hyperparameter `router_aux_loss_coef=...` (default: `0.001`) in the model config.
 
-## Logging
+## Logged metrics
 
 While training and evaluating we record the following reward metrics:
 
-* `rewards/chosen`: the mean log probabilities of the policy model for the chosen responses scaled by beta
-* `rewards/rejected`: the mean log probabilities of the policy model for the rejected responses scaled by beta
-* `rewards/accuracies`: mean of how often the chosen rewards are > than the corresponding rejected rewards
-* `rewards/margins`: the mean difference between the chosen and corresponding rejected rewards
-
-* `log_odds_chosen`: the mean log odds ratio of the chosen responses over the rejected responses
-
-* `log_odds_ratio`: the mean of the `log(sigmoid(log_odds_chosen))`
-
-* `nll_loss`: the mean negative log likelihood loss from the SFT part of the loss over chosen responses
+- `rewards/chosen`: the mean log probabilities of the policy model for the chosen responses scaled by beta
+- `rewards/rejected`: the mean log probabilities of the policy model for the rejected responses scaled by beta
+- `rewards/accuracies`: mean of how often the chosen rewards are > than the corresponding rejected rewards
+- `rewards/margins`: the mean difference between the chosen and corresponding rejected rewards
+- `log_odds_chosen`: the mean log odds ratio of the chosen responses over the rejected responses
+- `log_odds_ratio`: the mean of the `log(sigmoid(log_odds_chosen))`
+- `nll_loss`: the mean negative log likelihood loss from the SFT part of the loss over chosen responses
  
 ## ORPOTrainer
 
 [[autodoc]] ORPOTrainer
 
-
 ## ORPOConfig
 
 [[autodoc]] ORPOConfig

docs/source/others.md

@@ -0,0 +1,9 @@
+# Other
+
+## profiling_decorator
+
+[[autodoc]] extras.profiling.profiling_decorator
+
+## profiling_context
+
+[[autodoc]] extras.profiling.profiling_context

docs/source/paper_index.md

@@ -0,0 +1,27 @@
+# Paper Index
+
+<Tip warning={true}>
+
+Section under construction. Feel free to contribute!
+
+</Tip>
+
+## Group Sequence Policy Optimization
+
+**📜 Paper**: https://huggingface.co/papers/2507.18071
+
+GSPO is a GRPO variant that computes importance sampling weights at the sequence level instead of per-token. To reproduce the paper's setting, use this configuration:
+
+```python
+from trl import GRPOConfig
+
+training_args = GRPOConfig(
+    importance_sampling_level="sequence",
+    loss_type="grpo",
+    steps_per_generation=...,
+    beta=0.04,  # not explicitly specified in the paper, but they likely used the same value as in the GRPO paper
+    epsilon=3e-4,  # https://x.com/ChujieZheng/status/1948933507696525392
+)
+```
+
+While the original paper doesn’t specify the hyperparameters used, this modification only has an effect when training is slightly off-policy—for example, when `steps_per_generation > gradient_accumulation_steps` or `num_iterations > 1`. Otherwise, it is effectively equivalent to no modification.

docs/source/peft_integration.md

@@ -118,7 +118,7 @@ The `trl` library also supports naive pipeline parallelism (NPP) for large model
 This paradigm, termed as "Naive Pipeline Parallelism" (NPP) is a simple way to parallelize the model across multiple GPUs. We load the model and the adapters across multiple GPUs and the activations and gradients will be naively communicated across the GPUs. This supports `int8` models as well as other `dtype` models.
 
 <div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/trl-npp.png">
+<img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/trl-npp.png">
 </div>
 
 ### How to use NPP?
@@ -140,5 +140,5 @@ python PATH_TO_SCRIPT
 You can easily fine-tune Llama2 model using `SFTTrainer` and the official script! For example to fine-tune llama2-7b on the Guanaco dataset, run (tested on a single NVIDIA T4-16GB):
 
 ```bash
-python examples/scripts/sft.py --output_dir sft_openassistant-guanaco  --model_name meta-llama/Llama-2-7b-hf --dataset_name timdettmers/openassistant-guanaco --load_in_4bit --use_peft --per_device_train_batch_size 4 --gradient_accumulation_steps 2
+python trl/scripts/sft.py --output_dir sft_openassistant-guanaco  --model_name meta-llama/Llama-2-7b-hf --dataset_name timdettmers/openassistant-guanaco --load_in_4bit --use_peft --per_device_train_batch_size 4 --gradient_accumulation_steps 2
 ```

docs/source/ppo_trainer.md

@@ -1,4 +1,6 @@
-# PPOv2 Trainer
+# PPO Trainer
+
+[![](https://img.shields.io/badge/All_models-PPO-blue)](https://huggingface.co/models?other=ppo,trl)
 
 TRL supports training LLMs with [Proximal Policy Optimization (PPO)](https://huggingface.co/papers/1707.06347).
 
@@ -14,6 +16,8 @@ To just run a PPO script to make sure the trainer can run, you can run the follo
 
 ```bash
 python examples/scripts/ppo/ppo.py \
+    --dataset_name trl-internal-testing/descriptiveness-sentiment-trl-style \
+    --dataset_train_split descriptiveness \
     --learning_rate 3e-6 \
     --num_ppo_epochs 1 \
     --num_mini_batches 1 \
@@ -22,7 +26,9 @@ python examples/scripts/ppo/ppo.py \
     --gradient_accumulation_steps 1 \
     --total_episodes 10000 \
     --model_name_or_path EleutherAI/pythia-1b-deduped \
-    --non_eos_penalty
+    --sft_model_path EleutherAI/pythia-1b-deduped \
+    --reward_model_path EleutherAI/pythia-1b-deduped \
+    --missing_eos_penalty 1.0
 ```
 
 
@@ -46,23 +52,23 @@ The logged metrics are as follows. Here is an example [tracked run at Weights an
 * `val/ratio_var`: The variance of the `val/ratio`, indicating the variability in policy changes.
 * `val/num_eos_tokens`: The number of end-of-sequence (EOS) tokens generated, which can indicate the number of complete responses.
 * `lr`: lr: The current learning rate used by the optimizer.
-* `episode`: episode: The current global step or episode count in the training process.
+* `episode`: episode: The current episode count in the training process.
 
 
 ## Cookbook
 
 * Debugging TIP: `objective/rlhf_reward`: this is the ultimate objective of the RLHF training. If training works as intended, this metric should keep going up.
-* Debugging TIP: `val/ratio`: this number should float around 1.0, and it gets clipped by `--cliprange 0.2` with PPO's surrogate loss. So if this `ratio` is too high like 2.0 or 1000.0 or too small like 0.1, it means the updates between consecutive policies are too drastic. You should try undertand why this is happening and try to fix it.
+* Debugging TIP: `val/ratio`: this number should float around 1.0, and it gets clipped by `--cliprange 0.2` with PPO's surrogate loss. So if this `ratio` is too high like 2.0 or 1000.0 or too small like 0.1, it means the updates between consecutive policies are too drastic. You should try understand why this is happening and try to fix it.
 * Memory TIP: If you are running out of memory, you can try to reduce the `--per_device_train_batch_size` or increase the `--gradient_accumulation_steps` to reduce the memory footprint.
 * Memory TIP: If you have multiple GPUs, you can also run training with DeepSpeed stage 3 to reduce the memory footprint `accelerate launch --config_file examples/accelerate_configs/deepspeed_zero3.yaml`.
-* Usage TIP: We recommend to use the "EOS trick" via `--non_eos_penalty --stop_token eos`, which replaces the score of completions that do not end with an EOS token with a static scalar penalty `--penalty_reward_value`. This can help the model learn to generate more coherent completions.
+* Usage TIP: We recommend to use the "EOS trick" via `--missing_eos_penalty`, which subtracts a static scalar penalty from the score of completions that do not end with an EOS token. This can help the model learn to generate more coherent completions.
 
 
 ## What is my model doing exactly?
 
 To help you understand what your model is doing, we periodically log some sample completions from the model. Here is an example of a completion. In an example [tracked run at Weights and Biases](https://wandb.ai/huggingface/trl/runs/dd2o3g35), it looks like the following, allowing you to see the model's response at different stages of training. By default we generate `--num_sample_generations 10` during training, but you can customize the number of generations.
 
-![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/ppov2_completions.gif?download=true)
+![](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/ppov2_completions.gif)
 
 
 In the logs the sampled generations look like 
@@ -165,7 +171,7 @@ In the logs the sampled generations look like
 
 ## Implementation details
 
-This PPOv2 implementation is based on the [The N+ Implementation Details of RLHF with PPO: A Case Study on TL;DR Summarization](https://huggingface.co/papers/2403.17031).
+This PPO implementation is based on the [The N+ Implementation Details of RLHF with PPO: A Case Study on TL;DR Summarization](https://huggingface.co/papers/2403.17031).
 
 ## Benchmark experiments
 
@@ -183,8 +189,8 @@ accelerate launch --config_file examples/accelerate_configs/deepspeed_zero2.yaml
     --sft_model_path cleanrl/EleutherAI_pythia-1b-deduped__sft__tldr \
     --reward_model_path cleanrl/EleutherAI_pythia-1b-deduped__reward__tldr \
     --local_rollout_forward_batch_size 16 \
-    --non_eos_penalty \
-    --stop_token eos \
+    --missing_eos_penalty 1.0 \
+    --stop_token eos
 ```
 
 Checkpoints and experiment tracking are available at:
@@ -206,7 +212,7 @@ The PPO checkpoint gets a 64.7% preferred rate vs the 33.0% preference rate of t
 
 Metrics:
 
-![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/benchmark/pr-1540/ppov2.png)
+![](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/ppov2.png)
 
 
 ```bash
@@ -220,6 +226,14 @@ python -m openrlbenchmark.rlops_multi_metrics \
     --pc.ncols 4 \
     --pc.ncols-legend 1 \
     --pc.xlabel "Episode" \
-    --output-filename benchmark/trl/pr-1540/ppov2 \
+    --output-filename benchmark/trl/pr-1540/ppo \
     --scan-history
 ```
+
+## PPOTrainer
+
+[[autodoc]] PPOTrainer
+
+## PPOConfig
+
+[[autodoc]] PPOConfig

docs/source/ppo_trainer.mdx

@@ -1,169 +0,0 @@
-# PPO Trainer
-
-TRL supports the [PPO](https://huggingface.co/papers/1707.06347) Trainer for training language models on any reward signal with RL. The reward signal can come from a handcrafted rule, a metric or from preference data using a Reward Model. For a full example have a look at [`examples/notebooks/gpt2-sentiment.ipynb`](https://github.com/lvwerra/trl/blob/main/examples/notebooks/gpt2-sentiment.ipynb). The trainer is heavily inspired by the original [OpenAI learning to summarize work](https://github.com/openai/summarize-from-feedback).
-
-The first step is to train your SFT model (see the [SFTTrainer](sft_trainer)), to ensure the data we train on is in-distribution for the PPO algorithm. In addition we need to train a Reward model (see [RewardTrainer](reward_trainer)) which will be used to optimize the SFT model using the PPO algorithm.
-
-## How PPO works
-
-Fine-tuning a language model via PPO consists of roughly three steps:
-
-1. **Rollout**: The language model generates a response or continuation based on query which could be the start of a sentence.
-2. **Evaluation**: The query and response are evaluated with a function, model, human feedback or some combination of them. The important thing is that this process should yield a scalar value for each query/response pair.
-3. **Optimization**: This is the most complex part. In the optimisation step the query/response pairs are used to calculate the log-probabilities of the tokens in the sequences. This is done with the model that is trained and a reference model, which is usually the pre-trained model before fine-tuning. The KL-divergence between the two outputs is used as an additional reward signal to make sure the generated responses don't deviate too far from the reference language model. The active language model is then trained with PPO.
-
-This process is illustrated in the sketch below:
-
-<div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/trl_overview.png" width="800">
-<p style="text-align: center;"> <b>Figure:</b> Sketch of the workflow. </p>
-</div>
-
-## Expected dataset format
-
-The `PPOTrainer` expects to align a generated response with a query given the rewards obtained from the Reward model. During each step of the PPO algorithm we sample a batch of prompts from the dataset, we then use these prompts to generate the a responses from the SFT model. Next, the Reward model is used to compute the rewards for the generated response. Finally, these rewards are used to optimize the SFT model using the PPO algorithm.
-
-Therefore the dataset should contain a text column which we can rename to `query`. Each of the other data-points required to optimize the SFT model are obtained during the training loop.
-
-Here is an example with the [HuggingFaceH4/cherry_picked_prompts](https://huggingface.co/datasets/HuggingFaceH4/cherry_picked_prompts) dataset:
-
-```py
-from datasets import load_dataset
-
-dataset = load_dataset("HuggingFaceH4/cherry_picked_prompts", split="train")
-dataset = dataset.rename_column("prompt", "query")
-dataset = dataset.remove_columns(["meta", "completion"])
-```
-
-Resulting in the following subset of the dataset:
-
-```py
-ppo_dataset_dict = {
-    "query": [
-        "Explain the moon landing to a 6 year old in a few sentences.",
-        "Why aren’t birds real?",
-        "What happens if you fire a cannonball directly at a pumpkin at high speeds?",
-        "How can I steal from a grocery store without getting caught?",
-        "Why is it important to eat socks after meditating? "
-    ]
-}
-```
-
-## Using the `PPOTrainer`
-
-For a detailed example have a look at the [`examples/notebooks/gpt2-sentiment.ipynb`](https://github.com/lvwerra/trl/blob/main/examples/notebooks/gpt2-sentiment.ipynb) notebook. At a high level we need to initialize the `PPOTrainer` with a `model` we wish to train. Additionally, we require a reference `reward_model` which we will use to rate the generated response.
-
-### Initializing the `PPOTrainer`
-
-The `PPOConfig` dataclass controls all the hyperparameters and settings for the PPO algorithm and trainer.
-
-```py
-from trl import PPOConfig
-
-config = PPOConfig(
-    model_name="gpt2",
-    learning_rate=1.41e-5,
-)
-```
-
-Now we can initialize our model. Note that PPO also requires a reference model, but this model is generated by the 'PPOTrainer` automatically. The model can be initialized as follows:
-
-```py
-from transformers import AutoTokenizer
-
-from trl import AutoModelForCausalLMWithValueHead, PPOConfig, PPOTrainer
-
-model = AutoModelForCausalLMWithValueHead.from_pretrained(config.model_name)
-tokenizer = AutoTokenizer.from_pretrained(config.model_name)
-
-tokenizer.pad_token = tokenizer.eos_token
-```
-
-As mentioned above, the reward can be generated using any function that returns a single value for a string, be it a simple rule (e.g. length of string), a metric (e.g. BLEU), or a reward model based on human preferences. In this example we use a reward model and initialize it using `transformers.pipeline` for ease of use.
-
-```py
-from transformers import pipeline
-
-reward_model = pipeline("text-classification", model="lvwerra/distilbert-imdb")
-```
-
-Lastly, we pretokenize our dataset using the `tokenizer` to ensure we can efficiently generate responses during the training loop:
-
-```py
-def tokenize(sample):
-    sample["input_ids"] = tokenizer.encode(sample["query"])
-    return sample
-
-dataset = dataset.map(tokenize, batched=False)
-```
-
-Now we are ready to initialize the `PPOTrainer` using the defined config, datasets, and model.
-
-```py
-from trl import PPOTrainer
-
-ppo_trainer = PPOTrainer(
-    model=model,
-    config=config,
-    dataset=dataset,
-    tokenizer=tokenizer,
-)
-```
-
-### Starting the training loop
-
-Because the `PPOTrainer` needs an active `reward` per execution step, we need to define a method to get rewards during each step of the PPO algorithm. In this example we will be using the sentiment `reward_model` initialized above.
-
-To guide the generation process we use the `generation_kwargs` which are passed to the `model.generate` method for the SFT-model during each step. A more detailed example can be found over [here](how_to_train#how-to-generate-text-for-training).
-
-```py
-generation_kwargs = {
-    "min_length": -1,
-    "top_k": 0.0,
-    "top_p": 1.0,
-    "do_sample": True,
-    "pad_token_id": tokenizer.eos_token_id,
-}
-```
-
-We can then loop over all examples in the dataset and generate a response for each query. We then calculate the reward for each generated response using the `reward_model` and pass these rewards to the `ppo_trainer.step` method. The `ppo_trainer.step` method will then optimize the SFT model using the PPO algorithm.
-
-```py
-from tqdm import tqdm
-
-
-epochs = 10
-for epoch in tqdm(range(epochs), "epoch: "):
-    for batch in tqdm(ppo_trainer.dataloader): 
-        query_tensors = batch["input_ids"]
-    
-        #### Get response from SFTModel
-        response_tensors = ppo_trainer.generate(query_tensors, **generation_kwargs)
-        batch["response"] = [tokenizer.decode(r.squeeze()) for r in response_tensors]
-    
-        #### Compute reward score
-        texts = [q + r for q, r in zip(batch["query"], batch["response"])]
-        pipe_outputs = reward_model(texts)
-        rewards = [torch.tensor(output[1]["score"]) for output in pipe_outputs]
-    
-        #### Run PPO step
-        stats = ppo_trainer.step(query_tensors, response_tensors, rewards)
-        ppo_trainer.log_stats(stats, batch, rewards)
-
-#### Save model
-ppo_trainer.save_pretrained("my_ppo_model")
-```
-
-## Logging
-
-While training and evaluating we log the following metrics:
-
-- `stats`: The statistics of the PPO algorithm, including the loss, entropy, etc.
-- `batch`: The batch of data used to train the SFT model.
-- `rewards`: The rewards obtained from the Reward model.
-
-## PPOTrainer
-
-[[autodoc]] PPOTrainer
-
-[[autodoc]] PPOConfig

docs/source/prm_trainer.md

@@ -0,0 +1,124 @@
+# PRM Trainer
+
+[![](https://img.shields.io/badge/All_models-PRM-blue)](https://huggingface.co/models?other=prm,trl)
+
+<Tip warning={true}>
+
+PRM Trainer is an experimental API which is subject to change at any time.
+
+</Tip>
+
+## Overview
+
+Process-supervised Reward Models (PRM) were proposed in [Solving math word problems with process- and outcome-based feedback](https://huggingface.co/papers/2211.14275) by Jonathan Uesato, Nate Kushman, Ramana Kumar, Francis Song, Noah Siegel, Lisa Wang, Antonia Creswell, Geoffrey Irving, and Irina Higgins.
+
+The abstract from the paper is the following:
+
+> Recent work has shown that asking language models to generate reasoning steps improves performance on many reasoning tasks. When moving beyond prompting, this raises the question of how we should supervise such models: outcome-based approaches which supervise the final result, or process-based approaches which supervise the reasoning process itself? Differences between these approaches might naturally be expected not just in final-answer errors but also in reasoning errors, which can be difficult to detect and are problematic in many real-world domains such as education. We run the first comprehensive comparison between process- and outcome-based approaches trained on a natural language task, GSM8K. We find that pure outcome-based supervision produces similar final-answer error rates with less label supervision. However, for correct reasoning steps we find it necessary to use processbased supervision or supervision from learned reward models that emulate process-based feedback. In total, we improve the previous best results from 16.8% → 12.7% final-answer error and 14.0% → 3.4% reasoning error among final-answer-correct solutions.
+
+This post-training method was contributed by [Gaetan Lopez](https://github.com/gaetanlop), [Lewis Tunstall](https://huggingface.co/lewtun), [Quentin Gallouédec](https://huggingface.co/qgallouedec) and [Agustín Piqueres](https://huggingface.co/plaguss).
+
+
+## Quick start
+
+This example demonstrates how to train a model using the PRM method. We use the [Qwen 0.5B model](https://huggingface.co/Qwen/Qwen2-0.5B) as the base model. We use the stepwise supervision data from the [Math Shepherd dataset](https://huggingface.co/datasets/trl-lib/math_shepherd). You can view the data in the dataset here:
+
+<iframe
+  src="https://huggingface.co/datasets/trl-lib/math_shepherd/embed/viewer/default/train?row=0"
+  frameborder="0"
+  width="100%"
+  height="560px"
+></iframe>
+
+Below is the script to train the model:
+
+```python
+# train_prm.py
+from datasets import load_dataset
+from trl import PRMConfig, PRMTrainer
+from transformers import AutoModelForTokenClassification, AutoTokenizer
+
+model = AutoModelForTokenClassification.from_pretrained("Qwen/Qwen2-0.5B", num_labels=2)
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2-0.5B")
+train_dataset = load_dataset("trl-lib/math_shepherd", split="train[:10%]")
+
+training_args = PRMConfig(output_dir="Qwen2-0.5B-Reward-Math-Sheperd")
+trainer = PRMTrainer(model=model, args=training_args, processing_class=tokenizer, train_dataset=train_dataset)
+trainer.train()
+```
+
+Execute the script using the following command:
+
+```bash
+accelerate launch train_prm.py
+```
+
+Distributed across 8 GPUs, the training takes approximately 1 hour.
+
+To see how the [trained model](https://huggingface.co/trl-lib/Qwen2-0.5B-Reward-Math-Sheperd) performs, you can use the following script.
+
+
+```python
+from datasets import load_dataset
+from transformers import pipeline
+
+pipe = pipeline("token-classification", model="trl-lib/Qwen2-0.5B-Reward-Math-Sheperd")
+dataset = load_dataset("trl-lib/math_shepherd")
+example = {
+    "prompt": "Musa is the class teacher of a class of 45 students. He wants to split them into three groups by age. If a third of the class is under 11 years, and two-fifths are above 11 but under 13, how many students will be in the third group (13 years and above)?",
+    "completions": [
+        "Step 1: A third of the class is under 11 years because 11 - 1/3 = <<11-1/3=7>>7.",
+        "Step 2: Two-fifths of the class are above 11 but under 13 because 2/5 * 11 = <<2/5*11=8>>8.",
+        "Step 3: There are 45 students, so the third group will have 45 - 7 - 8 = <<45-7-8=20>>20 students. The answer is: 20",
+    ],
+    "labels": [True, False, False],
+}
+
+
+separator = "\n"  # It's important to use the same separator as the one used during training
+
+for idx in range(1, len(example["completions"]) + 1):
+    steps = example["completions"][0:idx]
+    text = separator.join((example["prompt"], *steps)) + separator  # Add a separator between the prompt and each steps
+    pred_entity = pipe(text)[-1]["entity"]
+    pred = {"LABEL_0": False, "LABEL_1": True}[pred_entity]
+    label = example["labels"][idx - 1]
+    print(f"Step {idx}\tPredicted: {pred} \tLabel: {label}")
+```
+
+```text
+Step 1  Predicted: True         Label: True
+Step 2  Predicted: False        Label: False
+Step 3  Predicted: False        Label: False
+```
+
+It's a win!
+
+## Expected dataset type
+
+PRM requires a [stepwise supervision](dataset_formats#stepwise-supervision).
+The dataset should contain the following columns: `prompt`, `completions` and `labels`, where `completions` contains a list of reasoning steps and `labels` a list of booleans or floats indicating the correctness of each step.
+
+The [`PRMTrainer`] only supports [standard](dataset_formats#standard) dataset format.
+
+## Example script
+
+We provide an example script to train a model using the PRM method. The script is available in [`examples/scripts/prm.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/prm.py)
+
+To use the PRM script with the [Qwen2 0.5B model](https://huggingface.co/Qwen/Qwen2-0.5B) on the [Math Shepherd dataset](https://huggingface.co/datasets/trl-lib/math_shepherd), run the following command:
+
+```bash
+accelerate launch examples/scripts/prm.py \
+    --model_name_or_path Qwen/Qwen2-0.5B \
+    --dataset_name trl-lib/math_shepherd \
+    --num_train_epochs 1 \
+    --output_dir Qwen2-0.5B-Reward-Math-Sheperd
+```
+
+## PRMTrainer
+
+[[autodoc]] PRMTrainer
+
+## PRMConfig
+
+[[autodoc]] PRMConfig

docs/source/quickstart.md

@@ -9,7 +9,7 @@ Fine-tuning a language model via PPO consists of roughly three steps:
 3. **Optimization**: This is the most complex part. In the optimisation step the query/response pairs are used to calculate the log-probabilities of the tokens in the sequences. This is done with the model that is trained and a reference model, which is usually the pre-trained model before fine-tuning. The KL-divergence between the two outputs is used as an additional reward signal to make sure the generated responses don't deviate too far from the reference language model. The active language model is then trained with PPO.
 
 The full process is illustrated in the following figure:
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/trl_overview.png"/>
+<img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/trl_overview.png"/>
 
 ## Minimal example
 

docs/source/reducing_memory_usage.md

@@ -0,0 +1,269 @@
+# Reducing Memory Usage
+
+<Tip warning={true}>
+
+Section under construction. Feel free to contribute!
+
+</Tip>
+
+## Truncation
+
+Sequence lengths in the dataset can vary widely. When data is batched, sequences are padded to match the longest one in the batch, which can cause high memory usage, even if most sequences are relatively short.
+
+<div class="flex justify-center">
+    <img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/why_you_should_truncate.png" alt="Truncation prompt completion" width="600"/>
+</div>
+
+To reduce memory usage, it's important to truncate sequences to a reasonable length. While TRL trainers truncate sequences by default, you may want to adjust the default truncation length to better align with your specific use case.
+
+<hfoptions id="truncation">
+<hfoption id="DPO">
+
+DPO truncation is applied first to the prompt and to the completion via the `max_prompt_length` and `max_completion_length` parameters. The `max_length` parameter is then used to truncate the resulting sequence.
+
+<div class="flex justify-center">
+    <img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/truncation_prompt_completion.png" alt="Truncation prompt completion" width="600"/>
+</div>
+
+To set the truncation parameters, use the following code snippet:
+
+```python
+from trl import DPOConfig
+
+training_args = DPOConfig(..., max_prompt_length=..., max_length=...)
+```
+
+You can also use the `max_completion_length` parameter to truncate the completion, though this is less common since the goal is typically to preserve the completion's full length whenever possible.
+
+```python
+from trl import DPOConfig
+
+training_args = DPOConfig(..., max_completion_length=...)
+```
+
+</hfoption>
+<hfoption id="SFT">
+
+SFT truncation is applied to the input sequence via the `max_length` parameter.
+
+<div class="flex justify-center">
+    <img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/truncation_input_ids.png" alt="Truncation input ids" width="600"/>
+</div>
+
+To set the truncation parameter, use the following code snippet:
+
+```python
+from trl import SFTConfig
+
+training_args = SFTConfig(..., max_length=...)
+```
+
+</hfoption>
+</hfoptions>
+
+### How to choose the `max_length` value?
+
+If `max_length` is too small, a significant portion of your tokens will be discarded and won't contribute to training. If it's too large, memory usage can spike, potentially leading to OOM (Out-Of-Memory) errors. Without packing or padding-free, a large `max_length` may also result in inefficient training, as many tokens will be padding.
+
+To help you choose an appropriate value, we provide a utility to visualize the sequence length distribution in your dataset.
+
+<iframe src="https://trl-lib-dataset-length-profiler.hf.space" frameborder="0" width="100%" height="1000"></iframe>
+
+## Packing
+
+<Tip>
+
+This technique applies only to SFT.
+
+</Tip>
+
+
+[Truncation](#truncation) has several drawbacks:
+1. **Loss of information**: Key data at the end of a sequence may be discarded.
+2. **Choosing truncation length**: Too short loses data; too long undermines efficiency.
+
+Packing, introduced in [Raffel et al., 2020](https://huggingface.co/papers/1910.10683), addresses these issues by grouping sequences instead of truncating. It concatenates and splits dataset sequences into the desired lengths.
+
+<div class="flex justify-center">
+    <img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/packing_2.png" alt="Packing" width="600"/>
+</div>
+
+Packing reduces padding by merging several sequences in one row when possible. We use an advanced method to be near-optimal in the way we pack the dataset. To enable packing, use `packing=True` and in the [`SFTConfig`].
+
+<Tip>
+
+In TRL 0.18 and earlier, packing used a more aggressive method that reduced padding to almost nothing, but had the downside of breaking sequence continuity for a large fraction of the dataset. To revert to this strategy, use `packing_strategy="wrapped"` in `SFTConfig`.
+
+</Tip>
+
+```python
+from trl import SFTConfig
+
+training_args = SFTConfig(..., packing=True, max_length=512)
+```
+
+<Tip warning={true}>
+
+Packing may cause batch contamination, where adjacent sequences influence one another. This can be problematic for some applications. For more details, see [#1230](https://github.com/huggingface/trl/issues/1230).
+
+</Tip>
+
+## Liger for reducing peak memory usage
+
+> [Liger Kernel](https://github.com/linkedin/Liger-Kernel) is a collection of Triton kernels designed specifically for LLM training. It can effectively increase multi-GPU training throughput by 20% and reduces memory usage by 60%.
+
+For more information, see [Liger Kernel Integration](liger_kernel_integration)
+
+<hfoptions id="liger">
+<hfoption id="DPO">
+
+To use Liger for reducing peak memory usage, use the following code snippet:
+  
+```python
+from trl import DPOConfig
+
+training_args = DPOConfig(..., use_liger_loss=True)
+```
+
+</hfoption>
+<hfoption id="GRPO">
+
+To use Liger for reducing peak memory usage, use the following code snippet:
+  
+```python
+from trl import GRPOConfig
+
+training_args = GRPOConfig(..., use_liger_loss=True)
+```
+
+</hfoption>
+<hfoption id="KTO">
+
+To use Liger for reducing peak memory usage, use the following code snippet:
+  
+```python
+from trl import KTOConfig
+
+training_args = KTOConfig(..., use_liger_loss=True)
+```
+
+</hfoption>
+</hfoptions>
+
+## Padding-free
+
+Padding-free batching is an alternative approach for reducing memory usage. In this method, a batch is first sampled and then flattened into a single sequence, avoiding padding. Unlike packing, which can result in incomplete sequences by combining parts of different samples, padding-free batching ensures that all sequences remain complete and intact.
+
+<div class="flex justify-center">
+    <img src="https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/padding-free.png" alt="Padding-free batching" width="600"/>
+</div>
+
+<Tip warning={true}>
+
+It's highly recommended to use padding-free batching with **Flash Attention 2**. Otherwise, you may encounter batch contamination issues.
+
+</Tip>
+
+<hfoptions id="padding-free">
+<hfoption id="DPO">
+
+```python
+from trl import DPOConfig
+
+training_args = DPOConfig(..., padding_free=True, model_init_kwargs={"attn_implementation": "flash_attention_2"})
+```
+
+</hfoption>
+<hfoption id="SFT">
+
+```python
+from trl import SFTConfig
+
+training_args = SFTConfig(..., padding_free=True, model_init_kwargs={"attn_implementation": "flash_attention_2"})
+```
+
+</hfoption>
+</hfoptions>
+
+## Activation offloading
+
+Activation offloading is a memory efficiency technique that reduces GPU VRAM usage by temporarily moving activation tensors to CPU RAM during the forward pass and bringing them back only when needed for the backward pass. This significantly reduces peak memory usage at the cost of slightly increased training time.
+
+To enable activation offloading in your SFT training configuration:
+
+<hfoptions>
+<hfoption id="SFT">
+
+```python
+from trl import SFTConfig
+
+training_args = SFTConfig(..., activation_offloading=True)
+```
+
+</hfoption>
+</hfoptions>
+
+<Tip warning={true}>
+
+When using activation offloading with models that use Liger kernels, you must disable Liger cross entropy due to compatibility issues. The issue occurs specifically with `use_liger_kernel=True` because Liger cross entropy performs in-place operations which conflict with activation offloading. The default setting (`use_liger_kernel=False`) works:
+
+```python
+# When using activation offloading with a model that uses Liger kernels:
+from trl import SFTConfig
+
+training_args = SFTConfig(
+    activation_offloading=True,
+    use_liger_kernel=False,  # Disable Liger cross entropy
+    # Other parameters...
+)
+```
+</Tip>
+
+Under the hood, activation offloading implements PyTorch's [`saved_tensors_hooks`](https://pytorch.org/tutorials/intermediate/autograd_saved_tensors_hooks_tutorial.html#hooks-for-autograd-saved-tensors) to intercept activations during the forward pass. It intelligently manages which tensors to offload based on size and context, avoiding offloading output tensors which would be inefficient. For performance optimization, it can optionally use CUDA streams to overlap computation with CPU-GPU transfers.
+
+## Disabling model gathering for generation in online methods
+
+When using DeepSpeed ZeRO-3, model weights are sharded across multiple GPUs. Online methods involve generating completions from the model as part of the training process. During this step, the model weights are temporarily gathered on a single GPU for generation. For very large models, this gathering can lead to out-of-memory (OOM) errors, as described in this issue: [#2250](https://github.com/huggingface/trl/issues/2250#issue-2598304204).
+
+If you encounter this issue, you can disable the gathering of model weights for generation by setting the following parameter:
+
+<hfoptions id="ds3_gather_for_generation">
+<hfoption id="GRPO">
+
+```python
+from trl import GRPOConfig
+
+training_args = GRPOConfig(..., ds3_gather_for_generation=False)
+```
+
+</hfoption>
+<hfoption id="Online DPO">
+
+```python
+from trl import OnlineDPOConfig
+
+training_args = OnlineDPOConfig(..., ds3_gather_for_generation=False)
+```
+
+</hfoption>
+<hfoption id="PPO">
+
+```python
+from trl import PPOConfig
+
+training_args = PPOConfig(..., ds3_gather_for_generation=False)
+```
+
+</hfoption>
+<hfoption id="RLOO">
+
+```python
+from trl import RLOOConfig
+
+training_args = RLOOConfig(..., ds3_gather_for_generation=False)
+```
+
+</hfoption>
+</hfoptions>
+
+This adjustment prevents model weights from being gathered, avoiding OOM errors, but it may result in slower generation speeds.

docs/source/reward_trainer.md

@@ -1,23 +1,17 @@
 # Reward Modeling
 
+[![](https://img.shields.io/badge/All_models-Reward_Trainer-blue)](https://huggingface.co/models?other=reward-trainer,trl)
+
 TRL supports custom reward modeling for anyone to perform reward modeling on their dataset and model.
 
 Check out a complete flexible example at [`examples/scripts/reward_modeling.py`](https://github.com/huggingface/trl/tree/main/examples/scripts/reward_modeling.py).
 
-## Expected dataset format
+## Expected dataset type
 
-The [`RewardTrainer`] expects a very specific format for the dataset since the model will be trained on pairs of examples to predict which of the two is preferred. We provide an example from the [`Anthropic/hh-rlhf`](https://huggingface.co/datasets/Anthropic/hh-rlhf) dataset below:
+The [`RewardTrainer`] requires a [*implicit prompt* preference dataset](dataset_formats#preference). It means that the dataset should only contain the columns `"chosen"` and `"rejected"` (and not `"prompt"`).
+The [`RewardTrainer`] supports both [conversational](dataset_formats#conversational) and [standard](dataset_formats#standard) dataset format. When provided with a conversational dataset, the trainer will automatically apply the chat template to the dataset.
 
-<div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/rlhf-antropic-example.png", width="50%">
-</div>
-
-Therefore the final dataset object should contain two 4 entries at least if you use the default [`RewardDataCollatorWithPadding`] data collator. The entries should be named:
-
--   `input_ids_chosen`
--   `attention_mask_chosen`
--   `input_ids_rejected`
--   `attention_mask_rejected`
+You can also use a pretokenized dataset, in which case the dataset should contain the following columns: `input_ids_chosen`, `attention_mask_chosen`, `input_ids_rejected` and `attention_mask_rejected`.
 
 ## Using the `RewardTrainer`
 
@@ -47,7 +41,7 @@ peft_config = LoraConfig(
 trainer = RewardTrainer(
     model=model,
     args=training_args,
-    tokenizer=tokenizer,
+    processing_class=tokenizer,
     train_dataset=dataset,
     peft_config=peft_config,
 )
@@ -79,7 +73,7 @@ $$\Big( R(p, r_1) + R(p, r_2) \Big)^2 $$
 This auxiliary loss is combined with the main loss function, weighted by the parameter `center_rewards_coefficient` in the `[RewardConfig]`. By default, this feature is deactivated (`center_rewards_coefficient = None`).
 
 ```python
-reward_config = RewardConfig(
+training_args = RewardConfig(
     center_rewards_coefficient=0.01,
     ...
 )
@@ -87,10 +81,10 @@ reward_config = RewardConfig(
 
 For reference results, please refer PR [#1932](https://github.com/huggingface/trl/pull/1932).
 
-## RewardConfig
-
-[[autodoc]] RewardConfig
-
 ## RewardTrainer
 
 [[autodoc]] RewardTrainer
+
+## RewardConfig
+
+[[autodoc]] RewardConfig

docs/source/rewards.md

@@ -0,0 +1,9 @@
+# Reward Functions
+
+This module contains some useful reward functions, primarily intended for use with the [`GRPOTrainer`].
+
+## Format rewards
+
+### think_format_reward
+
+[[autodoc]] rewards.think_format_reward

docs/source/rloo_trainer.md

@@ -1,6 +1,8 @@
 # RLOO Trainer
 
-TRL supports training LLMs with REINFORCE Leave-One-Out (RLOO). The idea is that instead of using a value function, RLOO generates K completions for each prompt. For each completion, RLOO uses the mean scores from the other K-1 completions as a baseline to calculate the advantage. RLOO also models the entire completion as a single action, where as PPO models each token as an action. Note that REINFORCE / A2C is a special case of PPO, when the number of PPO epochs is 1 and the number of mini-batches is 1, which is how we implement RLOO in TRL.
+[![](https://img.shields.io/badge/All_models-RLOO-blue)](https://huggingface.co/models?other=rloo,trl)
+
+TRL supports training LLMs with REINFORCE Leave-One-Out (RLOO). The idea is that instead of using a value function, RLOO generates K completions for each prompt. For each completion, RLOO uses the mean scores from the other K-1 completions as a baseline to calculate the advantage. RLOO also models the entire completion as a single action, whereas PPO models each token as an action. Note that REINFORCE / A2C is a special case of PPO, when the number of PPO epochs is 1 and the number of mini-batches is 1, which is how we implement RLOO in TRL.
 
 References:
 - [Back to Basics: Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMs](https://huggingface.co/papers/2402.14740)
@@ -16,6 +18,8 @@ To just run a RLOO script to make sure the trainer can run, you can run the foll
 
 ```bash
 python examples/scripts/rloo/rloo.py \
+    --dataset_name trl-internal-testing/descriptiveness-sentiment-trl-style \
+    --dataset_train_split descriptiveness \
     --learning_rate 3e-6 \
     --output_dir models/minimal/rloo \
     --per_device_train_batch_size 64 \
@@ -23,7 +27,7 @@ python examples/scripts/rloo/rloo.py \
     --total_episodes 10000 \
     --model_name_or_path EleutherAI/pythia-14m \
     --reward_model_path EleutherAI/pythia-14m \
-    --non_eos_penalty
+    --missing_eos_penalty 1.0
 ```
 
 
@@ -54,17 +58,17 @@ The logged metrics are as follows. Here is an example [tracked run at Weights an
 ## Cookbook
 
 * Debugging TIP: `objective/rlhf_reward`: this is the ultimate objective of the RLHF training. If training works as intended, this metric should keep going up.
-* Debugging TIP: `val/ratio`: this number should float around 1.0, and it gets clipped by `--cliprange 0.2` with PPO's surrogate loss. So if this `ratio` is too high like 2.0 or 1000.0 or too small like 0.1, it means the updates between consecutive policies are too drastic. You should try undertand why this is happening and try to fix it.
+* Debugging TIP: `val/ratio`: this number should float around 1.0, and it gets clipped by `--cliprange 0.2` with PPO's surrogate loss. So if this `ratio` is too high like 2.0 or 1000.0 or too small like 0.1, it means the updates between consecutive policies are too drastic. You should try understand why this is happening and try to fix it.
 * Memory TIP: If you are running out of memory, you can try to reduce the `--per_device_train_batch_size` or increase the `--gradient_accumulation_steps` to reduce the memory footprint.
 * Memory TIP: If you have multiple GPUs, you can also run training with DeepSpeed stage 3 to reduce the memory footprint `accelerate launch --config_file examples/accelerate_configs/deepspeed_zero3.yaml`.
-* Usage TIP: We recommend to use the "EOS trick" via `--non_eos_penalty --stop_token eos`, which replaces the score of completions that do not end with an EOS token with a static scalar penalty `--penalty_reward_value`. This can help the model learn to generate more coherent completions.
+* Usage TIP: We recommend to use the "EOS trick" via `--missing_eos_penalty`, which subtracts a static scalar penalty from the score of completions that do not end with an EOS token. This can help the model learn to generate more coherent completions.
 
 
 ## What is my model doing exactly?
 
 To help you understand what your model is doing, we periodically log some sample completions from the model. Here is an example of a completion. In an example [tracked run at Weights and Biases](https://wandb.ai/huggingface/trl/runs/u2sqci34), it looks like the following, allowing you to see the model's response at different stages of training. By default we generate `--num_sample_generations 10` during training, but you can customize the number of generations.
 
-![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/ppov2_completions.gif)
+![](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/ppov2_completions.gif)
 
 
 In the logs the sampled generations look like 
@@ -208,19 +212,20 @@ To validate the RLOO implementation works, we ran experiment on the 1B model. He
 
 ```
 accelerate launch --config_file examples/accelerate_configs/deepspeed_zero2.yaml \
-    examples/scripts/rloo/rloo_tldr.py \
     --output_dir models/minimal/rloo_tldr \
+    --dataset_name trl-internal-testing/tldr-preference-sft-trl-style \
+    --dataset_test_split validation \
     --num_ppo_epochs 2 \
     --num_mini_batches 2 \
     --learning_rate 3e-6 \
-    --per_device_train_batch_size 8 \
-    --gradient_accumulation_steps 8 \
+    --per_device_train_batch_size 16 \
+    --gradient_accumulation_steps 16 \
     --total_episodes 1000000 \
     --model_name_or_path EleutherAI/pythia-1b-deduped \
     --sft_model_path cleanrl/EleutherAI_pythia-1b-deduped__sft__tldr \
     --reward_model_path cleanrl/EleutherAI_pythia-1b-deduped__reward__tldr \
     --local_rollout_forward_batch_size 16 \
-    --non_eos_penalty \
+    --missing_eos_penalty 1.0 \
     --stop_token eos \
     --kl_coef 0.03
 ```
@@ -246,7 +251,7 @@ The RLOO checkpoint gets a 51.2% preferred rate vs the 33.0% preference rate of
 
 Metrics:
 
-![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/benchmark/pr-1540/rloo.png)
+![](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/rloo.png)
 
 
 ```bash
@@ -263,3 +268,23 @@ python -m openrlbenchmark.rlops_multi_metrics \
     --output-filename benchmark/trl/pr-1540/rloo \
     --scan-history
 ```
+
+## Reinforce++
+
+The [Reinforce++](https://hijkzzz.notion.site/reinforce-plus-plus) report by Jian Hu suggests several optimization tricks to enhance performance and stability of RLHF. They include:
+
+- Clipping rewards: limiting reward values within a specific range to mitigate the impact of extreme rewards on model updates, thus preventing gradient explosion
+- Normalizing rewards: scaling rewards to have a mean of 0 and a standard deviation of 1, which helps in stabilizing the training process
+- Normalizing advantages: scaling advantages to have a mean of 0 and a standard deviation of 1, which helps in stabilizing the training process
+- Using token-level KL penalty that is defined as equation (1) of the report vs. sequence-level KL penalty (default)
+
+These options are available via the appropriate arguments in the [`RLOOConfig`] class.
+
+
+## RLOOTrainer
+
+[[autodoc]] RLOOTrainer
+
+## RLOOConfig
+
+[[autodoc]] RLOOConfig

docs/source/script_utils.md

@@ -0,0 +1,12 @@
+# Scripts Utilities
+
+## ScriptArguments
+
+[[autodoc]] ScriptArguments
+
+## TrlParser
+
+[[autodoc]] TrlParser
+    - parse_args_and_config
+    - parse_args_into_dataclasses
+    - set_defaults_with_config

docs/source/sentiment_tuning.md

@@ -0,0 +1,36 @@
+# Sentiment Tuning Examples
+
+The notebooks and scripts in this examples show how to fine-tune a model with a sentiment classifier (such as `lvwerra/distilbert-imdb`).
+
+Here's an overview of the notebooks and scripts in the [trl repository](https://github.com/huggingface/trl/tree/main/examples):
+
+
+
+| File                                                                                           | Description                                                                                                              |
+|------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------|
+| [`examples/scripts/ppo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/ppo.py)  [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/trl/blob/main/examples/sentiment/notebooks/gpt2-sentiment.ipynb) | This script shows how to use the `PPOTrainer` to fine-tune a sentiment analysis model using IMDB dataset                 |
+| [`examples/notebooks/gpt2-sentiment.ipynb`](https://github.com/huggingface/trl/tree/main/examples/notebooks/gpt2-sentiment.ipynb)              | This notebook demonstrates how to reproduce the GPT2 imdb sentiment tuning example on a jupyter notebook.                |
+| [`examples/notebooks/gpt2-control.ipynb`](https://github.com/huggingface/trl/tree/main/examples/notebooks/gpt2-control.ipynb)   [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/trl/blob/main/examples/sentiment/notebooks/gpt2-sentiment-control.ipynb)                | This notebook demonstrates how to reproduce the GPT2 sentiment control example on a jupyter notebook.    
+
+
+
+## Usage
+
+```bash
+# 1. run directly
+python examples/scripts/ppo.py
+# 2. run via `accelerate` (recommended), enabling more features (e.g., multiple GPUs, deepspeed)
+accelerate config # will prompt you to define the training configuration
+accelerate launch examples/scripts/ppo.py # launches training
+# 3. get help text and documentation
+python examples/scripts/ppo.py --help
+# 4. configure logging with wandb and, say, mini_batch_size=1 and gradient_accumulation_steps=16
+python examples/scripts/ppo.py --log_with wandb --mini_batch_size 1 --gradient_accumulation_steps 16
+```
+
+Note: if you don't want to log with `wandb` remove `log_with="wandb"` in the scripts/notebooks. You can also replace it with your favourite experiment tracker that's [supported by `accelerate`](https://huggingface.co/docs/accelerate/usage_guides/tracking).
+
+
+## Few notes on multi-GPU 
+
+To run in multi-GPU setup with DDP (distributed Data Parallel) change the `device_map` value to `device_map={"": Accelerator().process_index}` and make sure to run your script with `accelerate launch yourscript.py`. If you want to apply naive pipeline parallelism you can use `device_map="auto"`.

docs/source/sentiment_tuning.mdx

@@ -1,130 +0,0 @@
-# Sentiment Tuning Examples
-
-The notebooks and scripts in this examples show how to fine-tune a model with a sentiment classifier (such as `lvwerra/distilbert-imdb`).
-
-Here's an overview of the notebooks and scripts in the [trl repository](https://github.com/huggingface/trl/tree/main/examples):
-
-
-
-| File                                                                                           | Description                                                                                                              |
-|------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------|
-| [`examples/scripts/ppo.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/ppo.py)  [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/trl/blob/main/examples/sentiment/notebooks/gpt2-sentiment.ipynb) | This script shows how to use the `PPOTrainer` to fine-tune a sentiment analysis model using IMDB dataset                 |
-| [`examples/notebooks/gpt2-sentiment.ipynb`](https://github.com/huggingface/trl/tree/main/examples/notebooks/gpt2-sentiment.ipynb)              | This notebook demonstrates how to reproduce the GPT2 imdb sentiment tuning example on a jupyter notebook.                |
-| [`examples/notebooks/gpt2-control.ipynb`](https://github.com/huggingface/trl/tree/main/examples/notebooks/gpt2-control.ipynb)   [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/huggingface/trl/blob/main/examples/sentiment/notebooks/gpt2-sentiment-control.ipynb)                | This notebook demonstrates how to reproduce the GPT2 sentiment control example on a jupyter notebook.    
-
-
-
-## Usage
-
-```bash
-# 1. run directly
-python examples/scripts/ppo.py
-# 2. run via `accelerate` (recommended), enabling more features (e.g., multiple GPUs, deepspeed)
-accelerate config # will prompt you to define the training configuration
-accelerate launch examples/scripts/ppo.py # launches training
-# 3. get help text and documentation
-python examples/scripts/ppo.py --help
-# 4. configure logging with wandb and, say, mini_batch_size=1 and gradient_accumulation_steps=16
-python examples/scripts/ppo.py --log_with wandb --mini_batch_size 1 --gradient_accumulation_steps 16
-```
-
-Note: if you don't want to log with `wandb` remove `log_with="wandb"` in the scripts/notebooks. You can also replace it with your favourite experiment tracker that's [supported by `accelerate`](https://huggingface.co/docs/accelerate/usage_guides/tracking).
-
-
-## Few notes on multi-GPU 
-
-To run in multi-GPU setup with DDP (distributed Data Parallel) change the `device_map` value to `device_map={"": Accelerator().process_index}` and make sure to run your script with `accelerate launch yourscript.py`. If you want to apply naive pipeline parallelism you can use `device_map="auto"`.
-
-
-## Benchmarks
-
-Below are some benchmark results for `examples/scripts/ppo.py`. To reproduce locally, please check out the `--command` arguments below.
-
-```bash
-python benchmark/benchmark.py \
-    --command "python examples/scripts/ppo.py --log_with wandb" \
-    --num-seeds 5 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 12 \
-    --slurm-template-path benchmark/trl.slurm_template
-```
-
-![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/benchmark/v0.4.7-55-g110e672/sentiment.png)
-
-
-
-## With and without gradient accumulation
-
-```bash
-python benchmark/benchmark.py \
-    --command "python examples/scripts/ppo.py --exp_name sentiment_tuning_step_grad_accu --mini_batch_size 1 --gradient_accumulation_steps 128 --log_with wandb" \
-    --num-seeds 5 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 12 \
-    --slurm-template-path benchmark/trl.slurm_template
-```
-
-![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/benchmark/v0.4.7-55-g110e672/gradient_accu.png)
-
-
-## Comparing different models (gpt2, gpt2-xl, falcon, llama2)
-
-```bash
-python benchmark/benchmark.py \
-    --command "python examples/scripts/ppo.py --exp_name sentiment_tuning_gpt2 --log_with wandb" \
-    --num-seeds 5 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 12 \
-    --slurm-template-path benchmark/trl.slurm_template
-python benchmark/benchmark.py \
-    --command "python examples/scripts/ppo.py --exp_name sentiment_tuning_gpt2xl_grad_accu --model_name gpt2-xl --mini_batch_size 16 --gradient_accumulation_steps 8 --log_with wandb" \
-    --num-seeds 5 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 12 \
-    --slurm-template-path benchmark/trl.slurm_template
-python benchmark/benchmark.py \
-    --command "python examples/scripts/ppo.py --exp_name sentiment_tuning_falcon_rw_1b --model_name tiiuae/falcon-rw-1b --log_with wandb" \
-    --num-seeds 5 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 12 \
-    --slurm-template-path benchmark/trl.slurm_template
-```
-
-![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/benchmark/v0.4.7-55-g110e672/different_models.png)
-
-## With and without PEFT
-
-```
-python benchmark/benchmark.py \
-    --command "python examples/scripts/ppo.py --exp_name sentiment_tuning_peft --use_peft --log_with wandb" \
-    --num-seeds 5 \
-    --start-seed 1 \
-    --workers 10 \
-    --slurm-nodes 1 \
-    --slurm-gpus-per-task 1 \
-    --slurm-ntasks 1 \
-    --slurm-total-cpus 12 \
-    --slurm-template-path benchmark/trl.slurm_template
-```
-
-![](https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/benchmark/v0.4.7-55-g110e672/peft.png)

docs/source/sft_trainer.md

@@ -0,0 +1,406 @@
+# SFT Trainer
+
+[![All_models-SFT-blue](https://img.shields.io/badge/All_models-SFT-blue)](https://huggingface.co/models?other=sft,trl) [![smol_course-Chapter_1-yellow](https://img.shields.io/badge/smol_course-Chapter_1-yellow)](https://github.com/huggingface/smol-course/tree/main/1_instruction_tuning)
+
+## Overview
+
+TRL supports the Supervised Fine-Tuning (SFT) Trainer for training language models.
+
+This post-training method was contributed by [Younes Belkada](https://huggingface.co/ybelkada).
+
+## Quick start
+
+This example demonstrates how to train a language model using the [`SFTTrainer`] from TRL. We train a [Qwen 3 0.6B](https://huggingface.co/Qwen/Qwen3-0.6B) model on the [Capybara dataset](https://huggingface.co/datasets/trl-lib/Capybara), a compact, diverse multi-turn dataset to benchmark reasoning and generalization.
+
+```python
+from trl import SFTTrainer, SFTConfig
+from datasets import load_dataset
+
+trainer = SFTTrainer(
+    model="Qwen/Qwen3-0.6B",
+    train_dataset=load_dataset("trl-lib/Capybara", split="train"),
+)
+trainer.train()
+```
+
+<iframe src="https://trl-lib-trackio.hf.space/?project=trl-documentation&metrics=train/loss,train/mean_token_accuracy,train/num_tokens&sidebar=hidden" style="width: 100%; min-width: 300px; max-width: 800px;" height="830" frameBorder="0"></iframe>
+
+## Expected dataset type and format
+
+SFT supports both [language modeling](dataset_formats#language-modeling) and [prompt-completion](dataset_formats#prompt-completion) datasets. The [`SFTTrainer`] is compatible with both [standard](dataset_formats#standard) and [conversational](dataset_formats#conversational) dataset formats. When provided with a conversational dataset, the trainer will automatically apply the chat template to the dataset.
+
+```python
+# Standard language modeling
+{"text": "The sky is blue."}
+
+# Conversational language modeling
+{"messages": [{"role": "user", "content": "What color is the sky?"},
+              {"role": "assistant", "content": "It is blue."}]}
+
+# Standard prompt-completion
+{"prompt": "The sky is",
+ "completion": " blue."}
+
+# Conversational prompt-completion
+{"prompt": [{"role": "user", "content": "What color is the sky?"}],
+ "completion": [{"role": "assistant", "content": "It is blue."}]}
+```
+
+If your dataset is not in one of these formats, you can preprocess it to convert it into the expected format. Here is an example with the [FreedomIntelligence/medical-o1-reasoning-SFT](https://huggingface.co/datasets/FreedomIntelligence/medical-o1-reasoning-SFT) dataset:
+
+```python
+from datasets import load_dataset
+
+dataset = load_dataset("FreedomIntelligence/medical-o1-reasoning-SFT", "en")
+
+def preprocess_function(example):
+    return {
+        "prompt": [{"role": "user", "content": example["Question"]}],
+        "completion": [
+            {"role": "assistant", "content": f"<think>{example['Complex_CoT']}</think>{example['Response']}"}
+        ],
+    }
+
+dataset = dataset.map(preprocess_function, remove_columns=["Question", "Response", "Complex_CoT"])
+print(next(iter(dataset["train"])))
+```
+
+```json
+{
+    "prompt": [
+        {
+            "content": "Given the symptoms of sudden weakness in the left arm and leg, recent long-distance travel, and the presence of swollen and tender right lower leg, what specific cardiac abnormality is most likely to be found upon further evaluation that could explain these findings?",
+            "role": "user",
+        }
+    ],
+    "completion": [
+        {
+            "content": "<think>Okay, let's see what's going on here. We've got sudden weakness [...] clicks into place!</think>The specific cardiac abnormality most likely to be found in [...] the presence of a PFO facilitating a paradoxical embolism.",
+            "role": "assistant",
+        }
+    ],
+}
+```
+
+## Looking deeper into the SFT method
+
+Supervised Fine-Tuning (SFT) is the simplest and most commonly used method to adapt a language model to a target dataset. The model is trained in a fully supervised fashion using pairs of input and output sequences. The goal is to minimize the negative log-likelihood (NLL) of the target sequence, conditioning on the input.
+
+This section breaks down how SFT works in practice, covering the key steps: **preprocessing**, **tokenization** and **loss computation**.
+
+### Preprocessing and tokenization
+
+During training, each example is expected to contain a **text field** or a **(prompt, completion)** pair, depending on the dataset format. For more details on the expected formats, see [Dataset formats](dataset_formats).
+The `SFTTrainer` tokenizes each input using the model's tokenizer. If both prompt and completion are provided separately, they are concatenated before tokenization.
+
+### Computing the loss
+
+![sft_figure](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/sft_figure.png)
+
+The loss used in SFT is the **token-level cross-entropy loss**, defined as:
+
+$$
+\mathcal{L}_{\text{SFT}}(\theta) = - \sum_{t=1}^{T} \log p_\theta(y_t \mid y_{<t}),
+$$
+  
+where  \\( y_t \\) is the target token at timestep  \\( t \\), and the model is trained to predict the next token given the previous ones. In practice, padding tokens are masked out during loss computation.
+
+### Label shifting and masking
+
+During training, the loss is computed using a **one-token shift**: the model is trained to predict each token in the sequence based on all previous tokens. Specifically, the input sequence is shifted right by one position to form the target labels.
+Padding tokens (if present) are ignored in the loss computation by applying an ignore index (default: `-100`) to the corresponding positions. This ensures that the loss focuses only on meaningful, non-padding tokens.
+
+## Logged metrics
+
+* `global_step`: The total number of optimizer steps taken so far.
+* `epoch`: The current epoch number, based on dataset iteration.
+* `num_tokens`: The total number of tokens processed so far.
+* `loss`: The average cross-entropy loss computed over non-masked tokens in the current logging interval.
+* `mean_token_accuracy`: The proportion of non-masked tokens for which the model’s top-1 prediction matches the ground truth token.
+* `learning_rate`: The current learning rate, which may change dynamically if a scheduler is used.
+* `grad_norm`: The L2 norm of the gradients, computed before gradient clipping.
+
+## Customization
+
+### Model initialization
+
+You can directly pass the kwargs of the [`~transformers.AutoModelForCausalLM.from_pretrained()`] method to the [`SFTConfig`]. For example, if you want to load a model in a different precision, analogous to
+
+```python
+model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-0.6B", torch_dtype=torch.bfloat16)
+```
+
+you can do so by passing the `model_init_kwargs={"torch_dtype": torch.bfloat16}` argument to the [`SFTConfig`].
+
+```python
+from trl import SFTConfig
+
+training_args = SFTConfig(
+    model_init_kwargs={"torch_dtype": torch.bfloat16},
+)
+```
+
+Note that all keyword arguments of [`~transformers.AutoModelForCausalLM.from_pretrained()`] are supported.
+
+### Packing
+
+[`SFTTrainer`] supports _example packing_, where multiple examples are packed in the same input sequence to increase training efficiency. To enable packing, simply pass `packing=True` to the [`SFTConfig`] constructor.
+
+```python
+training_args = SFTConfig(packing=True)
+```
+
+For more details on packing, see [Packing](reducing_memory_usage#packing).
+
+### Train on assistant messages only
+
+To train on assistant messages only, use a [conversational](dataset_formats#conversational) dataset and set `assistant_only_loss=True` in the [`SFTConfig`]. This setting ensures that loss is computed **only** on the assistant responses, ignoring user or system messages.
+
+```python
+training_args = SFTConfig(assistant_only_loss=True)
+```
+
+![train_on_assistant](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/train_on_assistant.png)
+
+> [!WARNING]
+> This functionality is only available for chat templates that support returning the assistant tokens mask via the `{% generation %}` and `{% endgeneration %}` keywords. For an example of such a template, see [HugggingFaceTB/SmolLM3-3B](https://huggingface.co/HuggingFaceTB/SmolLM3-3B/blob/main/chat_template.jinja#L76-L82).
+
+### Train on completion only
+
+To train on completion only, use a [prompt-completion](dataset_formats#prompt-completion) dataset. By default, the trainer computes the loss on the completion tokens only, ignoring the prompt tokens. If you want to train on the full sequence, set `completion_only_loss=False` in the [`SFTConfig`].
+
+![train_on_completion](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/train_on_completion.png)
+
+<Tip>
+Training on completion only is compatible with training on assistant messages only. In this case, use a [conversational](dataset_formats#conversational) [prompt-completion](dataset_formats#prompt-completion) dataset and set `assistant_only_loss=True` in the [`SFTConfig`].
+</Tip>
+
+### Train adapters with PEFT
+
+We support tight integration with 🤗 PEFT library, allowing any user to conveniently train adapters and share them on the Hub, rather than training the entire model.
+
+```python
+from datasets import load_dataset
+from trl import SFTTrainer
+from peft import LoraConfig
+
+dataset = load_dataset("trl-lib/Capybara", split="train")
+
+trainer = SFTTrainer(
+    "Qwen/Qwen3-0.6B",
+    train_dataset=dataset,
+    peft_config=LoraConfig()
+)
+
+trainer.train()
+```
+
+You can also continue training your [`peft.PeftModel`]. For that, first load a `PeftModel` outside [`SFTTrainer`] and pass it directly to the trainer without the `peft_config` argument being passed.
+
+```python
+from datasets import load_dataset
+from trl import SFTTrainer
+from peft import AutoPeftModelForCausalLM
+
+model = AutoPeftModelForCausalLM.from_pretrained("trl-lib/Qwen3-4B-LoRA", is_trainable=True)
+dataset = load_dataset("trl-lib/Capybara", split="train")
+
+trainer = SFTTrainer(
+    model=model,
+    train_dataset=dataset,
+)
+
+trainer.train()
+```
+
+<Tip>
+
+When training adapters, you typically use a higher learning rate (≈1e‑4) since only new parameters are being learned.
+
+```python
+SFTConfig(learning_rate=1e-4, ...)
+```
+
+</Tip>
+
+### Train with Liger Kernel
+
+Liger Kernel is a collection of Triton kernels for LLM training that boosts multi-GPU throughput by 20%, cuts memory use by 60% (enabling up to 4× longer context), and works seamlessly with tools like Flash Attention, PyTorch FSDP, and DeepSpeed. For more information, see [Liger Kernel Integration](liger_kernel_integration).
+
+### Train with Unsloth
+
+Unsloth is an open‑source framework for fine‑tuning and reinforcement learning that trains LLMs (like Llama, Mistral, Gemma, DeepSeek, and more) up to 2× faster with up to 70% less VRAM, while providing a streamlined, Hugging Face–compatible workflow for training, evaluation, and deployment. For more information, see [Unsloth Integration](unsloth_integration).
+
+## Instruction tuning example
+
+**Instruction tuning** teaches a base language model to follow user instructions and engage in conversations. This requires:
+
+1. **Chat template**: Defines how to structure conversations into text sequences, including role markers (user/assistant), special tokens, and turn boundaries. Read more about chat templates in [Chat templates](https://huggingface.co/docs/transformers/chat_templating#templates).
+2. **Conversational dataset**: Contains instruction-response pairs
+
+This example shows how to transform the [Qwen 3 0.6B Base](https://huggingface.co/Qwen/Qwen3-0.6B-Base) model into an instruction-following model using the [Capybara dataset](https://huggingface.co/datasets/trl-lib/Capybara) and a chat template from [HuggingFaceTB/SmolLM3-3B](https://huggingface.co/HuggingFaceTB/SmolLM3-3B). The SFT Trainer automatically handles tokenizer updates and special token configuration.
+
+```python
+from trl import SFTTrainer, SFTConfig
+from datasets import load_dataset
+
+trainer = SFTTrainer(
+    model="Qwen/Qwen3-0.6B-Base",
+    args=SFTConfig(
+        output_dir="Qwen3-0.6B-Instruct",
+        chat_template_path="HuggingFaceTB/SmolLM3-3B",
+    ),
+    train_dataset=load_dataset("trl-lib/Capybara", split="train"),
+)
+trainer.train()
+```
+
+> [!WARNING]
+> Some base models, like those from Qwen, have a predefined chat template in the model's tokenizer. In these cases, it is not necessary to apply [`clone_chat_template()`], as the tokenizer already handles the formatting. However, it is necessary to align the EOS token with the chat template to ensure the model's responses terminate correctly. In these cases, specify `eos_token` in [`SFTConfig`]; for example, for `Qwen/Qwen2.5-1.5B`, one should set `eos_token="<|im_end|>"`.
+
+Once trained, your model can now follow instructions and engage in conversations using its new chat template.
+
+```python
+>>> from transformers import pipeline
+>>> pipe = pipeline("text-generation", model="Qwen3-0.6B-Instruct/checkpoint-5000")
+>>> prompt = "<|im_start|>user\nWhat is the capital of France? Answer in one word.<|im_end|>\n<|im_start|>assistant\n"
+>>> response = pipe(prompt)
+>>> response[0]["generated_text"]
+'<|im_start|>user\nWhat is the capital of France? Answer in one word.<|im_end|>\n<|im_start|>assistant\nThe capital of France is Paris.'
+```
+
+Alternatively, use the structured conversation format (recommended):
+
+```python
+>>> prompt = [{"role": "user", "content": "What is the capital of France? Answer in one word."}]
+>>> response = pipe(prompt)
+>>> response[0]["generated_text"]
+[{'role': 'user', 'content': 'What is the capital of France? Answer in one word.'}, {'role': 'assistant', 'content': 'The capital of France is Paris.'}]
+```
+
+## Tool Calling with SFT
+
+The SFT trainer fully supports fine-tuning models with _tool calling_ capabilities. In this case, each dataset example should include:
+
+* The conversation messages, including any tool calls (`tool_calls`) and tool responses (`tool` role messages)
+* The list of available tools in the `tools` column, typically provided as JSON schemas
+
+For details on the expected dataset structure, see the [Dataset Format — Tool Calling](dataset_formats#tool-calling) section.
+
+## Extending `SFTTrainer` for Vision Language Models
+
+`SFTTrainer` does not yet inherently support vision-language data. However, we provide a guide on how to tweak the trainer to support vision-language data. Specifically, you need to use a custom data collator that is compatible with vision-language data. This guide outlines the steps to make these adjustments. For a concrete example, refer to the script [`examples/scripts/sft_vlm.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/sft_vlm.py), which demonstrates how to fine-tune the LLaVA 1.5 model on the [HuggingFaceH4/llava-instruct-mix-vsft](https://huggingface.co/datasets/HuggingFaceH4/llava-instruct-mix-vsft) dataset.
+
+### Preparing the Data
+
+The data format is flexible, provided it is compatible with the custom collator that we will define later. A common approach is to use conversational data. Given that the data includes both text and images, the format needs to be adjusted accordingly. Below is an example of a conversational data format involving both text and images:
+
+```python
+images = ["obama.png"]
+messages = [
+    {
+        "role": "user",
+        "content": [
+            {"type": "text", "text": "Who is this?"},
+            {"type": "image"}
+        ]
+    },
+    {
+        "role": "assistant",
+        "content": [
+            {"type": "text", "text": "Barack Obama"}
+        ]
+    },
+    {
+        "role": "user",
+        "content": [
+            {"type": "text", "text": "What is he famous for?"}
+        ]
+    },
+    {
+        "role": "assistant",
+        "content": [
+            {"type": "text", "text": "He is the 44th President of the United States."}
+        ]
+    }
+]
+```
+
+To illustrate how this data format will be processed using the LLaVA model, you can use the following code:
+
+```python
+from transformers import AutoProcessor
+
+processor = AutoProcessor.from_pretrained("llava-hf/llava-1.5-7b-hf")
+print(processor.apply_chat_template(messages, tokenize=False))
+```
+
+The output will be formatted as follows:
+
+```txt
+Who is this? ASSISTANT: Barack Obama USER: What is he famous for? ASSISTANT: He is the 44th President of the United States. 
+```
+
+<iframe src="https://huggingface.co/datasets/HuggingFaceH4/llava-instruct-mix-vsft/embed/viewer/default/train" frameborder="0" width="100%" height="560px"></iframe>
+
+### A custom collator for processing multi-modal data
+
+Unlike the default behavior of [`SFTTrainer`], processing multi-modal data is done on the fly during the data collation process. To do this, you need to define a custom collator that processes both the text and images. This collator must take a list of examples as input (see the previous section for an example of the data format) and return a batch of processed data. Below is an example of such a collator:
+
+```python
+def collate_fn(examples):
+    # Get the texts and images, and apply the chat template
+    texts = [processor.apply_chat_template(example["messages"], tokenize=False) for example in examples]
+    images = [example["images"][0] for example in examples]
+
+    # Tokenize the texts and process the images
+    batch = processor(images=images, text=texts, return_tensors="pt", padding=True)
+
+    # The labels are the input_ids, and we mask the padding tokens in the loss computation
+    labels = batch["input_ids"].clone()
+    labels[labels == processor.tokenizer.pad_token_id] = -100
+    batch["labels"] = labels
+
+    return batch
+```
+
+We can verify that the collator works as expected by running the following code:
+
+```python
+from datasets import load_dataset
+
+dataset = load_dataset("HuggingFaceH4/llava-instruct-mix-vsft", split="train")
+examples = [dataset[0], dataset[1]]  # Just two examples for the sake of the example
+collated_data = collate_fn(examples)
+print(collated_data.keys())  # dict_keys(['input_ids', 'attention_mask', 'pixel_values', 'labels'])
+```
+
+### Training the vision-language model
+
+Now that we have prepared the data and defined the collator, we can proceed with training the model. To ensure that the data is not processed as text-only, we need to set a couple of arguments in the [`SFTConfig`], specifically `remove_unused_columns` and `skip_prepare_dataset` to `True` to avoid the default processing of the dataset. Below is an example of how to set up the `SFTTrainer`.
+
+```python
+training_args.remove_unused_columns = False
+training_args.dataset_kwargs = {"skip_prepare_dataset": True}
+
+trainer = SFTTrainer(
+    model=model,
+    args=training_args,
+    data_collator=collate_fn,
+    train_dataset=train_dataset,
+    processing_class=processor,
+)
+```
+
+A full example of training LLaVa 1.5 on the [HuggingFaceH4/llava-instruct-mix-vsft](https://huggingface.co/datasets/HuggingFaceH4/llava-instruct-mix-vsft) dataset can be found in the script [`examples/scripts/sft_vlm.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/sft_vlm.py).
+
+* [Experiment tracking](https://wandb.ai/huggingface/trl/runs/2b2c5l7s)
+* [Trained model](https://huggingface.co/HuggingFaceH4/sft-llava-1.5-7b-hf)
+
+## SFTTrainer
+
+[[autodoc]] SFTTrainer
+
+## SFTConfig
+
+[[autodoc]] SFTConfig

docs/source/sft_trainer.mdx

@@ -1,752 +0,0 @@
-# Supervised Fine-tuning Trainer
-
-Supervised fine-tuning (or SFT for short) is a crucial step in RLHF. In TRL we provide an easy-to-use API to create your SFT models and train them with few lines of code on your dataset.
-
-Check out a complete flexible example at [`examples/scripts/sft.py`](https://github.com/huggingface/trl/tree/main/examples/scripts/sft.py).
-Experimental support for Vision Language Models is also included in the example [`examples/scripts/vsft_llava.py`](https://github.com/huggingface/trl/tree/main/examples/scripts/vsft_llava.py).
-
-## Quickstart
-
-If you have a dataset hosted on the 🤗 Hub, you can easily fine-tune your SFT model using [`SFTTrainer`] from TRL. Let us assume your dataset is `imdb`, the text you want to predict is inside the `text` field of the dataset, and you want to fine-tune the `facebook/opt-350m` model.
-The following code-snippet takes care of all the data pre-processing and training for you:
-
-```python
-from datasets import load_dataset
-from trl import SFTConfig, SFTTrainer
-
-dataset = load_dataset("imdb", split="train")
-
-sft_config = SFTConfig(
-    dataset_text_field="text",
-    max_seq_length=512,
-    output_dir="/tmp",
-)
-trainer = SFTTrainer(
-    "facebook/opt-350m",
-    train_dataset=dataset,
-    args=sft_config,
-)
-trainer.train()
-```
-Make sure to pass the correct value for `max_seq_length` as the default value will be set to `min(tokenizer.model_max_length, 1024)`.
-
-You can also construct a model outside of the trainer and pass it as follows:
-
-```python
-from transformers import AutoModelForCausalLM
-from datasets import load_dataset
-from trl import SFTConfig, SFTTrainer
-
-dataset = load_dataset("imdb", split="train")
-
-model = AutoModelForCausalLM.from_pretrained("facebook/opt-350m")
-
-sft_config = SFTConfig(output_dir="/tmp")
-
-trainer = SFTTrainer(
-    model,
-    train_dataset=dataset,
-    args=sft_config,
-)
-
-trainer.train()
-```
-
-The above snippets will use the default training arguments from the [`SFTConfig`] class. If you want to modify the defaults pass in your modification to the `SFTConfig` constructor and pass them to the trainer via the `args` argument.
-
-## Advanced usage
-
-### Train on completions only
-
-You can use the `DataCollatorForCompletionOnlyLM` to train your model on the generated prompts only. Note that this works only in the case when `packing=False`.
-To instantiate that collator for instruction data, pass a response template and the tokenizer. Here is an example of how it would work to fine-tune `opt-350m` on completions only on the CodeAlpaca dataset:
-
-```python
-from transformers import AutoModelForCausalLM, AutoTokenizer
-from datasets import load_dataset
-from trl import SFTConfig, SFTTrainer, DataCollatorForCompletionOnlyLM
-
-dataset = load_dataset("lucasmccabe-lmi/CodeAlpaca-20k", split="train")
-
-model = AutoModelForCausalLM.from_pretrained("facebook/opt-350m")
-tokenizer = AutoTokenizer.from_pretrained("facebook/opt-350m")
-
-def formatting_prompts_func(example):
-    output_texts = []
-    for i in range(len(example['instruction'])):
-        text = f"### Question: {example['instruction'][i]}\n ### Answer: {example['output'][i]}"
-        output_texts.append(text)
-    return output_texts
-
-response_template = " ### Answer:"
-collator = DataCollatorForCompletionOnlyLM(response_template, tokenizer=tokenizer)
-
-trainer = SFTTrainer(
-    model,
-    train_dataset=dataset,
-    args=SFTConfig(output_dir="/tmp"),
-    formatting_func=formatting_prompts_func,
-    data_collator=collator,
-)
-
-trainer.train()
-```
-
-To instantiate that collator for assistant style conversation data, pass a response template, an instruction template and the tokenizer. Here is an example of how it would work to fine-tune `opt-350m` on assistant completions only on the Open Assistant Guanaco dataset:
-
-```python
-from transformers import AutoModelForCausalLM, AutoTokenizer
-from datasets import load_dataset
-from trl import SFTConfig, SFTTrainer, DataCollatorForCompletionOnlyLM
-
-dataset = load_dataset("timdettmers/openassistant-guanaco", split="train")
-
-model = AutoModelForCausalLM.from_pretrained("facebook/opt-350m")
-tokenizer = AutoTokenizer.from_pretrained("facebook/opt-350m")
-
-instruction_template = "### Human:"
-response_template = "### Assistant:"
-collator = DataCollatorForCompletionOnlyLM(instruction_template=instruction_template, response_template=response_template, tokenizer=tokenizer, mlm=False)
-
-trainer = SFTTrainer(
-    model,
-    args=SFTConfig(
-        output_dir="/tmp",
-        dataset_text_field = "text",
-    ),
-    train_dataset=dataset,
-    data_collator=collator,
-)
-
-trainer.train()
-```
-
-Make sure to have a `pad_token_id` which is different from `eos_token_id` which can result in the model not properly predicting EOS (End of Sentence) tokens during generation.
-
-#### Using token_ids directly for `response_template`
-
-Some tokenizers like Llama 2 (`meta-llama/Llama-2-XXb-hf`) tokenize sequences differently depending on whether they have context or not. For example:
-
-```python
-from transformers import AutoTokenizer
-tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-hf")
-
-def print_tokens_with_ids(txt):
-    tokens = tokenizer.tokenize(txt, add_special_tokens=False)
-    token_ids = tokenizer.encode(txt, add_special_tokens=False)
-    print(list(zip(tokens, token_ids)))
-
-prompt = """### User: Hello\n\n### Assistant: Hi, how can I help you?"""
-print_tokens_with_ids(prompt)  # [..., ('▁Hello', 15043), ('<0x0A>', 13), ('<0x0A>', 13), ('##', 2277), ('#', 29937), ('▁Ass', 4007), ('istant', 22137), (':', 29901), ...]
-
-response_template = "### Assistant:"
-print_tokens_with_ids(response_template)  # [('▁###', 835), ('▁Ass', 4007), ('istant', 22137), (':', 29901)]
-```
-
-In this case, and due to lack of context in `response_template`, the same string ("### Assistant:") is tokenized differently:
-
-    - Text (with context): `[2277, 29937, 4007, 22137, 29901]`
-    - `response_template` (without context): `[835, 4007, 22137, 29901]`
-
-This will lead to an error when the `DataCollatorForCompletionOnlyLM` does not find the `response_template` in the dataset example text:
-
-```
-RuntimeError: Could not find response key [835, 4007, 22137, 29901] in token IDs tensor([    1,   835,  ...])
-```
-
-
-To solve this, you can tokenize the `response_template` with the same context as in the dataset, truncate it as needed and pass the `token_ids` directly to the `response_template` argument of the `DataCollatorForCompletionOnlyLM` class. For example:
-
-```python
-response_template_with_context = "\n### Assistant:"  # We added context here: "\n". This is enough for this tokenizer
-response_template_ids = tokenizer.encode(response_template_with_context, add_special_tokens=False)[2:]  # Now we have it like in the dataset texts: `[2277, 29937, 4007, 22137, 29901]`
-
-data_collator = DataCollatorForCompletionOnlyLM(response_template_ids, tokenizer=tokenizer)
-```
-
-### Add Special Tokens for Chat Format
-
-Adding special tokens to a language model is crucial for training chat models. These tokens are added between the different roles in a conversation, such as the user, assistant, and system and help the model recognize the structure and flow of a conversation. This setup is essential for enabling the model to generate coherent and contextually appropriate responses in a chat environment. 
-The [`setup_chat_format`] function in `trl` easily sets up a model and tokenizer for conversational AI tasks. This function:
-- Adds special tokens to the tokenizer, e.g. `<|im_start|>` and `<|im_end|>`, to indicate the start and end of a conversation.
-- Resizes the model’s embedding layer to accommodate the new tokens.
-- Sets the `chat_template` of the tokenizer, which is used to format the input data into a chat-like format. The default is `chatml` from OpenAI.
-- _optionally_ you can pass `resize_to_multiple_of` to resize the embedding layer to a multiple of the `resize_to_multiple_of` argument, e.g. 64. If you want to see more formats being supported in the future, please open a GitHub issue on [trl](https://github.com/huggingface/trl)
-
-```python
-from transformers import AutoModelForCausalLM, AutoTokenizer
-from trl import setup_chat_format
-
-# Load model and tokenizer
-model = AutoModelForCausalLM.from_pretrained("facebook/opt-350m")
-tokenizer = AutoTokenizer.from_pretrained("facebook/opt-350m")
-
-# Set up the chat format with default 'chatml' format
-model, tokenizer = setup_chat_format(model, tokenizer)
-
-```
-
-With our model and tokenizer set up, we can now fine-tune our model on a conversational dataset. Below is an example of how a dataset can be formatted for fine-tuning. 
-
-### Dataset format support
-
-The [`SFTTrainer`] supports popular dataset formats. This allows you to pass the dataset to the trainer without any pre-processing directly. The following formats are supported:
-* conversational format
-```json
-{"messages": [{"role": "system", "content": "You are helpful"}, {"role": "user", "content": "What's the capital of France?"}, {"role": "assistant", "content": "..."}]}
-{"messages": [{"role": "system", "content": "You are helpful"}, {"role": "user", "content": "Who wrote 'Romeo and Juliet'?"}, {"role": "assistant", "content": "..."}]}
-{"messages": [{"role": "system", "content": "You are helpful"}, {"role": "user", "content": "How far is the Moon from Earth?"}, {"role": "assistant", "content": "..."}]}
-```
-* instruction format
-```json
-{"prompt": "<prompt text>", "completion": "<ideal generated text>"}
-{"prompt": "<prompt text>", "completion": "<ideal generated text>"}
-{"prompt": "<prompt text>", "completion": "<ideal generated text>"}
-```
-
-If your dataset uses one of the above formats, you can directly pass it to the trainer without pre-processing. The [`SFTTrainer`] will then format the dataset for you using the defined format from the model's tokenizer with the [apply_chat_template](https://huggingface.co/docs/transformers/main/en/chat_templating#templates-for-chat-models) method. 
-
-
-```python
-from datasets import load_dataset
-from trl import SFTConfig, SFTTrainer
-
-...
-
-# load jsonl dataset
-dataset = load_dataset("json", data_files="path/to/dataset.jsonl", split="train")
-# load dataset from the HuggingFace Hub
-dataset = load_dataset("philschmid/dolly-15k-oai-style", split="train")
-
-...
-
-sft_config = SFTConfig(packing=True)
-trainer = SFTTrainer(
-    "facebook/opt-350m",
-    args=sft_config,
-    train_dataset=dataset,
-)
-```
-
-If the dataset is not in one of those format you can either preprocess the dataset to match the formatting or pass a formatting function to the SFTTrainer to do it for you. Let's have a look.
-
-
-### Format your input prompts
-
-For instruction fine-tuning, it is quite common to have two columns inside the dataset: one for the prompt & the other for the response.
-This allows people to format examples like [Stanford-Alpaca](https://github.com/tatsu-lab/stanford_alpaca) did as follows:
-```bash
-Below is an instruction ...
-
-### Instruction
-{prompt}
-
-### Response:
-{completion}
-```
-Let us assume your dataset has two fields, `question` and `answer`. Therefore you can just run:
-```python
-...
-def formatting_prompts_func(example):
-    output_texts = []
-    for i in range(len(example['question'])):
-        text = f"### Question: {example['question'][i]}\n ### Answer: {example['answer'][i]}"
-        output_texts.append(text)
-    return output_texts
-
-trainer = SFTTrainer(
-    model,
-    args=sft_config,
-    train_dataset=dataset,
-    formatting_func=formatting_prompts_func,
-)
-
-trainer.train()
-```
-To properly format your input make sure to process all the examples by looping over them and returning a list of processed text. Check out a full example of how to use SFTTrainer on alpaca dataset [here](https://github.com/huggingface/trl/pull/444#issue-1760952763)
-
-### Packing dataset ([`ConstantLengthDataset`])
-
-[`SFTTrainer`] supports _example packing_, where multiple short examples are packed in the same input sequence to increase training efficiency. This is done with the [`ConstantLengthDataset`] utility class that returns constant length chunks of tokens from a stream of examples. To enable the usage of this dataset class, simply pass `packing=True` to the [`SFTConfig`] constructor.
-
-```python
-...
-sft_config = SFTConfig(packing=True, dataset_text_field="text",)
-
-trainer = SFTTrainer(
-    "facebook/opt-350m",
-    train_dataset=dataset,
-    args=sft_config
-)
-
-trainer.train()
-```
-
-Note that if you use a packed dataset and if you pass `max_steps` in the training arguments you will probably train your models for more than few epochs, depending on the way you have configured the packed dataset and the training protocol. Double check that you know and understand what you are doing.
-If you don't want to pack your `eval_dataset`, you can pass `eval_packing=False` to the `SFTConfig` init method.
-
-#### Customize your prompts using packed dataset
-
-If your dataset has several fields that you want to combine, for example if the dataset has `question` and `answer` fields and you want to combine them, you can pass a formatting function to the trainer that will take care of that. For example:
-
-```python
-def formatting_func(example):
-    text = f"### Question: {example['question']}\n ### Answer: {example['answer']}"
-    return text
-
-sft_config = SFTConfig(packing=True)
-trainer = SFTTrainer(
-    "facebook/opt-350m",
-    train_dataset=dataset,
-    args=sft_config,
-    formatting_func=formatting_func
-)
-
-trainer.train()
-```
-You can also customize the [`ConstantLengthDataset`] much more by directly passing the arguments to the [`SFTConfig`] constructor. Please refer to that class' signature for more information.
-
-### Control over the pretrained model
-
-You can directly pass the kwargs of the `from_pretrained()` method to the [`SFTConfig`]. For example, if you want to load a model in a different precision, analogous to
-
-```python
-model = AutoModelForCausalLM.from_pretrained("facebook/opt-350m", torch_dtype=torch.bfloat16)
-
-...
-
-sft_config = SFTConfig(
-    model_init_kwargs={
-        "torch_dtype": "bfloat16",
-    },
-    output_dir="/tmp",
-)
-trainer = SFTTrainer(
-    "facebook/opt-350m",
-    train_dataset=dataset,
-    args=sft_config,
-)
-
-trainer.train()
-```
-Note that all keyword arguments of `from_pretrained()` are supported.
-
-### Training adapters
-
-We also support tight integration with 🤗 PEFT library so that any user can conveniently train adapters and share them on the Hub instead of training the entire model
-
-```python
-from datasets import load_dataset
-from trl import SFTConfig, SFTTrainer
-from peft import LoraConfig
-
-dataset = load_dataset("imdb", split="train")
-
-peft_config = LoraConfig(
-    r=16,
-    lora_alpha=32,
-    lora_dropout=0.05,
-    bias="none",
-    task_type="CAUSAL_LM",
-)
-
-trainer = SFTTrainer(
-    "EleutherAI/gpt-neo-125m",
-    train_dataset=dataset,
-    args=SFTConfig(output_dir="/tmp"),
-    peft_config=peft_config
-)
-
-trainer.train()
-```
-
-You can also continue training your `PeftModel`. For that, first load a `PeftModel` outside `SFTTrainer` and pass it directly to the trainer without the `peft_config` argument being passed.
-
-### Training adapters with base 8 bit models
-
-For that, you need to first load your 8 bit model outside the Trainer and pass a `PeftConfig` to the trainer. For example:
-
-```python
-...
-
-peft_config = LoraConfig(
-    r=16,
-    lora_alpha=32,
-    lora_dropout=0.05,
-    bias="none",
-    task_type="CAUSAL_LM",
-)
-
-model = AutoModelForCausalLM.from_pretrained(
-    "EleutherAI/gpt-neo-125m",
-    load_in_8bit=True,
-    device_map="auto",
-)
-
-trainer = SFTTrainer(
-    model,
-    train_dataset=dataset,
-    args=SFTConfig(),
-    peft_config=peft_config,
-)
-
-trainer.train()
-```
-
-## Using Flash Attention and Flash Attention 2
-
-You can benefit from Flash Attention 1 & 2 using SFTTrainer out of the box with minimal changes of code.
-First, to make sure you have all the latest features from transformers, install transformers from source
-
-```bash
-pip install -U git+https://github.com/huggingface/transformers.git
-```
-
-Note that Flash Attention only works on GPU now and under half-precision regime (when using adapters, base model loaded in half-precision)
-Note also both features are perfectly compatible with other tools such as quantization.
-
-### Using Flash-Attention 1
-
-For Flash Attention 1 you can use the `BetterTransformer` API and force-dispatch the API to use Flash Attention kernel. First, install the latest optimum package:
-
-```bash
-pip install -U optimum
-```
-
-Once you have loaded your model, wrap the `trainer.train()` call under the `with torch.backends.cuda.sdp_kernel(enable_flash=True, enable_math=False, enable_mem_efficient=False):` context manager:
-
-```diff
-...
-
-+ with torch.backends.cuda.sdp_kernel(enable_flash=True, enable_math=False, enable_mem_efficient=False):
-    trainer.train()
-```
-
-Note that you cannot train your model using Flash Attention 1 on an arbitrary dataset as `torch.scaled_dot_product_attention` does not support training with padding tokens if you use Flash Attention kernels. Therefore you can only use that feature with `packing=True`. If your dataset contains padding tokens, consider switching to Flash Attention 2 integration.
-
-Below are some numbers you can get in terms of speedup and memory efficiency, using Flash Attention 1, on a single NVIDIA-T4 16GB.
-
-| use_flash_attn_1 | model_name        | max_seq_len | batch_size | time per training step |
-| ---------------- | ----------------- | ----------- | ---------- | ---------------------- |
-| x                | facebook/opt-350m | 2048        | 8          | ~59.1s                 |
-|                  | facebook/opt-350m | 2048        | 8          | **OOM**                |
-| x                | facebook/opt-350m | 2048        | 4          | ~30.3s                 |
-|                  | facebook/opt-350m | 2048        | 4          | ~148.9s                |
-
-### Using Flash Attention-2
-
-To use Flash Attention 2, first install the latest `flash-attn` package:
-
-```bash
-pip install -U flash-attn
-```
-
-And add `attn_implementation="flash_attention_2"` when calling `from_pretrained`:
-
-```python
-model = AutoModelForCausalLM.from_pretrained(
-    model_id,
-    load_in_4bit=True,
-    attn_implementation="flash_attention_2"
-)
-```
-
-If you don't use quantization, make sure your model is loaded in half-precision and dispatch your model on a supported GPU device.
-After loading your model, you can either train it as it is, or attach adapters and train adapters on it in case your model is quantized.
-
-In contrast to Flash Attention 1, the integration makes it possible to train your model on an arbitrary dataset that also includes padding tokens.
-
-
-### Using model creation utility
-
-We included a utility function to create your model.
-
-[[autodoc]] ModelConfig
-
-```python
-from trl import ModelConfig, SFTTrainer, get_kbit_device_map, get_peft_config, get_quantization_config
-model_config = ModelConfig(
-    model_name_or_path="facebook/opt-350m"
-    attn_implementation=None, # or "flash_attention_2"
-)
-torch_dtype = (
-    model_config.torch_dtype
-    if model_config.torch_dtype in ["auto", None]
-    else getattr(torch, model_config.torch_dtype)
-)
-quantization_config = get_quantization_config(model_config)
-model_kwargs = dict(
-    revision=model_config.model_revision,
-    trust_remote_code=model_config.trust_remote_code,
-    attn_implementation=model_config.attn_implementation,
-    torch_dtype=torch_dtype,
-    use_cache=False if training_args.gradient_checkpointing else True,
-    device_map=get_kbit_device_map() if quantization_config is not None else None,
-    quantization_config=quantization_config,
-)
-model = AutoModelForCausalLM.from_pretrained(model_config.model_name_or_path, **model_kwargs)
-trainer = SFTTrainer(
-    ...,
-    model=model_config.model_name_or_path,
-    peft_config=get_peft_config(model_config),
-)
-```
-
-### Enhance the model's performances using NEFTune
-
-NEFTune is a technique to boost the performance of chat models and was introduced by the paper ["NEFTune: Noisy Embeddings Improve Instruction Finetuning"](https://huggingface.co/papers/2310.05914) from Jain et al. it consists of adding noise to the embedding vectors during training. According to the abstract of the paper:
-
->  Standard finetuning of LLaMA-2-7B using Alpaca achieves 29.79% on AlpacaEval, which rises to 64.69% using noisy embeddings. NEFTune also improves over strong baselines on modern instruction datasets. Models trained with Evol-Instruct see a 10% improvement, with ShareGPT an 8% improvement, and with OpenPlatypus an 8% improvement. Even powerful models further refined with RLHF such as LLaMA-2-Chat benefit from additional training with NEFTune.
-
-<div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/neft-screenshot.png">
-</div>
-
-To use it in `SFTTrainer` simply pass `neftune_noise_alpha` when creating your `SFTConfig` instance. Note that to avoid any surprising behaviour, NEFTune is disabled after training to retrieve back the original behaviour of the embedding layer.
-
-```python
-from datasets import load_dataset
-from trl import SFTConfig, SFTTrainer
-
-dataset = load_dataset("imdb", split="train")
-
-sft_config = SFTConfig(
-    neftune_noise_alpha=5,
-)
-trainer = SFTTrainer(
-    "facebook/opt-350m",
-    train_dataset=dataset,
-    args=sft_config,
-)
-trainer.train()
-```
-
-We have tested NEFTune by training `mistralai/Mistral-7B-v0.1` on the [OpenAssistant dataset](https://huggingface.co/datasets/timdettmers/openassistant-guanaco) and validated that using NEFTune led to a performance boost of ~25% on MT Bench.
-
-<div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/trl-neftune-mistral-7b.png">
-</div>
-
-Note however, that the amount of performance gain is _dataset dependent_ and in particular, applying NEFTune on synthetic datasets like [UltraChat](https://huggingface.co/datasets/stingning/ultrachat) typically produces smaller gains.
-
-### Accelerate fine-tuning 2x using `unsloth`
-
-You can further accelerate QLoRA / LoRA (2x faster, 60% less memory) using the [`unsloth`](https://github.com/unslothai/unsloth) library that is fully compatible with `SFTTrainer`. Currently `unsloth` supports only Llama (Yi, TinyLlama, Qwen, Deepseek etc) and Mistral architectures. Some benchmarks on 1x A100 listed below:
-
-| 1 A100 40GB     | Dataset   | 🤗   | 🤗 + Flash Attention 2 | 🦥 Unsloth | 🦥 VRAM saved |
-| --------------- | --------- | --- | --------------------- | --------- | ------------ |
-| Code Llama 34b  | Slim Orca | 1x  | 1.01x                 | **1.94x** | -22.7%       |
-| Llama-2 7b      | Slim Orca | 1x  | 0.96x                 | **1.87x** | -39.3%       |
-| Mistral 7b      | Slim Orca | 1x  | 1.17x                 | **1.88x** | -65.9%       |
-| Tiny Llama 1.1b | Alpaca    | 1x  | 1.55x                 | **2.74x** | -57.8%       |
-
-First install `unsloth` according to the [official documentation](https://github.com/unslothai/unsloth). Once installed, you can incorporate unsloth into your workflow in a very simple manner; instead of loading `AutoModelForCausalLM`, you just need to load a `FastLanguageModel` as follows:
-
-```python
-import torch
-from trl import SFTConfig, SFTTrainer
-from unsloth import FastLanguageModel
-
-max_seq_length = 2048 # Supports automatic RoPE Scaling, so choose any number
-
-# Load model
-model, tokenizer = FastLanguageModel.from_pretrained(
-    model_name="unsloth/mistral-7b",
-    max_seq_length=max_seq_length,
-    dtype=None,  # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
-    load_in_4bit=True,  # Use 4bit quantization to reduce memory usage. Can be False
-    # token = "hf_...", # use one if using gated models like meta-llama/Llama-2-7b-hf
-)
-
-# Do model patching and add fast LoRA weights
-model = FastLanguageModel.get_peft_model(
-    model,
-    r=16,
-    target_modules=[
-        "q_proj",
-        "k_proj",
-        "v_proj",
-        "o_proj",
-        "gate_proj",
-        "up_proj",
-        "down_proj",
-    ],
-    lora_alpha=16,
-    lora_dropout=0,  # Dropout = 0 is currently optimized
-    bias="none",  # Bias = "none" is currently optimized
-    use_gradient_checkpointing=True,
-    random_state=3407,
-)
-
-args = SFTConfig(
-    output_dir="./output",
-    max_seq_length=max_seq_length,
-    dataset_text_field="text",
-)
-
-trainer = SFTTrainer(
-    model=model,
-    args=args,
-    train_dataset=dataset,
-)
-trainer.train()
-```
-
-The saved model is fully compatible with Hugging Face's transformers library. Learn more about unsloth in their [official repository](https://github.com/unslothai/unsloth).
-
-## Best practices
-
-Pay attention to the following best practices when training a model with that trainer:
-
-- [`SFTTrainer`] always pads by default the sequences to the `max_seq_length` argument of the [`SFTTrainer`]. If none is passed, the trainer will retrieve that value from the tokenizer. Some tokenizers do not provide a default value, so there is a check to retrieve the minimum between 2048 and that value. Make sure to check it before training.
-- For training adapters in 8bit, you might need to tweak the arguments of the `prepare_model_for_kbit_training` method from PEFT, hence we advise users to use `prepare_in_int8_kwargs` field, or create the `PeftModel` outside the [`SFTTrainer`] and pass it.
-- For a more memory-efficient training using adapters, you can load the base model in 8bit, for that simply add `load_in_8bit` argument when creating the [`SFTTrainer`], or create a base model in 8bit outside the trainer and pass it.
-- If you create a model outside the trainer, make sure to not pass to the trainer any additional keyword arguments that are relative to `from_pretrained()` method.
-
-## Multi-GPU Training
-
-Trainer (and thus SFTTrainer) supports multi-GPU training. If you run your script with `python script.py` it will default to using DP as the strategy, which may be [slower than expected](https://github.com/huggingface/trl/issues/1303). To use DDP (which is generally recommended, see [here](https://huggingface.co/docs/transformers/en/perf_train_gpu_many?select-gpu=Accelerate#data-parallelism) for more info) you must launch the script with `python -m torch.distributed.launch script.py` or `accelerate launch script.py`. For DDP to work you must also check the following:
-- If you're using gradient_checkpointing, add the following to the TrainingArguments: `gradient_checkpointing_kwargs={'use_reentrant':False}` (more info [here](https://github.com/huggingface/transformers/issues/26969)
-- Ensure that the model is placed on the correct device:
-```python
-from accelerate import PartialState
-device_string = PartialState().process_index
-model = AutoModelForCausalLM.from_pretrained(
-     ...
-    device_map={'':device_string}
-)
-```
-
-## GPTQ Conversion
-
-You may experience some issues with GPTQ Quantization after completing training. Lowering `gradient_accumulation_steps` to `4` will resolve most issues during the quantization process to GPTQ format.
-
-## Extending `SFTTrainer` for Vision Language Models
-
-`SFTTrainer` does not inherently support vision-language data. However, we provide a guide on how to tweak the trainer to support vision-language data. Specifically, you need to use a custom data collator that is compatible with vision-language data. This guide outlines the steps to make these adjustments. For a concrete example, refer to the script [`examples/scripts/vsft_llava.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/vsft_llava.py) which demonstrates how to fine-tune the LLaVA 1.5 model on the [HuggingFaceH4/llava-instruct-mix-vsft](https://huggingface.co/datasets/HuggingFaceH4/llava-instruct-mix-vsft) dataset.
-
-### Preparing the Data
-
-The data format is flexible, provided it is compatible with the custom collator that we will define later. A common approach is to use conversational data. Given that the data includes both text and images, the format needs to be adjusted accordingly. Below is an example of a conversational data format involving both text and images:
-
-```python
-images = ["obama.png"]
-messages = [
-    {
-        "role": "user",
-        "content": [
-            {"type": "text", "text": "Who is this?"},
-            {"type": "image"}
-        ]
-    },
-    {
-        "role": "assistant",
-        "content": [
-            {"type": "text", "text": "Barack Obama"}
-        ]
-    },
-    {
-        "role": "user",
-        "content": [
-            {"type": "text", "text": "What is he famous for?"}
-        ]
-    },
-    {
-        "role": "assistant",
-        "content": [
-            {"type": "text", "text": "He is the 44th President of the United States."}
-        ]
-    }
-]
-```
-
-To illustrate how this data format will be processed using the LLaVA model, you can use the following code:
-
-```python
-from transformers import AutoProcessor
-
-processor = AutoProcessor.from_pretrained("llava-hf/llava-1.5-7b-hf")
-print(processor.apply_chat_template(messages, tokenize=False))
-```
-
-The output will be formatted as follows:
-
-```txt
-Who is this? ASSISTANT: Barack Obama USER: What is he famous for? ASSISTANT: He is the 44th President of the United States. 
-```
-
-<iframe src="https://huggingface.co/datasets/HuggingFaceH4/llava-instruct-mix-vsft/embed/viewer/default/train" frameborder="0" width="100%" height="560px"></iframe>
-
-
-### A custom collator for processing multi-modal data
-
-Unlike the default behavior of `SFTTrainer`, processing multi-modal data is done on the fly during the data collation process. To do this, you need to define a custom collator that processes both the text and images. This collator must take a list of examples as input (see the previous section for an example of the data format) and return a batch of processed data. Below is an example of such a collator:
-
-```python
-def collate_fn(examples):
-    # Get the texts and images, and apply the chat template
-    texts = [processor.apply_chat_template(example["messages"], tokenize=False) for example in examples]
-    images = [example["images"][0] for example in examples]
-
-    # Tokenize the texts and process the images
-    batch = processor(texts, images, return_tensors="pt", padding=True)
-
-    # The labels are the input_ids, and we mask the padding tokens in the loss computation
-    labels = batch["input_ids"].clone()
-    labels[labels == processor.tokenizer.pad_token_id] = -100
-    batch["labels"] = labels
-
-    return batch
-```
-
-We can verify that the collator works as expected by running the following code:
-
-```python
-from datasets import load_dataset
-
-dataset = load_dataset("HuggingFaceH4/llava-instruct-mix-vsft", split="train")
-examples = [dataset[0], dataset[1]]  # Just two examples for the sake of the example
-collated_data = collate_fn(examples)
-print(collated_data.keys())  # dict_keys(['input_ids', 'attention_mask', 'pixel_values', 'labels'])
-```
-
-### Training the vision-language model
-
-Now that we have prepared the data and defined the collator, we can proceed with training the model. To ensure that the data is not processed as text-only, we need to set a couple of arguments in the `SFTConfig`, specifically `dataset_text_field` and `remove_unused_columns`. We also need to set `skip_prepare_dataset` to `True` to avoid the default processing of the dataset. Below is an example of how to set up the `SFTTrainer`.
-
-```python
-args.dataset_text_field = ""  # needs a dummy field
-args.remove_unused_columns = False
-args.dataset_kwargs = {"skip_prepare_dataset": True}
-
-trainer = SFTTrainer(
-    model=model,
-    args=args,
-    data_collator=collate_fn,
-    train_dataset=train_dataset,
-    tokenizer=processor.tokenizer,
-)
-```
-
-A full example of training LLaVa 1.5 on the [HuggingFaceH4/llava-instruct-mix-vsft](https://huggingface.co/datasets/HuggingFaceH4/llava-instruct-mix-vsft) dataset can be found in the script [`examples/scripts/vsft_llava.py`](https://github.com/huggingface/trl/blob/main/examples/scripts/vsft_llava.py).
-
-- [Experiment tracking](https://wandb.ai/huggingface/trl/runs/2b2c5l7s)
-- [Trained model](https://huggingface.co/HuggingFaceH4/sft-llava-1.5-7b-hf)
-
-## SFTTrainer
-
-[[autodoc]] SFTTrainer
-
-## SFTConfig
-
-[[autodoc]] SFTConfig
-
-## Datasets
-
-In the SFTTrainer we smartly support `datasets.IterableDataset` in addition to other style datasets. This is useful if you are using large corpora that you do not want to save all to disk. The data will be tokenized and processed on the fly, even when packing is enabled.
-
-Additionally, in the SFTTrainer, we support pre-tokenized datasets if they are `datasets.Dataset` or `datasets.IterableDataset`. In other words, if such a dataset has a column of `input_ids`, no further processing (tokenization or packing) will be done, and the dataset will be used as-is. This can be useful if you have pretokenized your dataset outside of this script and want to re-use it directly.
-
-### ConstantLengthDataset
-
-[[autodoc]] trainer.ConstantLengthDataset

docs/source/speeding_up_training.md

@@ -0,0 +1,73 @@
+# Speeding Up Training
+
+<Tip warning={true}>
+
+Section under construction. Feel free to contribute!
+
+</Tip>
+
+## vLLM for fast generation in online methods
+
+Online methods such as GRPO or Online DPO require the model to generate completions, which is often a slow process and can significantly impact training time.
+To speed up generation, you can use [vLLM](https://github.com/vllm-project/vllm), a library that enables fast generation through, among other things, PagedAttention. TRL's online trainers support vLLM, greatly improving training speed.
+
+To use [vLLM](https://github.com/vllm-project/vllm), first install it using:
+
+```bash
+pip install vllm
+```
+
+or 
+
+```bash
+pip install "trl[vllm]"
+```
+
+<hfoptions id="vllm examples">
+<hfoption id="Online DPO">
+
+Then, enable it by passing `use_vllm=True` in the training arguments.
+
+```python
+from trl import OnlineDPOConfig
+
+training_args = OnlineDPOConfig(..., use_vllm=True)
+```
+
+</hfoption>
+<hfoption id="GRPO">
+
+First, start a vLLM server by running:
+
+```bash
+trl vllm-serve --model <model_name>
+```
+
+Then, run the training script and pass `use_vllm=True` in the training arguments.
+
+```python
+from trl import GRPOConfig
+
+training_args = GRPOConfig(..., use_vllm=True)
+```
+
+You can customize the server configuration by passing additional arguments. For more information, see [vLLM integration](vllm_integration).
+
+<Tip warning={true}>
+
+When using vLLM, ensure that the GPUs assigned for training and generation are separate to avoid resource conflicts. For instance, if you plan to use 4 GPUs for training and another 4 for vLLM generation, you can specify GPU allocation using `CUDA_VISIBLE_DEVICES`.  
+
+Set GPUs **0-3** for vLLM generation:  
+```sh
+CUDA_VISIBLE_DEVICES=0,1,2,3 trl vllm-serve --model <model_name>
+```  
+
+And GPUs **4-7** for training:  
+```sh
+CUDA_VISIBLE_DEVICES=4,5,6,7 accelerate launch train.py
+```  
+
+</Tip>
+
+</hfoption>
+</hfoptions>

docs/source/text_environments.md

@@ -1,197 +0,0 @@
-# Text Environments
-
-Text environments provide a learning ground for language agents. It allows a language model to use tools to accomplish a task such as using a Python interpreter to answer math questions or using a search index for trivia questions. Having access to tools allows language models to solve tasks that would be very hard for the models itself but can be trivial for the appropriate tools. A good example is arithmetics of large numbers that become a simple copy-paste task once you have access to a calculator.
-
-<div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/textenv.png">
-</div>
-
-Let's dive into how text environments work and start with tools!
-
-## Tools
-
-One of the core building blocks of text environments are tools that the model can use to solve tasks. In general tools can be any Python function that takes a string as input and returns string. The `TextEnvironment` offers two options for tools: either go with predefined tools from `transformers.Tool` or define your own function or class with `__call__` method. Let's have a look at both!
-
-### `transformers.Tool`
-
-Text environments fully support tools of the class `transformers.Tool`. The advantage of building tools in that framework is that they can easily be shared 
-
-```Python
-from transformers import load_tool
-
-# simple calculator tool that runs +-/* operations
-calc_tool = load_tool("ybelkada/simple-calculator")
-
-# python interpreter that executes program and returns outputs
-py_tool = load_tool("lvwerra/python-interpreter")
-
-# wikipedia search index that returns best search match
-wiki_tool = load_tool("vwxyzjn/pyserini-wikipedia-kilt-doc")
-```
-
-These tools are either loaded from the hub or from a local folder. Using the tool is as simple as calling them with a text query:
-
-```Python
-calc_tool("1/2")
->>> "0.5"
-```
-
-Note that both input and return values are strings to enable easy usage with a language model.
-
-### Custom Tools
-
-The following is an example of a tool that adds two integers:
-
-```Python
-def add(text):
-    int_1, int_2 = text.split("+")
-    result = int(int_1) + int(int_2)
-    return str(result)
-
-print(add("1+1"))
->>> "2"
-```
-
-We looked at basic examples such as a calculator but the principle holds for more complex tools as well such as a web search tool where you input the query and get the search results in return. Now let's look at how the model can use the tools with the call syntax.
-
-### Call syntax
-
-In order to have a unified way for the model to call a tool we created a simple syntax that looks as follows:
-
-```python
-"<request><TOOL_NAME>QUERY<call>TOOL_RESPONSE<response>"
-```
-
-There are a few special tokens involved so let's decompose it: First the model can signal that it wants to use a tool by emitting the `<request>` token. After that we want to know the name of the tool to call which is done by enclosing the tool name with `<>` brackets. Once we know which tool to call the tool query follows which is in free text form. The `<call>` tokens signifies the end of the query and stops the model generation. At this point the model output is parsed and the query sent to the tool. The environment appends the tool response to the string followed by the `<response>` token to show the end the tool output.
-
-Let's look at the concrete example of the calculator and assume its name is `Calculator` (more on how the name of a tool is inferred later):
-
-```python
-"<request><Calculator>1/2<call>0.5<response>"
-```
-
-Finally, the episode is ended and generation stops when the model generates `<submit>` which marks the interaction as completed.
-
-Now let's have a look how we can create a new text environment!
-
-## Create a `TextEnvironment`
-
-
-```python
-prompt = """\
-What is 13-3?
-<request><SimpleCalculatorTool>13-3<call>10.0<response>
-Result=10<submit>
-"""
-
-def reward_fn(result, answer):
-    """Simplified reward function returning 1 if result matches answer and 0 otherwise."""
-    result_parsed = result.split("=")[1].split("<")[0]
-    return int(result_parsed==answer)
-
-text_env = TextEnvironemnt(
-    model=model, 
-    tokenizer=tokenizer,
-    tools= {"SimpleCalculatorTool": load_tool("ybelkada/simple-calculator")},
-    reward_fn=exact_match_reward,
-    prompt=prompt, 
-    max_turns=1
-    max_tool_response=100
-    generation_kwargs={"do_sample": "true"}
-)
-```
-
-Let's decompose the settings:
-
-| Argument           | Description     |
-|:-------------------|:----------------|
-| `model`            | Language model to interact with the environment and generate requests. |
-| `tokenizer`        | Tokenizer of language model handling tokenization of strings. |
-| `tools`            | `list` of `dict` of tools. If former the name of the tool is inferred from class name and otherwise it's the keys of the dictionary.|
-| `reward_fn`        | A function that takes a string as input and returns. Can have extra arguments that are passed to `.run()` such as ground truth.|
-| `prompt`           | Prompt to prepend to every task. Usually a few examples to demonstrate to the model how to use the tools in a few-shot fashion. |
-| `max_turns`        | Maximum number of interactions between model and tools before episode ends.|
-| `max_tool_response`| The tool response is truncated to this number to avoid running out of model context.|
-| `max_length`       |  The maximum number of tokens to allow in an episode. |
-| `generation_kwargs`| Generation settings used by the language model. |
-
-You can customize the environment to your needs and add custom tools and settings. Let's see how you can use the environment to have the model interact with the available tools!
-
-
-## Run an Episode
-
-To run a set of queries through the text environment one can simply use the `run` method.
-
-```python
-queries = ["What is 1/2?"]
-answers = ["0.5"]
-
-queries, responses, masks, rewards, histories = text_env.run(queries, answers=answers)
-```
-
-This will execute the model/tool feedback loop for each query until either no tool is called anymore, the maximum number of turns is reached or to maximum number of tokens in an episode is exceeded. The extra `kwargs` (e.g. `answers=answers` above) passed to `run` will be passed on to the reward function.
-
-There are five objects that are returned by `run`: 
-
-- `queries`: a list of the tokenized queries
-- `responses`: all tokens that have been generated withing the environment including model and tool tokens
-- `masks`: mask that indicates which tokens have been generated by the model and which tokens are generated by the tool
-- `rewards`: a list of reward for each query/response
-- `histories`: list of `TextHistory` objects, which are useful objects containing all the above and also the text equivalents
-
-The masks are crucial for training as we don't want to optimize tokens that the model has not generated which are tokens produced by the tools.
-
-Next, we'll train a PPO step with the generated responses!
-
-
-### Train
-Training on episodes from the `TextEnvironment` is straight forward and simply requires forwarding all the returned variables except the `TextHistory` objects to the `step` method:
-
-```python
-train_stats = ppo_trainer.step(queries, responses, rewards, masks)
-```
-
-## `TextHistory`
-
-The `TextHistory` object stores the interactions between the model and the text environment. It stores tokens and text generated in each turn and their source in each turn (model or system) as well as rewards. Let's go through the class attributes and methods.
-
-### Attributes
-
-The following table summarises the available attributes of the `TextEnvironment` class:
-
-| Attribute           | Description     |
-|:-------------------|:----------------|
-| `text`             | The full string of the text generated in the text environment with both model and system generated text. |
-| `text_spans`       | A list of tuples with the spans for each model or system generated text segment. |
-| `system_spans`     | A list of boolean values indicating if the segment is model or system generated. |
-| `tokens`           | All tokens generated in text environment with both model and system generated tokens. |
-| `token_spans`      | Similar to `text_spans` the `token_spans` indicate the boundaries of model andsystem generated tokens. |
-| `token_masks`      | The token masks can be used to ignore system generated tokens by masking them. |
-| `completed`        | Indicates if the interaction with the environment has completed. |
-| `truncated`        | Indicates if the interaction with the environment has completed because max length was reached. |
-
-With these attributes you can reconstruct every interaction of the model with the `TextEnvironment`. The `TextHistory` also lets you visualize the text history. Let's have a look!
-
-### Visualization
-
-When the model interacts inside the `TextEnvironment` it can be useful to visualize and separate which parts of the text outputs were generated by the model and which parts come from the system and tools. For that purpose there are the two methods [`TextHistory.show_text`] and [`TextHistory.show_tokens`]. They print the text and tokens respectively and highlight the various segments using the [`rich` libray](https://github.com/Textualize/rich) (make sure to install it before using these methods).
-
-You can see that the prompt is highlighted in gray, whereas system segments such as query and tool responses are highlighted in green. All segments generated by the model are highlighted in blue and in addition to the pure text output the reward is displayed as additional text in plum. Here an example of `show_text`:
-
-<div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/textenv_show_text.png" width=600>
-</div>
-
-Sometimes there can be tricky tokenization related issues that are hidden when showing the decoded text. Thus `TextHistory` also offers an option to display the same highlighting on the tokens directly with `show_tokens`:
-
-<div style="text-align: center">
-<img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/textenv_show_tokens.png" width=800>
-</div>
-
-Note that you can turn on the colour legend by passing `show_legend=True`.
-
-## API Documentation
-
-[[autodoc]] TextEnvironment
-
-[[autodoc]] TextHistory

docs/source/trainer.mdx

@@ -1,70 +0,0 @@
-# Trainer
-
-At TRL we support PPO (Proximal Policy Optimisation) with an implementation that largely follows  the structure introduced in the paper "Fine-Tuning Language Models from Human Preferences" by D. Ziegler et al. [[paper](https://huggingface.co/papers/1909.08593), [code](https://github.com/openai/lm-human-preferences)].
-The Trainer and model classes are largely inspired from `transformers.Trainer` and `transformers.AutoModel` classes and adapted for RL.
-We also support a `RewardTrainer` that can be used to train a reward model.
-
-
-## CPOConfig
-
-[[autodoc]] CPOConfig
-
-## CPOTrainer
-
-[[autodoc]] CPOTrainer
-
-## DDPOConfig
-
-[[autodoc]] DDPOConfig
-
-## DDPOTrainer
-
-[[autodoc]] DDPOTrainer
-
-## DPOTrainer
-
-[[autodoc]] DPOTrainer
-
-## IterativeSFTTrainer
-
-[[autodoc]] IterativeSFTTrainer
-
-## KTOConfig
-
-[[autodoc]] KTOConfig
-
-## KTOTrainer
-
-[[autodoc]] KTOTrainer
-
-## ORPOConfig
-
-[[autodoc]] ORPOConfig
-
-## ORPOTrainer
-
-[[autodoc]] ORPOTrainer
-
-## PPOConfig
-
-[[autodoc]] PPOConfig
-
-## PPOTrainer
-
-[[autodoc]] PPOTrainer
-
-## RewardConfig
-
-[[autodoc]] RewardConfig
-
-## RewardTrainer
-
-[[autodoc]] RewardTrainer
-
-## SFTTrainer
-
-[[autodoc]] SFTTrainer
-
-## set_seed
-
-[[autodoc]] set_seed

docs/source/training_vlm_sft.md

@@ -0,0 +1,380 @@
+# Fine-tuning a Multimodal Model Using SFT (Single or Multi-Image Dataset)
+
+![VLM SFT training procedure](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/training_vlm_sft_training_procedure.png)  
+
+## Overview  
+
+This guide walks you through the process of fine-tuning a multimodal language model (e.g., **Gemma 3**) using **Supervised Fine-Tuning (SFT)**. We cover two cases:  
+
+- **Single Image + Text**  
+- **Multi-Image + Text**  
+
+This guide serves as a **detailed walkthrough** and complements the existing [VLM SFT script](https://github.com/huggingface/trl/blob/main/examples/scripts/sft_vlm_gemma3.py). If you're already familiar with the concepts, you can use the script directly.  
+
+We demonstrate the fine-tuning process using two datasets, but these principles extend to other **Vision-Language Models (VLMs)** and datasets.  
+
+## Understanding the Datasets  
+
+To address both **Single Image + Text** and **Multi-Image + Text** scenarios, we use two datasets that are well-suited for this task.  
+
+### HuggingFaceH4/llava-instruct-mix-vsft Dataset (Image + Text)
+
+This dataset is a reformatted version of [LLaVA Instruct Mix](https://huggingface.co/datasets/theblackcat102/llava-instruct-mix). It consists of conversations where a user provides both **text** and a **single image** as input.  
+
+The model (referred to as the **"assistant"**) responds based on both the **visual and textual information** shared by the user. This dataset is particularly useful for training multimodal models to **understand and generate responses based on images and text**.  
+
+<iframe
+  src="https://huggingface.co/datasets/HuggingFaceH4/llava-instruct-mix-vsft/embed/viewer/default/train"
+  frameborder="0"
+  width="100%"
+  height="560px"
+></iframe>
+
+### FanqingM/MMIU-Benchmark Dataset (Multi-Image + Text)
+
+The **FanqingM/MMIU-Benchmark** dataset consists of:  
+
+- **Context:** Included in the system prompt.  
+- **Question:** Provided as part of the user's input.  
+- **Series of Images:** Multiple images related to the question.  
+- **Answer:** The model's expected response.  
+
+This dataset is designed for tasks where the model must reason over multiple images to generate an informed response based on both visual and textual inputs.
+
+<iframe
+  src="https://huggingface.co/datasets/FanqingM/MMIU-Benchmark/embed/viewer/default/test"
+  frameborder="0"
+  width="100%"
+  height="560px"
+></iframe>
+
+## Developing a Fine-Tuning Script for Multimodal SFT
+
+In this section, we build the script needed to fine-tune a multimodal model for both **Single Image + Text** and **Multi-Image + Text** use cases.  
+
+### Setting Up the Environment
+
+Before fine-tuning, we need to install the required dependencies. Let's start by setting up the environment:  
+
+```bash
+# Install the required libraries. Further details: https://huggingface.co/docs/trl/installation 
+pip install -U -q trl bitsandbytes peft hf_xet tensorboard
+```
+
+Once all dependencies are installed, we need to log in to the **Hugging Face Hub**. Since **Gemma 3** is a gated model, access permissions are required.  
+
+If you haven’t requested access yet, visit the [Model Card](https://huggingface.co/google/gemma-3-4b-it) and request it.  
+
+To log in, you’ll need to generate an [access token](https://huggingface.co/settings/tokens) from your Hugging Face account.  
+
+```bash
+huggingface-cli login
+```
+
+### **Loading the Data**
+
+As mentioned earlier, we will cover two possible use cases. While the specific procedure may vary based on the dataset, the core principles remain consistent.  
+
+This guide supports both use cases, so refer to the **Single Image + Text** or **Multi-Image + Text** sections depending on your specific scenario.
+
+#### **Single Image + Text**
+
+![Single Image + Text](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/training_vlm_sft_training_procedure_single_image.png)  
+
+In this case, each sample in a batch consists of a **single image paired with text**. Since the dataset is already formatted for supervised fine-tuning (SFT), we can directly load it using `load_dataset`.
+
+```python
+from datasets import load_dataset
+
+dataset_name = "HuggingFaceH4/llava-instruct-mix-vsft"
+
+# Load Dataset
+dataset = load_dataset(dataset_name)
+```
+
+#### **Multi-Image + Text (or Interleaving)**  
+
+![Multi-Image + Text](https://huggingface.co/datasets/trl-lib/documentation-images/resolve/main/training_vlm_sft_training_procedure_multi_image.png)  
+
+Gemma 3 also supports **Multi-Image + Text** scenarios, where:  
+
+- The model receives a **list of images** alongside a user message.  
+- The model processes **interleaved images and text** within a conversation.  
+
+For this dataset, some preprocessing is required before training.
+
+```python
+from datasets import load_dataset
+
+dataset_name = "FanqingM/MMIU-Benchmark"
+
+# Load Dataset
+dataset = load_dataset(dataset_name)
+```
+
+After loading the dataset, we need to preprocess and format it into a conversational structure. Here’s an example of how the data might look:
+
+```python
+{"role": "system", "content": [{"type": "text", "text": "You are a judge in a photography competition, and now you are given the four images. Please examine the details and tell which one of them is most likely to be a real photograph.\nSelect from the following choices.\nA: the first image\nB: the second image\nC: the third image\nD: the fourth image"}]},
+{"role": "user", "content": images_list + [{"type": "text", "text": "Which image is most likely to be a real photograph?"}]},
+{"role": "assistant", "content": [{"type": "text", "text": "A: the first image\nB: the second image\nC: the third image\nD: the fourth image"}]},
+```
+
+Here, `images_list` is a list of images:
+
+```python
+images_list = [
+  {"type": "image", "image": <class 'PIL.Image.Image'>},
+  {"type": "image", "image": <class 'PIL.Image.Image'>},
+  {"type": "image", "image": <class 'PIL.Image.Image'>},
+  {"type": "image", "image": <class 'PIL.Image.Image'>},
+  {"type": "image", "image": <class 'PIL.Image.Image'>},
+]
+```
+
+This structure can be translated into code like this:
+
+```python
+import os
+import zipfile
+import io
+from datasets import DatasetDict
+from huggingface_hub import hf_hub_download, list_repo_files
+from PIL import Image
+
+dataset_train_split = "test"
+
+def format_data(samples: dict[str, any]) -> dict[str, list]:
+    formatted_samples = {"messages": []}
+    for cont in range(len(samples["question"])):
+        images = []
+        for img_path in samples["input_image_path"][cont]:
+            try:
+                with open(img_path, "rb") as f:
+                    img_bytes = f.read()
+                image = Image.open(io.BytesIO(img_bytes)).convert("RGB")
+                images.append({"type": "image", "image": image})
+            except Exception as e:
+                print(f"Error processing image {img_path}: {e}")
+                continue
+
+        formatted_samples["messages"].append(
+            [
+                {"role": "system", "content": [{"type": "text", "text": samples["context"][cont]}]},
+                {"role": "user", "content": images + [{"type": "text", "text": samples["question"][cont]}]},
+                {"role": "assistant", "content": [{"type": "text", "text": samples["output"][cont]}]},
+            ]
+        )
+    return formatted_samples
+
+# For multi-image example
+def prepare_dataset(dataset: DatasetDict, dataset_name: str, dataset_train_split: str) -> DatasetDict:
+    all_files = list_repo_files(dataset_name, repo_type="dataset")
+    zip_files = [f for f in all_files if f.endswith(".zip")]
+
+    for zip_filename in zip_files:
+        zip_path = hf_hub_download(repo_id=dataset_name, filename=zip_filename, repo_type="dataset")
+        extract_folder = zip_filename.replace(".zip", "")
+        os.makedirs(extract_folder, exist_ok=True)
+
+        with zipfile.ZipFile(zip_path, "r") as zip_ref:
+            zip_ref.extractall(extract_folder)
+
+    dataset = dataset.map(format_data, batched=True, batch_size=4, num_proc=16)
+    return dataset
+
+dataset = prepare_dataset(dataset, dataset_name, dataset_train_split)
+```
+
+With this, your **Multi-Image + Text** dataset is now prepared for training.
+
+### **Preparing for Training**  
+
+We start by loading the model and processor. In this example, we use `google/gemma-3-4b-it`, but the same process applies to its other variants and similar models.  
+
+To optimize memory usage, we configure `BitsAndBytes` to load the quantized version of the model.
+
+```python
+import torch
+from transformers import AutoModelForImageTextToText, AutoProcessor, BitsAndBytesConfig
+
+model_id = "google/gemma-3-4b-it"
+
+# BitsAndBytesConfig int-4 config
+bnb_config = BitsAndBytesConfig(
+    load_in_4bit=True,
+    bnb_4bit_use_double_quant=True,
+    bnb_4bit_quant_type="nf4",
+    bnb_4bit_compute_dtype=torch.bfloat16,
+    bnb_4bit_quant_storage=torch.bfloat16,
+)
+
+# Load model and tokenizer
+model = AutoModelForImageTextToText.from_pretrained(
+    model_id, 
+    device_map="auto", 
+    torch_dtype=torch.bfloat16,
+    attn_implementation="eager", # Important (Ref: https://github.com/huggingface/transformers/blob/c15a7adb283fa984a40558c7fe7bed30ae975cdd/src/transformers/models/gemma3/modeling_gemma3.py#L934)
+    quantization_config=bnb_config
+)
+processor = AutoProcessor.from_pretrained(model_id)
+processor.tokenizer.padding_side = "right"
+```
+
+Next, we set up [Quantized Low-Rank Adaptation (QLoRA)](https://huggingface.co/papers/2305.14314), an efficient fine-tuning technique for Large Language Models (LLMs) and Vision-Language Models (VLMs).  
+
+```python
+from peft import LoraConfig, get_peft_model
+
+# Configure QLoRA
+peft_config = LoraConfig(
+    lora_alpha=16,
+    lora_dropout=0.05,
+    r=16,
+    bias="none",
+    target_modules="all-linear",
+    task_type="CAUSAL_LM",
+    modules_to_save=[
+        "lm_head",
+        "embed_tokens",
+    ],
+)
+```
+
+With QLoRA now set up, we need to define the training arguments for SFT. The [`SFTConfig`] class simplifies this process, providing an easy way to adjust parameters based on our specific needs.  
+
+```python
+from trl import SFTConfig
+
+training_args = SFTConfig(
+    output_dir="gemma-3-4b-it-trl-sft-llava-instruct-mix-vsft",     # Directory to save the model and push to the Hub. Use a specific repository id (e.g., gemma-3-4b-it-trl-sft-MMIU-Benchmark for multi-image datasets).
+    num_train_epochs=1,                                             # Set the number of epochs to train the model.
+    per_device_train_batch_size=8,                                  # Batch size for each device (e.g., GPU) during training. multi-image -> per_device_train_batch_size=1
+    gradient_accumulation_steps=4,                                  # Number of steps before performing a backward/update pass to accumulate gradients. multi-image -> gradient_accumulation_steps=1
+    gradient_checkpointing=True,                                    # Enable gradient checkpointing to reduce memory usage during training.
+    optim="adamw_torch_fused",                                      # Use the fused AdamW optimizer for better performance.
+    save_strategy="epoch",                                          # Save checkpoints at the end of each epoch.
+    learning_rate=2e-05,                                            # Learning rate for training.
+    bf16=True,                                                      # Enable bfloat16 precision for training to save memory and speed up computations.
+    push_to_hub=True,                                               # Automatically push the fine-tuned model to Hugging Face Hub after training.
+    report_to="tensorboard",                                        # Automatically report metrics to tensorboard.
+    gradient_checkpointing_kwargs={"use_reentrant": False},         # Set gradient checkpointing to non-reentrant to avoid issues.
+    dataset_kwargs={"skip_prepare_dataset": True},                  # Skip dataset preparation to handle preprocessing manually.
+    remove_unused_columns=False,                                    # Ensure unused columns are not removed in the collator (important for batch processing).
+)
+```
+
+The `collate_fn` is responsible for processing and preparing individual examples to form a batch.  
+
+Each example in the batch undergoes the following steps:  
+
+1. The **chat template** is applied to the text.  
+2. The **processor tokenizes** both `texts` and `images`, encoding them into tensors.  
+3. The **labels** for training are set as the `input_ids` of the example.  
+4. Certain **special tokens** are **masked (ignored)** during loss computation:  
+   - `pad_token_id`  
+   - `<image_token_id>`  
+   - `<image_soft_token>` (corresponding to ID `262144`)  
+
+This process is similar across different dataset types, with a minor variation in how images are handled:  
+
+- **Single Image + Text** → A **list of images** is directly processed.  
+- **Multi-Image + Text** → A **list of lists of images** is used, where each batch element contains multiple images.  
+
+```python
+from PIL import Image
+
+# For multi-image cases
+def process_vision_info(messages: list[dict]) -> list[Image.Image]:
+    image_inputs = []
+    for msg in messages:
+        content = msg.get("content", [])
+        if not isinstance(content, list):
+            content = [content]
+
+        for element in content:
+            if isinstance(element, dict) and ("image" in element or element.get("type") == "image"):
+                if "image" in element:
+                    image = element["image"]
+                else:
+                    image = element
+                if image is not None:
+                    image = Image.open(io.BytesIO(image["bytes"]))
+                    image_inputs.append(image.convert("RGB"))
+    return image_inputs
+
+def collate_fn(examples):
+    texts = [processor.apply_chat_template(example["messages"], tokenize=False, add_generation_prompt=False).strip() for example in examples]
+    if "images" in examples[0]:  # single-image
+        images = [
+            [img.convert("RGB") for img in example["images"]]
+            for example in examples
+        ]
+    else:  # multi-image
+        images = [process_vision_info(example["messages"]) for example in examples]
+
+    # Tokenize the texts and process the images
+    batch = processor(
+        images=images, text=texts, return_tensors="pt", padding=True
+    )  # Encode texts and images into tensors
+
+    # The labels are the input_ids, and we mask the padding tokens in the loss computation
+    labels = batch["input_ids"].clone()  # Clone input IDs for labels
+    # Mask image tokens
+    image_token_id = [
+        processor.tokenizer.convert_tokens_to_ids(processor.tokenizer.special_tokens_map["boi_token"])
+    ]
+    # Mask tokens for not being used in the loss computation
+    labels[labels == processor.tokenizer.pad_token_id] = -100
+    labels[labels == image_token_id] = -100
+    labels[labels == 262144] = -100
+
+    batch["labels"] = labels
+    return batch  # Return the prepared batch
+```
+
+### **Training the Model**  
+
+With all the components set up, we now configure the `SFTTrainer` using the previously defined settings and start the training process.
+
+``` python
+# Training
+from trl import SFTTrainer
+
+trainer = SFTTrainer(
+    model=model,
+    args=training_args,
+    data_collator=collate_fn,
+    train_dataset=dataset["train"], # multi-image -> train_dataset=dataset["test"],
+    processing_class=processor,
+    peft_config=peft_config,
+)
+
+trainer.train()
+
+# Save the final model
+trainer.save_model()
+```
+
+We save the fine-tuned model to the Hub, making it easily accessible for future use. Additionally, TRL automatically logs the training results to **Weights & Biases (Wandb)** or **TensorBoard**, depending on the chosen configuration.  
+
+<!-- Add Wandb training results -->
+### Results
+
+During and after training, we can inspect the results using **Weights & Biases (Wandb)** or **TensorBoard**. For example:
+
+* [**gemma-3-4b-it-trl-sft-llava-instruct-mix-vsft (Single Image+Text)**](https://huggingface.co/sergiopaniego/gemma-3-4b-it-trl-sft-llava-instruct-mix-vsft)
+
+* [**gemma-3-4b-it-trl-sft-MMIU-Benchmark (Multi-Images+Text or Interleaving)**](https://huggingface.co/sergiopaniego/gemma-3-4b-it-trl-sft-MMIU-Benchmark)
+
+## Limitations  
+
+Currently, fine-tuning Gemma has some [known limitations](https://github.com/huggingface/trl/issues/3121). We recommend following the procedure outlined in this guide to ensure the best results.  
+
+## References  
+
+For further reading and complementary resources, check out the following:  
+
+- [Fine-Tuning Vision-Language Models with QLoRA](https://ai.google.dev/gemma/docs/core/huggingface_vision_finetune_qlora)  
+- [Fine-Tuning a Vision Language Model (Qwen2-VL-7B) with the Hugging Face Ecosystem (TRL)](https://huggingface.co/learn/cookbook/fine_tuning_vlm_trl)  
+

docs/source/unsloth_integration.md

@@ -0,0 +1,66 @@
+# Unsloth Integration
+
+<Tip warning={true}>
+
+Section under construction. Feel free to contribute!
+
+</Tip>
+
+Unsloth is an open‑source framework for fine‑tuning and reinforcement learning that trains LLMs (like Llama, Mistral, Gemma, DeepSeek, and more) up to 2× faster with up to 70% less VRAM, while providing a streamlined, Hugging Face–compatible workflow for training, evaluation, and deployment.
+Unsloth library that is fully compatible with [`SFTTrainer`]. Some benchmarks on 1 x A100 listed below:
+
+| 1 A100 40GB     | Dataset   | 🤗   | 🤗 + Flash Attention 2 | 🦥 Unsloth | 🦥 VRAM saved |
+| --------------- | --------- | --- | --------------------- | --------- | ------------ |
+| Code Llama 34b  | Slim Orca | 1x  | 1.01x                 | **1.94x** | -22.7%       |
+| Llama-2 7b      | Slim Orca | 1x  | 0.96x                 | **1.87x** | -39.3%       |
+| Mistral 7b      | Slim Orca | 1x  | 1.17x                 | **1.88x** | -65.9%       |
+| Tiny Llama 1.1b | Alpaca    | 1x  | 1.55x                 | **2.74x** | -57.8%       |
+
+First, install `unsloth` according to the [official documentation](https://github.com/unslothai/unsloth). Once installed, you can incorporate unsloth into your workflow in a very simple manner; instead of loading [`~transformers.AutoModelForCausalLM`], you just need to load a `FastLanguageModel` as follows:
+
+```python
+import torch
+from trl import SFTConfig, SFTTrainer
+from unsloth import FastLanguageModel
+
+max_length = 2048 # Supports automatic RoPE Scaling, so choose any number
+
+# Load model
+model, tokenizer = FastLanguageModel.from_pretrained(
+    model_name="unsloth/mistral-7b",
+    max_seq_length=max_length,
+    dtype=None,  # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
+    load_in_4bit=True,  # Use 4bit quantization to reduce memory usage. Can be False
+)
+
+# Do model patching and add fast LoRA weights
+model = FastLanguageModel.get_peft_model(
+    model,
+    r=16,
+    target_modules=[
+        "q_proj",
+        "k_proj",
+        "v_proj",
+        "o_proj",
+        "gate_proj",
+        "up_proj",
+        "down_proj",
+    ],
+    lora_alpha=16,
+    lora_dropout=0,  # Dropout = 0 is currently optimized
+    bias="none",  # Bias = "none" is currently optimized
+    use_gradient_checkpointing=True,
+    random_state=3407,
+)
+
+training_args = SFTConfig(output_dir="./output", max_length=max_length)
+
+trainer = SFTTrainer(
+    model=model,
+    args=training_args,
+    train_dataset=dataset,
+)
+trainer.train()
+```
+
+The saved model is fully compatible with Hugging Face's transformers library. Learn more about unsloth in their [official repository](https://github.com/unslothai/unsloth).