fix: correct attention class retrieval for gemma3n model in lora_kernels.py

* Support --lora-on-cpu flag for DPO model merging

* fix: use device=cpu in _convert_embedding_modules_dtype when lora_on_cpu is set

.github/workflows/base.yml

@@ -16,8 +16,9 @@ on:
 jobs:
   build-base:
     if: github.repository_owner == 'axolotl-ai-cloud'
+    timeout-minutes: 480
     # this job needs to be run on self-hosted GPU runners...
-    runs-on: axolotl-gpu-runner
+    runs-on: ubuntu-latest-m
     strategy:
       fail-fast: false
       matrix:
@@ -28,42 +29,50 @@ jobs:
             python_version: "3.11"
             pytorch: 2.5.1
             torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+            dockerfile: "Dockerfile-base"
           - cuda: "124"
             cuda_version: 12.4.1
             cudnn_version: ""
             python_version: "3.11"
             pytorch: 2.6.0
             torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+            dockerfile: "Dockerfile-base"
           - cuda: "126"
             cuda_version: 12.6.3
             cudnn_version: ""
             python_version: "3.11"
             pytorch: 2.6.0
             torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+            dockerfile: "Dockerfile-base"
           - cuda: "126"
             cuda_version: 12.6.3
             cudnn_version: ""
             python_version: "3.11"
-            pytorch: 2.7.0
+            pytorch: 2.7.1
             torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+            dockerfile: "Dockerfile-base"
           - cuda: "128"
             cuda_version: 12.6.3
             cudnn_version: ""
             python_version: "3.11"
-            pytorch: 2.7.0
+            pytorch: 2.7.1
             torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+            dockerfile: "Dockerfile-base"
           - cuda: "128"
             cuda_version: 12.8.1
             cudnn_version: ""
             python_version: "3.11"
             pytorch: nightly
             torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
-          - cuda: "128"
-            cuda_version: 12.8.1
-            cudnn_version: ""
-            python_version: "3.11"
-            pytorch: next
-            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+            dockerfile: "Dockerfile-base-nightly"
+#          # "next" is for release candidates of pytorch
+#          - cuda: "128"
+#            cuda_version: 12.8.1
+#            cudnn_version: ""
+#            python_version: "3.11"
+#            pytorch: next
+#            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+#            dockerfile: "Dockerfile-base-next"
     steps:
       - name: Checkout
         uses: actions/checkout@v4
@@ -85,7 +94,60 @@ jobs:
         uses: docker/build-push-action@v4
         with:
           context: .
-          file: ${{ matrix.pytorch == 'nightly' && './docker/Dockerfile-base-nightly' || matrix.pytorch == 'next' && './docker/Dockerfile-base-next' || './docker/Dockerfile-base' }}
+          file: ./docker/${{ matrix.dockerfile }}
+          push: ${{ github.event_name != 'pull_request' }}
+          tags: ${{ steps.metadata.outputs.tags }}-base-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
+          labels: ${{ steps.metadata.outputs.labels }}
+          build-args: |
+            CUDA_VERSION=${{ matrix.cuda_version }}
+            CUDNN_VERSION=${{ matrix.cudnn_version }}
+            CUDA=${{ matrix.cuda }}
+            PYTHON_VERSION=${{ matrix.python_version }}
+            PYTORCH_VERSION=${{ matrix.pytorch }}
+            TORCH_CUDA_ARCH_LIST=${{ matrix.torch_cuda_arch_list }}
+  build-base-uv:
+    if: github.repository_owner == 'axolotl-ai-cloud'
+    timeout-minutes: 480
+    runs-on: ubuntu-latest-m
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - cuda: "126"
+            cuda_version: 12.6.3
+            cudnn_version: ""
+            python_version: "3.11"
+            pytorch: 2.6.0
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+            dockerfile: "Dockerfile-uv-base"
+          - cuda: "128"
+            cuda_version: 12.8.1
+            cudnn_version: ""
+            python_version: "3.11"
+            pytorch: 2.7.1
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+            dockerfile: "Dockerfile-uv-base"
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Docker metadata
+        id: metadata
+        uses: docker/metadata-action@v5
+        with:
+          images: |
+            axolotlai/axolotl-base-uv
+      - name: Login to Docker Hub
+        uses: docker/login-action@v2
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+      - name: Build
+        uses: docker/build-push-action@v4
+        with:
+          context: .
+          file: ./docker/${{ matrix.dockerfile }}
           push: ${{ github.event_name != 'pull_request' }}
           tags: ${{ steps.metadata.outputs.tags }}-base-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
           labels: ${{ steps.metadata.outputs.labels }}

.github/workflows/docs.yml

@@ -23,7 +23,7 @@ jobs:
         - name: Install dependencies
           run: |
             python3 -m pip install jupyter quartodoc
-            python3 -m pip install -e . --no-deps
+            python3 -m pip install -e .
         - name: Build autodoc
           run: quartodoc build
         - name: Publish to GitHub Pages (and render)

.github/workflows/lint.yml

@@ -9,6 +9,7 @@ on:
        - '.github/workflows/*.yml'
        - "*.[q]md"
        - "examples/**/*.y[a]?ml"
+       - ".pre-commit-config.yaml"
   workflow_dispatch:
 
 jobs:

.github/workflows/main.yml

@@ -29,7 +29,12 @@ jobs:
           - cuda: 126
             cuda_version: 12.6.3
             python_version: "3.11"
-            pytorch: 2.7.0
+            pytorch: 2.7.1
+            axolotl_extras:
+          - cuda: 128
+            cuda_version: 12.8.1
+            python_version: "3.11"
+            pytorch: 2.7.1
             axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:
@@ -92,7 +97,12 @@ jobs:
           - cuda: 126
             cuda_version: 12.6.3
             python_version: "3.11"
-            pytorch: 2.7.0
+            pytorch: 2.7.1
+            axolotl_extras:
+          - cuda: 128
+            cuda_version: 12.8.1
+            python_version: "3.11"
+            pytorch: 2.7.1
             axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:

.github/workflows/multi-gpu-e2e.yml

@@ -3,7 +3,7 @@ name: docker-multigpu-tests-biweekly
 on:
   pull_request:
     paths:
-      - 'tests/e2e/multigpu/*.py'
+      - 'tests/e2e/multigpu/**.py'
       - 'requirements.txt'
       - 'setup.py'
       - 'pyproject.toml'
@@ -43,7 +43,7 @@ jobs:
           - cuda: 126
             cuda_version: 12.6.3
             python_version: "3.11"
-            pytorch: 2.7.0
+            pytorch: 2.7.1
             axolotl_extras:
             num_gpus: 2
             nightly_build: "true"
@@ -59,7 +59,7 @@ jobs:
       - name: Install Modal
         run: |
           python -m pip install --upgrade pip
-          pip install modal==0.71.8 jinja2
+          pip install modal==1.0.2 jinja2
       - name: Update env vars
         run: |
           echo "BASE_TAG=main-base-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}" >> $GITHUB_ENV

.github/workflows/precommit-autoupdate.yml

@@ -25,7 +25,6 @@ jobs:
           pre-commit autoupdate
           if [[ -n $(git status --porcelain) ]]; then
             echo "changes=true" >> $GITHUB_OUTPUT
-            git diff .pre-commit-config.yaml > pre-commit-update.diff
           fi
 
       - name: Create Pull Request
@@ -39,11 +38,3 @@ jobs:
           commit-message: "chore: update pre-commit hooks"
           body: |
             Automated PR to update pre-commit hooks to their latest versions.
-
-            <details>
-            <summary>Changes:</summary>
-
-            ```diff
-            ${{ steps.update.outputs.diff }}
-            ```
-            </details>

.github/workflows/preview-docs.yml

@@ -8,7 +8,9 @@ on:
     paths:
       - '**/*.md'      # any Markdown file
       - '**/*.qmd'     # any Quarto file
-      - '_quarto.yaml'
+      - '_quarto.yml'
+      - docs/scripts/generate_config_docs.py
+      - src/axolotl/utils/schemas/**.py
 
 permissions:
   checks: write
@@ -38,7 +40,7 @@ jobs:
       - name: Install dependencies
         run: |
           python3 -m pip install jupyter quartodoc
-          python3 -m pip install -e . --no-deps
+          python3 -m pip install -e .
 
       - name: Build autodoc
         run: quartodoc build

.github/workflows/tests.yml

@@ -44,115 +44,26 @@ jobs:
         env:
           SKIP: no-commit-to-branch
 
-  preload-cache:
-    name: Preload HF cache
-    runs-on: ubuntu-latest
-    strategy:
-      fail-fast: false
-      matrix:
-        python_version: ["3.11"]
-        pytorch_version: ["2.6.0"]
-    timeout-minutes: 20
-
-    env:
-      AXOLOTL_IS_CI_CACHE_PRELOAD: "1"
-
-    steps:
-      - name: Check out repository code
-        uses: actions/checkout@v4
-
-      - name: Restore HF cache
-        id: hf-cache-restore
-        uses: actions/cache/restore@v4
-        with:
-          path: |
-            /home/runner/.cache/huggingface/hub/datasets--*
-            /home/runner/.cache/huggingface/hub/models--*
-          key: ${{ runner.os }}-hf-hub-cache-v2
-
-      - name: Setup Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: ${{ matrix.python_version }}
-          cache: 'pip' # caching pip dependencies
-
-      - name: upgrade pip
-        run: |
-          pip3 install --upgrade pip
-          pip3 install --upgrade packaging==23.2 setuptools==75.8.0 wheel
-
-      - name: Install PyTorch
-        run: |
-          pip3 install torch==${{ matrix.pytorch_version }}
-
-      - name: Install dependencies
-        run: |
-          pip3 show torch
-          pip3 install --no-build-isolation -U -e .
-          python scripts/unsloth_install.py | sh
-          python scripts/cutcrossentropy_install.py | sh
-          pip3 install -r requirements-dev.txt -r requirements-tests.txt
-
-      - name: Make sure PyTorch version wasn't clobbered
-        run: |
-          python -c "import torch; assert '${{ matrix.pytorch_version }}' in torch.__version__"
-
-      - name: Ensure axolotl CLI was installed
-        run: |
-          axolotl --help
-
-      - name: Pre-Download dataset fixture
-        run: |
-          huggingface-cli download --repo-type=dataset axolotl-ai-internal/axolotl-oss-dataset-fixtures
-
-      - name: Run tests
-        run: |
-          pytest -v tests/conftest.py
-
-      - name: Upload coverage to Codecov
-        uses: codecov/codecov-action@v5
-        with:
-          token: ${{ secrets.CODECOV_TOKEN }}
-          files: ./coverage.xml
-          flags: unittests,pytorch-${{ matrix.pytorch_version }}
-          fail_ci_if_error: false
-
-      - name: cleanup pip cache
-        run: |
-          find "$(pip cache dir)/http-v2" -type f -mtime +14 -exec rm {} \;
-
-      - name: Save HF cache
-        id: hf-cache
-        uses: actions/cache/save@v4
-        with:
-          path: |
-            /home/runner/.cache/huggingface/hub/datasets--*
-            /home/runner/.cache/huggingface/hub/models--*
-          key: ${{ steps.hf-cache-restore.outputs.cache-primary-key }}
-
   pytest:
     name: PyTest
     runs-on: ubuntu-latest
-    needs: [preload-cache]
+#    needs: [preload-cache]
     strategy:
       fail-fast: false
       matrix:
         python_version: ["3.11"]
-        pytorch_version: ["2.5.1", "2.6.0", "2.7.0"]
+        pytorch_version: ["2.5.1", "2.6.0", "2.7.1"]
     timeout-minutes: 20
 
     steps:
       - name: Check out repository code
         uses: actions/checkout@v4
 
-      - name: Restore HF cache
-        id: hf-cache-restore
-        uses: actions/cache/restore@v4
-        with:
-          path: |
-            /home/runner/.cache/huggingface/hub/datasets--*
-            /home/runner/.cache/huggingface/hub/models--*
-          key: ${{ runner.os }}-hf-hub-cache-v2
+      - name: Restore Cache from S3
+        id: hf-cache-restore-s3
+        run: |
+          mkdir -p /home/runner/.cache/huggingface/hub
+          curl -L https://d1dttdx32dkk5p.cloudfront.net/hf-cache.tar.zst | tar -xf - -C /home/runner/.cache/huggingface/hub/  --use-compress-program unzstd
 
       - name: Setup Python
         uses: actions/setup-python@v5
@@ -210,26 +121,22 @@ jobs:
   pytest-sdist:
     name: PyTest from Source Dist
     runs-on: ubuntu-latest
-    needs: [preload-cache]
     strategy:
       fail-fast: false
       matrix:
         python_version: ["3.11"]
-        pytorch_version: ["2.5.1", "2.6.0", "2.7.0"]
+        pytorch_version: ["2.5.1", "2.6.0", "2.7.1"]
     timeout-minutes: 20
 
     steps:
       - name: Check out repository code
         uses: actions/checkout@v4
 
-      - name: Restore HF cache
-        id: hf-cache-restore
-        uses: actions/cache/restore@v4
-        with:
-          path: |
-            /home/runner/.cache/huggingface/hub/datasets--*
-            /home/runner/.cache/huggingface/hub/models--*
-          key: ${{ runner.os }}-hf-hub-cache-v2
+      - name: Restore Cache from S3
+        id: hf-cache-restore-s3
+        run: |
+          mkdir -p /home/runner/.cache/huggingface/hub
+          curl -L https://d1dttdx32dkk5p.cloudfront.net/hf-cache.tar.zst | tar -xf - -C /home/runner/.cache/huggingface/hub/  --use-compress-program unzstd
 
       - name: Setup Python
         uses: actions/setup-python@v5
@@ -277,12 +184,124 @@ jobs:
           find "$(pip cache dir)/http-v2" -type f -mtime +14 -exec rm {} \;
 
   docker-e2e-tests-1st:
+    # Run this job first as a gate for running the remainder of the test matrix
     if: ${{ ! contains(github.event.commits[0].message, '[skip e2e]') && github.repository_owner == 'axolotl-ai-cloud' }}
     # this job needs to be run on self-hosted GPU runners...
     runs-on: [self-hosted, modal]
-    timeout-minutes: 90
+    timeout-minutes: 120
     needs: [pre-commit, pytest, pytest-sdist]
 
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.6.0
+            num_gpus: 1
+            axolotl_extras: vllm
+          - cuda: 126
+            cuda_version: 12.6.3
+            python_version: "3.11"
+            pytorch: 2.6.0
+            num_gpus: 1
+            axolotl_extras:
+            dockerfile: "Dockerfile-uv.jinja"
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Install Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+      - name: Install Modal
+        run: |
+          python -m pip install --upgrade pip
+          pip install modal==1.0.2 jinja2
+      - name: Update env vars
+        run: |
+          echo "BASE_TAG=main-base-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}" >> $GITHUB_ENV
+          echo "PYTORCH_VERSION=${{ matrix.pytorch}}" >> $GITHUB_ENV
+          echo "AXOLOTL_ARGS=${{ matrix.axolotl_args}}" >> $GITHUB_ENV
+          echo "AXOLOTL_EXTRAS=${{ matrix.axolotl_extras}}" >> $GITHUB_ENV
+          echo "CUDA=${{ matrix.cuda }}" >> $GITHUB_ENV
+          echo "MODAL_IMAGE_BUILDER_VERSION=2024.10" >> $GITHUB_ENV
+          echo "N_GPUS=${{ matrix.num_gpus }}" >> $GITHUB_ENV
+          echo "CODECOV_TOKEN=${{ secrets.CODECOV_TOKEN }}" >> $GITHUB_ENV
+          echo "E2E_DOCKERFILE=${{ matrix.dockerfile || 'Dockerfile.jinja'}}" >> $GITHUB_ENV
+      - name: Run tests job on Modal
+        run: |
+          modal run cicd.e2e_tests
+
+  docker-e2e-tests:
+    if: github.repository_owner == 'axolotl-ai-cloud'
+    # this job needs to be run on self-hosted GPU runners...
+    runs-on: [self-hosted, modal]
+    timeout-minutes: 120
+    # Only run the remainder of the matrix if the first e2e check passed;
+    # this is to save on wasted compute costs for known failures that get caught in the first run
+    needs: [pre-commit, pytest, docker-e2e-tests-1st]
+
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.6.0
+            num_gpus: 1
+            axolotl_extras: llmcompressor
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.5.1
+            num_gpus: 1
+            axolotl_extras:
+          - cuda: 126
+            cuda_version: 12.6.3
+            python_version: "3.11"
+            pytorch: 2.7.1
+            num_gpus: 1
+            axolotl_extras:
+          - cuda: 128
+            cuda_version: 12.8.1
+            python_version: "3.11"
+            pytorch: 2.7.1
+            num_gpus: 1
+            axolotl_extras:
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Install Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+      - name: Install Modal
+        run: |
+          python -m pip install --upgrade pip
+          pip install modal==1.0.2 jinja2
+      - name: Update env vars
+        run: |
+          echo "BASE_TAG=main-base-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}" >> $GITHUB_ENV
+          echo "PYTORCH_VERSION=${{ matrix.pytorch}}" >> $GITHUB_ENV
+          echo "AXOLOTL_ARGS=${{ matrix.axolotl_args}}" >> $GITHUB_ENV
+          echo "AXOLOTL_EXTRAS=${{ matrix.axolotl_extras}}" >> $GITHUB_ENV
+          echo "CUDA=${{ matrix.cuda }}" >> $GITHUB_ENV
+          echo "MODAL_IMAGE_BUILDER_VERSION=2024.10" >> $GITHUB_ENV
+          echo "N_GPUS=${{ matrix.num_gpus }}" >> $GITHUB_ENV
+          echo "CODECOV_TOKEN=${{ secrets.CODECOV_TOKEN }}" >> $GITHUB_ENV
+          echo "E2E_DOCKERFILE=${{ matrix.dockerfile || 'Dockerfile.jinja'}}" >> $GITHUB_ENV
+      - name: Run tests job on Modal
+        run: |
+          modal run cicd.e2e_tests
+
+  docker-e2e-cleanup:
+    runs-on: [self-hosted, modal]
+    timeout-minutes: 90
+    needs: [docker-e2e-tests]
+
     strategy:
       fail-fast: false
       matrix:
@@ -303,7 +322,7 @@ jobs:
       - name: Install Modal
         run: |
           python -m pip install --upgrade pip
-          pip install modal==0.71.8 jinja2
+          pip install modal==1.0.2 jinja2
       - name: Update env vars
         run: |
           echo "BASE_TAG=main-base-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}" >> $GITHUB_ENV
@@ -316,64 +335,4 @@ jobs:
           echo "CODECOV_TOKEN=${{ secrets.CODECOV_TOKEN }}" >> $GITHUB_ENV
       - name: Run tests job on Modal
         run: |
-          modal run cicd.e2e_tests
-
-  docker-e2e-tests:
-    if: github.repository_owner == 'axolotl-ai-cloud'
-    # this job needs to be run on self-hosted GPU runners...
-    runs-on: [self-hosted, modal]
-    timeout-minutes: 90
-    needs: [pre-commit, pytest, docker-e2e-tests-1st]
-
-    strategy:
-      fail-fast: false
-      matrix:
-        include:
-          - cuda: 124
-            cuda_version: 12.4.1
-            python_version: "3.11"
-            pytorch: 2.6.0
-            num_gpus: 1
-            axolotl_extras: llmcompressor
-          - cuda: 124
-            cuda_version: 12.4.1
-            python_version: "3.11"
-            pytorch: 2.4.1
-            num_gpus: 1
-            axolotl_extras:
-          - cuda: 124
-            cuda_version: 12.4.1
-            python_version: "3.11"
-            pytorch: 2.5.1
-            num_gpus: 1
-            axolotl_extras:
-          - cuda: 126
-            cuda_version: 12.6.3
-            python_version: "3.11"
-            pytorch: 2.7.0
-            num_gpus: 1
-            axolotl_extras:
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-      - name: Install Python
-        uses: actions/setup-python@v5
-        with:
-          python-version: "3.11"
-      - name: Install Modal
-        run: |
-          python -m pip install --upgrade pip
-          pip install modal==0.71.8 jinja2
-      - name: Update env vars
-        run: |
-          echo "BASE_TAG=main-base-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}" >> $GITHUB_ENV
-          echo "PYTORCH_VERSION=${{ matrix.pytorch}}" >> $GITHUB_ENV
-          echo "AXOLOTL_ARGS=${{ matrix.axolotl_args}}" >> $GITHUB_ENV
-          echo "AXOLOTL_EXTRAS=${{ matrix.axolotl_extras}}" >> $GITHUB_ENV
-          echo "CUDA=${{ matrix.cuda }}" >> $GITHUB_ENV
-          echo "MODAL_IMAGE_BUILDER_VERSION=2024.10" >> $GITHUB_ENV
-          echo "N_GPUS=${{ matrix.num_gpus }}" >> $GITHUB_ENV
-          echo "CODECOV_TOKEN=${{ secrets.CODECOV_TOKEN }}" >> $GITHUB_ENV
-      - name: Run tests job on Modal
-        run: |
-          modal run cicd.e2e_tests
+          modal run cicd.cleanup

.pre-commit-config.yaml

@@ -19,15 +19,15 @@ repos:
     hooks:
       - id: isort
 -   repo: https://github.com/PyCQA/flake8
-    rev: 7.1.2
+    rev: 7.3.0
     hooks:
     - id: flake8
 -   repo: https://github.com/pylint-dev/pylint
-    rev: v3.3.6
+    rev: v3.3.7
     hooks:
     - id: pylint
 -   repo: https://github.com/pre-commit/mirrors-mypy
-    rev: v1.15.0
+    rev: v1.16.1
     hooks:
     - id: mypy
       additional_dependencies:
@@ -36,7 +36,7 @@ repos:
             'pydantic>=2.5.3',
         ]
 -   repo: https://github.com/PyCQA/bandit
-    rev: 1.8.3
+    rev: 1.8.5
     hooks:
     -   id: bandit
         args: [

.runpod/README.md

@@ -328,7 +328,7 @@ The following optimizers are supported:
 - Use `gradient_checkpointing: true` to reduce memory usage
 - Adjust `micro_batch_size` and `gradient_accumulation_steps` based on your GPU memory
 
-For more detailed information, please refer to the [documentation](https://axolotl-ai-cloud.github.io/axolotl/docs/config.html).
+For more detailed information, please refer to the [documentation](https://axolotl-ai-cloud.github.io/axolotl/docs/config-reference.html).
 
 ### Errors:
 

.runpod/src/config/config.yaml

@@ -242,16 +242,12 @@
 # early_stopping_patience: 3
 
 # # Specify a scheduler and kwargs to use with the optimizer
-# lr_scheduler: # 'one_cycle' | 'log_sweep' | empty for cosine
+# lr_scheduler: # 'one_cycle' | empty for cosine
 # lr_scheduler_kwargs:
 
 # # For one_cycle optim
 # lr_div_factor: # Learning rate div factor
 
-# # For log_sweep optim
-# log_sweep_min_lr:
-# log_sweep_max_lr:
-
 # # Specify optimizer
 # # Valid values are driven by the Transformers OptimizerNames class, see:
 # # https://github.com/huggingface/transformers/blob/95b374952dc27d8511541d6f5a4e22c9ec11fb24/src/transformers/training_args.py#L134

.runpod/src/handler.py

@@ -57,8 +57,10 @@ async def handler(job):
     logger.info("Training Complete.")
 
     # Cleanup
-    del os.environ["WANDB_API_KEY"]
-    del os.environ["HF_TOKEN"]
+    if "WANDB_API_KEY" in os.environ:
+        del os.environ["WANDB_API_KEY"]
+    if "HF_TOKEN" in os.environ:
+        del os.environ["HF_TOKEN"]
 
 
 runpod.serverless.start({"handler": handler, "return_aggregate_stream": True})

README.md

@@ -22,28 +22,32 @@
     <img src="https://github.com/axolotl-ai-cloud/axolotl/actions/workflows/multi-gpu-e2e.yml/badge.svg" alt="multigpu-semi-weekly tests">
 </p>
 
-Axolotl is a tool designed to streamline post-training for various AI models.
-Post-training refers to any modifications or additional training performed on
-pre-trained models - including full model fine-tuning, parameter-efficient tuning (like
-LoRA and QLoRA), supervised fine-tuning (SFT), instruction tuning, and alignment
-techniques. With support for multiple model architectures and training configurations,
-Axolotl makes it easy to get started with these techniques.
 
-Axolotl is designed to work with YAML config files that contain everything you need to
-preprocess a dataset, train or fine-tune a model, run model inference or evaluation,
-and much more.
+## 🎉 Latest Updates
+
+- 2025/06: Magistral with mistral-common tokenizer support has been added to Axolotl. See [examples](https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/magistral) to start training your own Magistral models with Axolotl!
+- 2025/05: Quantization Aware Training (QAT) support has been added to Axolotl. Explore the [docs](https://docs.axolotl.ai/docs/qat.html) to learn more!
+- 2025/04: Llama 4 support has been added in Axolotl. See [examples](https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/llama-4) to start training your own Llama 4 models with Axolotl's linearized version!
+- 2025/03: Axolotl has implemented Sequence Parallelism (SP) support. Read the [blog](https://huggingface.co/blog/axolotl-ai-co/long-context-with-sequence-parallelism-in-axolotl) and [docs](https://docs.axolotl.ai/docs/sequence_parallelism.html) to learn how to scale your context length when fine-tuning.
+- 2025/03: (Beta) Fine-tuning Multimodal models is now supported in Axolotl. Check out the [docs](https://docs.axolotl.ai/docs/multimodal.html) to fine-tune your own!
+- 2025/02: Axolotl has added LoRA optimizations to reduce memory usage and improve training speed for LoRA and QLoRA in single GPU and multi-GPU training (DDP and DeepSpeed). Jump into the [docs](https://docs.axolotl.ai/docs/lora_optims.html) to give it a try.
+- 2025/02: Axolotl has added GRPO support. Dive into our [blog](https://huggingface.co/blog/axolotl-ai-co/training-llms-w-interpreter-feedback-wasm) and [GRPO example](https://github.com/axolotl-ai-cloud/grpo_code) and have some fun!
+- 2025/01: Axolotl has added Reward Modelling / Process Reward Modelling fine-tuning support. See [docs](https://docs.axolotl.ai/docs/reward_modelling.html).
+
+## ✨ Overview
+
+Axolotl is a tool designed to streamline post-training for various AI models.
 
 Features:
 
-- Train various Huggingface models such as llama, pythia, falcon, mpt
-- Supports fullfinetune, lora, qlora, relora, and gptq
-- Customize configurations using a simple yaml file or CLI overwrite
-- Load different dataset formats, use custom formats, or bring your own tokenized datasets
-- Integrated with [xformers](https://github.com/facebookresearch/xformers), flash attention, [liger kernel](https://github.com/linkedin/Liger-Kernel), rope scaling, and multipacking
-- Works with single GPU or multiple GPUs via FSDP or Deepspeed
-- Easily run with Docker locally or on the cloud
-- Log results and optionally checkpoints to wandb, mlflow or Comet
-- And more!
+- **Multiple Model Support**: Train various models like LLaMA, Mistral, Mixtral, Pythia, and more. We are compatible with HuggingFace transformers causal language models.
+- **Training Methods**: Full fine-tuning, LoRA, QLoRA, GPTQ, QAT, Preference Tuning (DPO, IPO, KTO, ORPO), RL (GRPO), Multimodal, and Reward Modelling (RM) / Process Reward Modelling (PRM).
+- **Easy Configuration**: Re-use a single YAML file between dataset preprocess, training, evaluation, quantization, and inference.
+- **Performance Optimizations**: [Multipacking](https://docs.axolotl.ai/docs/multipack.html), [Flash Attention](https://github.com/Dao-AILab/flash-attention), [Xformers](https://github.com/facebookresearch/xformers), [Flex Attention](https://pytorch.org/blog/flexattention/), [Liger Kernel](https://github.com/linkedin/Liger-Kernel), [Cut Cross Entropy](https://github.com/apple/ml-cross-entropy/tree/main), [Sequence Parallelism (SP)](https://docs.axolotl.ai/docs/sequence_parallelism.html), [LoRA optimizations](https://docs.axolotl.ai/docs/lora_optims.html), [Multi-GPU training (FSDP1, FSDP2, DeepSpeed)](https://docs.axolotl.ai/docs/multi-gpu.html), [Multi-node training (Torchrun, Ray)](https://docs.axolotl.ai/docs/multi-node.html), and many more!
+- **Flexible Dataset Handling**: Load from local, HuggingFace, and cloud (S3, Azure, GCP, OCI) datasets.
+- **Cloud Ready**: We ship [Docker images](https://hub.docker.com/u/axolotlai) and also [PyPI packages](https://pypi.org/project/axolotl/) for use on cloud platforms and local hardware.
+
+
 
 ## 🚀 Quick Start
 
@@ -51,7 +55,7 @@ Features:
 
 - NVIDIA GPU (Ampere or newer for `bf16` and Flash Attention) or AMD GPU
 - Python 3.11
-- PyTorch ≥2.4.1
+- PyTorch ≥2.5.1
 
 ### Installation
 
@@ -81,19 +85,12 @@ axolotl train examples/llama-3/lora-1b.yml
 
 That's it! Check out our [Getting Started Guide](https://docs.axolotl.ai/docs/getting-started.html) for a more detailed walkthrough.
 
-## ✨ Key Features
-
-- **Multiple Model Support**: Train various models like LLaMA, Mistral, Mixtral, Pythia, and more
-- **Training Methods**: Full fine-tuning, LoRA, QLoRA, and more
-- **Easy Configuration**: Simple YAML files to control your training setup
-- **Performance Optimizations**: Flash Attention, xformers, multi-GPU training
-- **Flexible Dataset Handling**: Use various formats and custom datasets
-- **Cloud Ready**: Run on cloud platforms or local hardware
 
 ## 📚 Documentation
 
 - [Installation Options](https://docs.axolotl.ai/docs/installation.html) - Detailed setup instructions for different environments
-- [Configuration Guide](https://docs.axolotl.ai/docs/config.html) - Full configuration options and examples
+- [Configuration Guide](https://docs.axolotl.ai/docs/config-reference.html) - Full configuration options and examples
+- [Dataset Loading](https://docs.axolotl.ai/docs/dataset_loading.html) - Loading datasets from various sources
 - [Dataset Guide](https://docs.axolotl.ai/docs/dataset-formats/) - Supported formats and how to use them
 - [Multi-GPU Training](https://docs.axolotl.ai/docs/multi-gpu.html)
 - [Multi-Node Training](https://docs.axolotl.ai/docs/multi-node.html)
@@ -112,31 +109,6 @@ That's it! Check out our [Getting Started Guide](https://docs.axolotl.ai/docs/ge
 
 Contributions are welcome! Please see our [Contributing Guide](https://github.com/axolotl-ai-cloud/axolotl/blob/main/.github/CONTRIBUTING.md) for details.
 
-## Supported Models
-
-|             | fp16/fp32 | lora | qlora | gptq | gptq w/flash attn | flash attn | xformers attn |
-|-------------|:----------|:-----|-------|------|-------------------|------------|--------------|
-| llama       | ✅         | ✅    | ✅     | ✅             | ✅                 | ✅          | ✅            |
-| Mistral     | ✅         | ✅    | ✅     | ✅             | ✅                 | ✅          | ✅            |
-| Mixtral-MoE | ✅         | ✅    | ✅     | ❓             | ❓                 | ❓          | ❓            |
-| Mixtral8X22 | ✅         | ✅    | ✅     | ❓             | ❓                 | ❓          | ❓            |
-| Pythia      | ✅         | ✅    | ✅     | ❌             | ❌                 | ❌          | ❓            |
-| cerebras    | ✅         | ✅    | ✅     | ❌             | ❌                 | ❌          | ❓            |
-| btlm        | ✅         | ✅    | ✅     | ❌             | ❌                 | ❌          | ❓            |
-| mpt         | ✅         | ❌    | ❓     | ❌             | ❌                 | ❌          | ❓            |
-| falcon      | ✅         | ✅    | ✅     | ❌             | ❌                 | ❌          | ❓            |
-| gpt-j       | ✅         | ✅    | ✅     | ❌             | ❌                 | ❓          | ❓            |
-| XGen        | ✅         | ❓    | ✅     | ❓             | ❓                 | ❓          | ✅            |
-| phi         | ✅         | ✅    | ✅     | ❓             | ❓                 | ❓          | ❓            |
-| RWKV        | ✅         | ❓    | ❓     | ❓             | ❓                 | ❓          | ❓            |
-| Qwen        | ✅         | ✅    | ✅     | ❓             | ❓                 | ❓          | ❓            |
-| Gemma       | ✅         | ✅    | ✅     | ❓             | ❓                 | ✅          | ❓            |
-| Jamba       | ✅         | ✅    | ✅     | ❓             | ❓                 | ✅          | ❓            |
-
-✅: supported
-❌: not supported
-❓: untested
-
 ## ❤️ Sponsors
 
 Thank you to our sponsors who help make Axolotl possible:

_quarto.yml

@@ -1,5 +1,6 @@
 project:
   type: website
+  pre-render: docs/scripts/generate_config_docs.py
 
 quartodoc:
   dir: docs/api
@@ -17,7 +18,9 @@ quartodoc:
         - convert
         - prompt_tokenizers
         - logging_config
-        - core.trainer_builder
+        - core.builders.base
+        - core.builders.causal
+        - core.builders.rl
         - core.training_args
         - core.chat.messages
         - core.chat.format.chatml
@@ -43,13 +46,37 @@ quartodoc:
         - cli.vllm_serve
         - cli.cloud.base
         - cli.cloud.modal_
+        - cli.quantize
     - title: Trainers
       desc: Training implementations
       contents:
         - core.trainers.base
         - core.trainers.trl
+        - core.trainers.mamba
+        - core.trainers.relora
         - core.trainers.dpo.trainer
         - core.trainers.grpo.trainer
+        - core.trainers.grpo.sampler
+        - core.trainers.utils
+    - title: Model Loading
+      desc: Functionality for loading and patching models, tokenizers, etc.
+      contents:
+        - loaders.model
+        - loaders.tokenizer
+        - loaders.processor
+        - loaders.adapter
+        - loaders.patch_manager
+        - loaders.constants
+    - title: Mixins
+      desc: Mixin classes for augmenting trainers
+      contents:
+        - core.trainers.mixins.optimizer
+        - core.trainers.mixins.rng_state_loader
+        - core.trainers.mixins.scheduler
+    - title: Context Managers
+      desc: Context managers for altering trainer behaviors
+      contents:
+        - utils.ctx_managers.sequence_parallel
     - title: Prompt Strategies
       desc: Prompt formatting strategies
       contents:
@@ -86,7 +113,7 @@ quartodoc:
         - kernels.swiglu
         - kernels.quantize
         - kernels.utils
-    - title: MonkeyPatches
+    - title: Monkey Patches
       desc: Runtime patches for model optimizations
       contents:
         - monkeypatch.llama_attn_hijack_flash
@@ -103,17 +130,16 @@ quartodoc:
         - monkeypatch.trainer_fsdp_optim
         - monkeypatch.transformers_fa_utils
         - monkeypatch.unsloth_
-        - monkeypatch.attention.mllama
         - monkeypatch.data.batch_dataset_fetcher
         - monkeypatch.mixtral
+        - monkeypatch.gradient_checkpointing.offload_cpu
+        - monkeypatch.gradient_checkpointing.offload_disk
     - title: Utils
       desc: Utility functions
       contents:
-        - utils.models
         - utils.tokenization
         - utils.chat_templates
         - utils.lora
-        - utils.lora_embeddings
         - utils.model_shard_quant
         - utils.bench
         - utils.freeze
@@ -124,7 +150,7 @@ quartodoc:
         - utils.optimizers.adopt
         - utils.data.pretraining
         - utils.data.sft
-        - utils.gradient_checkpointing.unsloth
+        - utils.quantization
     - title: Schemas
       desc: Pydantic data models for Axolotl config
       contents:
@@ -174,12 +200,14 @@ quartodoc:
         - utils.callbacks.lisa
         - utils.callbacks.mlflow_
         - utils.callbacks.comet_
-
+        - utils.callbacks.qat
 website:
   title: "Axolotl"
   description: "We make fine-tuning accessible, scalable, and fun"
   favicon: favicon.jpg
 
+  google-analytics: "G-9KYCVJBNMQ"
+
   navbar:
     logo: image/axolotl_logo_digital_white.svg
     title: false
@@ -208,7 +236,7 @@ website:
             - docs/installation.qmd
             - docs/inference.qmd
             - docs/cli.qmd
-            - docs/config.qmd
+            - docs/config-reference.qmd
             - text: "API Reference"
               href: docs/api
 
@@ -232,6 +260,8 @@ website:
             - docs/lr_groups.qmd
             - docs/lora_optims.qmd
             - docs/dataset_loading.qmd
+            - docs/qat.qmd
+            - docs/quantize.qmd
 
         - section: "Core Concepts"
           contents:

cicd/Dockerfile-uv.jinja

@@ -0,0 +1,52 @@
+FROM axolotlai/axolotl-base-uv:{{ BASE_TAG }}
+
+ENV TORCH_CUDA_ARCH_LIST="7.0 7.5 8.0 8.6 9.0+PTX"
+ENV AXOLOTL_EXTRAS="{{ AXOLOTL_EXTRAS }}"
+ENV AXOLOTL_ARGS="{{ AXOLOTL_ARGS }}"
+ENV CUDA="{{ CUDA }}"
+ENV PYTORCH_VERSION="{{ PYTORCH_VERSION }}"
+ENV GITHUB_REF="{{ GITHUB_REF }}"
+ENV GITHUB_SHA="{{ GITHUB_SHA }}"
+ENV NIGHTLY_BUILD="{{ NIGHTLY_BUILD }}"
+ENV HF_HOME="{{ HF_HOME }}"
+
+RUN apt-get update && \
+    apt-get install -y --allow-change-held-packages vim curl nano libnccl2 libnccl-dev
+
+WORKDIR /workspace
+
+RUN git clone --depth=1 https://github.com/axolotl-ai-cloud/axolotl.git
+
+WORKDIR /workspace/axolotl
+
+RUN git fetch origin +$GITHUB_REF && \
+    git checkout FETCH_HEAD
+
+# If AXOLOTL_EXTRAS is set, append it in brackets
+RUN if [ "$NIGHTLY_BUILD" = "true" ] ; then \
+        sed -i 's#^transformers.*#transformers @ git+https://github.com/huggingface/transformers.git@main#' requirements.txt; \
+        sed -i 's#^peft.*#peft @ git+https://github.com/huggingface/peft.git@main#' requirements.txt; \
+        sed -i 's#^accelerate.*#accelerate @ git+https://github.com/huggingface/accelerate.git@main#' requirements.txt; \
+        sed -i 's#^trl.*#trl @ git+https://github.com/huggingface/trl.git@main#' requirements.txt; \
+        sed -i 's#^datasets.*#datasets @ git+https://github.com/huggingface/datasets.git@main#' requirements.txt; \
+    fi
+
+RUN uv pip install packaging==23.2 setuptools==75.8.0
+RUN if [ "$AXOLOTL_EXTRAS" != "" ] ; then \
+        uv pip install --no-build-isolation -e .[deepspeed,flash-attn,ring-flash-attn,optimizers,ray,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
+    else \
+        uv pip install --no-build-isolation -e .[deepspeed,flash-attn,ring-flash-attn,optimizers,ray] $AXOLOTL_ARGS; \
+    fi
+
+RUN python scripts/unsloth_install.py --uv | sh
+RUN python scripts/cutcrossentropy_install.py --uv | sh
+
+# So we can test the Docker image
+RUN uv pip install -r requirements-dev.txt -r requirements-tests.txt
+
+# fix so that git fetch/pull from remote works
+RUN git config remote.origin.fetch "+refs/heads/*:refs/remotes/origin/*" && \
+    git config --get remote.origin.fetch
+
+# helper for huggingface-login cli
+RUN git config --global credential.helper store

cicd/cicd.sh

@@ -18,7 +18,7 @@ pytest -v --durations=10 \
   --cov-append
 
 # Run patched tests excluding lora kernels with coverage append
-pytest -v --durations=10 \
+pytest --full-trace -vvv --durations=10 \
   --ignore=tests/e2e/patched/lora_kernels \
   /workspace/axolotl/tests/e2e/patched \
   --cov=axolotl \

cicd/cleanup.py

@@ -0,0 +1,19 @@
+"""Modal app to run axolotl GPU cleanup"""
+
+from .single_gpu import VOLUME_CONFIG, app, cicd_image, run_cmd
+
+
+@app.function(
+    image=cicd_image,
+    timeout=60 * 60,
+    cpu=8.0,
+    memory=131072,
+    volumes=VOLUME_CONFIG,
+)
+def cleanup():
+    run_cmd("./cicd/cleanup.sh", "/workspace/axolotl")
+
+
+@app.local_entrypoint()
+def main():
+    cleanup.remote()

cicd/cleanup.sh

@@ -0,0 +1,6 @@
+#!/bin/bash
+set -e
+
+# cleanup old cache files for datasets processing and intermediate mappings
+find /workspace/data/huggingface-cache/hub/datasets -name "cache-*" -type f -mtime +1 -exec rm {} \;
+find /workspace/data/huggingface-cache/hub/datasets -name "*.lock" -type f -mtime +1 -exec rm {} \;

cicd/e2e_tests.py

@@ -1,75 +1,12 @@
 """Modal app to run axolotl GPU tests"""
 
-# pylint: disable=duplicate-code
-
-import os
-import pathlib
-import tempfile
-
-import jinja2
-import modal
-from jinja2 import select_autoescape
-from modal import App, Image
-
-cicd_path = pathlib.Path(__file__).parent.resolve()
-
-template_loader = jinja2.FileSystemLoader(searchpath=cicd_path)
-template_env = jinja2.Environment(
-    loader=template_loader, autoescape=select_autoescape()
-)
-df_template = template_env.get_template("Dockerfile.jinja")
-
-df_args = {
-    "AXOLOTL_EXTRAS": os.environ.get("AXOLOTL_EXTRAS", ""),
-    "AXOLOTL_ARGS": os.environ.get("AXOLOTL_ARGS", ""),
-    "PYTORCH_VERSION": os.environ.get("PYTORCH_VERSION", "2.4.1"),
-    "BASE_TAG": os.environ.get("BASE_TAG", "main-base-py3.11-cu121-2.4.1"),
-    "CUDA": os.environ.get("CUDA", "121"),
-    "GITHUB_REF": os.environ.get("GITHUB_REF", "refs/heads/main"),
-    "GITHUB_SHA": os.environ.get("GITHUB_SHA", ""),
-    "NIGHTLY_BUILD": os.environ.get("NIGHTLY_BUILD", ""),
-    "CODECOV_TOKEN": os.environ.get("CODECOV_TOKEN", ""),
-    "HF_HOME": "/workspace/data/huggingface-cache/hub",
-}
-
-dockerfile_contents = df_template.render(**df_args)
-
-temp_dir = tempfile.mkdtemp()
-with open(pathlib.Path(temp_dir) / "Dockerfile", "w", encoding="utf-8") as f:
-    f.write(dockerfile_contents)
-
-cicd_image = Image.from_dockerfile(
-    pathlib.Path(temp_dir) / "Dockerfile",
-    context_mount=None,
-    force_build=True,
-    gpu="A10G",
-).env(df_args)
-
-app = App("Axolotl CI/CD", secrets=[])
-
-hf_cache_volume = modal.Volume.from_name(
-    "axolotl-ci-hf-hub-cache", create_if_missing=True
-)
-VOLUME_CONFIG = {
-    "/workspace/data/huggingface-cache/hub": hf_cache_volume,
-}
-
-N_GPUS = int(os.environ.get("N_GPUS", 1))
-GPU_CONFIG = modal.gpu.L40S(count=N_GPUS)
-
-
-def run_cmd(cmd: str, run_folder: str):
-    import subprocess  # nosec
-
-    # Propagate errors from subprocess.
-    if exit_code := subprocess.call(cmd.split(), cwd=run_folder):  # nosec
-        exit(exit_code)  # pylint: disable=consider-using-sys-exit
+from .single_gpu import GPU_CONFIG, VOLUME_CONFIG, app, cicd_image, run_cmd
 
 
 @app.function(
     image=cicd_image,
     gpu=GPU_CONFIG,
-    timeout=60 * 60,
+    timeout=120 * 60,  # 90 min
     cpu=8.0,
     memory=131072,
     volumes=VOLUME_CONFIG,

cicd/multigpu.py

@@ -24,9 +24,9 @@ df_template = template_env.get_template("Dockerfile.jinja")
 df_args = {
     "AXOLOTL_EXTRAS": os.environ.get("AXOLOTL_EXTRAS", ""),
     "AXOLOTL_ARGS": os.environ.get("AXOLOTL_ARGS", ""),
-    "PYTORCH_VERSION": os.environ.get("PYTORCH_VERSION", "2.4.1"),
-    "BASE_TAG": os.environ.get("BASE_TAG", "main-base-py3.11-cu121-2.4.1"),
-    "CUDA": os.environ.get("CUDA", "121"),
+    "PYTORCH_VERSION": os.environ.get("PYTORCH_VERSION", "2.5.1"),
+    "BASE_TAG": os.environ.get("BASE_TAG", "main-base-py3.11-cu124-2.5.1"),
+    "CUDA": os.environ.get("CUDA", "124"),
     "GITHUB_REF": os.environ.get("GITHUB_REF", "refs/heads/main"),
     "GITHUB_SHA": os.environ.get("GITHUB_SHA", ""),
     "CODECOV_TOKEN": os.environ.get("CODECOV_TOKEN", ""),
@@ -55,7 +55,7 @@ VOLUME_CONFIG = {
 }
 
 N_GPUS = int(os.environ.get("N_GPUS", 2))
-GPU_CONFIG = modal.gpu.H100(count=N_GPUS)
+GPU_CONFIG = f"H100:{N_GPUS}"
 
 
 def run_cmd(cmd: str, run_folder: str):
@@ -69,8 +69,8 @@ def run_cmd(cmd: str, run_folder: str):
 @app.function(
     image=cicd_image,
     gpu=GPU_CONFIG,
-    timeout=90 * 60,
-    cpu=8.0,
+    timeout=120 * 60,
+    cpu=16.0,
     memory=131072 * N_GPUS,
     volumes=VOLUME_CONFIG,
 )

cicd/single_gpu.py

@@ -0,0 +1,68 @@
+"""Modal app to run axolotl GPU tests"""
+
+# pylint: disable=duplicate-code
+
+import os
+import pathlib
+import tempfile
+
+import jinja2
+import modal
+import modal.experimental
+from jinja2 import select_autoescape
+from modal import App
+
+cicd_path = pathlib.Path(__file__).parent.resolve()
+
+template_loader = jinja2.FileSystemLoader(searchpath=cicd_path)
+template_env = jinja2.Environment(
+    loader=template_loader, autoescape=select_autoescape()
+)
+dockerfile = os.environ.get("E2E_DOCKERFILE", "Dockerfile.jinja")
+df_template = template_env.get_template(dockerfile)
+
+df_args = {
+    "AXOLOTL_EXTRAS": os.environ.get("AXOLOTL_EXTRAS", ""),
+    "AXOLOTL_ARGS": os.environ.get("AXOLOTL_ARGS", ""),
+    "PYTORCH_VERSION": os.environ.get("PYTORCH_VERSION", "2.5.1"),
+    "BASE_TAG": os.environ.get("BASE_TAG", "main-base-py3.11-cu124-2.5.1"),
+    "CUDA": os.environ.get("CUDA", "124"),
+    "GITHUB_REF": os.environ.get("GITHUB_REF", "refs/heads/main"),
+    "GITHUB_SHA": os.environ.get("GITHUB_SHA", ""),
+    "NIGHTLY_BUILD": os.environ.get("NIGHTLY_BUILD", ""),
+    "CODECOV_TOKEN": os.environ.get("CODECOV_TOKEN", ""),
+    "HF_HOME": "/workspace/data/huggingface-cache/hub",
+}
+
+dockerfile_contents = df_template.render(**df_args)
+
+temp_dir = tempfile.mkdtemp()
+with open(pathlib.Path(temp_dir) / "Dockerfile", "w", encoding="utf-8") as f:
+    f.write(dockerfile_contents)
+
+cicd_image = modal.experimental.raw_dockerfile_image(
+    pathlib.Path(temp_dir) / "Dockerfile",
+    # context_mount=None,
+    force_build=True,
+    # gpu="A10G",
+).env(df_args)
+
+app = App("Axolotl CI/CD", secrets=[])
+
+hf_cache_volume = modal.Volume.from_name(
+    "axolotl-ci-hf-hub-cache", create_if_missing=True
+)
+VOLUME_CONFIG = {
+    "/workspace/data/huggingface-cache/hub": hf_cache_volume,
+}
+
+N_GPUS = int(os.environ.get("N_GPUS", 1))
+GPU_CONFIG = f"L40S:{N_GPUS}"
+
+
+def run_cmd(cmd: str, run_folder: str):
+    import subprocess  # nosec
+
+    # Propagate errors from subprocess.
+    if exit_code := subprocess.call(cmd.split(), cwd=run_folder):  # nosec
+        exit(exit_code)  # pylint: disable=consider-using-sys-exit

codecov.yml

@@ -19,7 +19,7 @@ coverage:
         if_no_uploads: error
         if_not_found: success
         if_ci_failed: error
-        only_pulls: false
+        only_pulls: true
         flags: null
         paths: null
     patch:

deepspeed_configs/zero2_torch_compile.json

@@ -0,0 +1,31 @@
+{
+  "compile": {
+    "disable": false,
+    "backend": "inductor"
+  },
+  "zero_optimization": {
+    "stage": 2,
+    "offload_optimizer": {
+      "device": "cpu"
+    },
+    "contiguous_gradients": true,
+    "overlap_comm": true
+  },
+  "bf16": {
+    "enabled": "auto"
+  },
+  "fp16": {
+    "enabled": "auto",
+    "auto_cast": false,
+    "loss_scale": 0,
+    "initial_scale_power": 32,
+    "loss_scale_window": 1000,
+    "hysteresis": 2,
+    "min_loss_scale": 1
+  },
+  "gradient_accumulation_steps": "auto",
+  "gradient_clipping": "auto",
+  "train_batch_size": "auto",
+  "train_micro_batch_size_per_gpu": "auto",
+  "wall_clock_breakdown": false
+}

docker/Dockerfile-base

@@ -38,6 +38,6 @@ RUN git lfs install --skip-repo && \
     # The base image ships with `pydantic==1.8.2` which is not working
     pip3 install -U --no-cache-dir pydantic==1.10.10
 
-RUN if [ "$PYTORCH_VERSION" = "2.7.0" ] ; then \
+RUN if [ "$PYTORCH_VERSION" = "2.7.1" ] ; then \
         pip3 install flash-attn==2.7.4.post1; \
     fi

docker/Dockerfile-base-next

@@ -29,7 +29,7 @@ ENV PATH="/root/miniconda3/envs/py${PYTHON_VERSION}/bin:${PATH}"
 WORKDIR /workspace
 
 RUN python3 -m pip install --upgrade pip && pip3 install packaging && \
-    python3 -m pip install --no-cache-dir -U torch==2.7.0 --extra-index-url https://download.pytorch.org/whl/test/cu$CUDA && \
+    python3 -m pip install --no-cache-dir -U torch==2.7.1 --extra-index-url https://download.pytorch.org/whl/test/cu$CUDA && \
     python3 -m pip install --no-cache-dir "causal_conv1d @ git+https://github.com/Dao-AILab/causal-conv1d.git@main" && \
     python3 -m pip install --no-cache-dir "mamba_ssm @ git+https://github.com/state-spaces/mamba.git@main"
 

docker/Dockerfile-uv-base

@@ -0,0 +1,40 @@
+ARG CUDA_VERSION="12.6.3"
+ARG CUDNN_VERSION=""
+ARG UBUNTU_VERSION="22.04"
+ARG MAX_JOBS=4
+
+FROM nvidia/cuda:$CUDA_VERSION-cudnn$CUDNN_VERSION-devel-ubuntu$UBUNTU_VERSION AS base-builder
+
+ARG PYTHON_VERSION="3.11"
+ARG PYTORCH_VERSION="2.6.0"
+ARG CUDA="126"
+ARG TORCH_CUDA_ARCH_LIST="7.0 7.5 8.0 8.6 9.0+PTX"
+
+ENV PYTHON_VERSION=$PYTHON_VERSION
+ENV TORCH_CUDA_ARCH_LIST=$TORCH_CUDA_ARCH_LIST
+ENV UV_TORCH_BACKEND="cu${CUDA}"
+
+RUN apt-get update \
+    && apt-get install -y wget git build-essential ninja-build git-lfs libaio-dev pkg-config curl && rm -rf /var/lib/apt/lists/* \
+    && git lfs install --skip-repo \
+    && curl -LsSf https://astral.sh/uv/install.sh | sh
+
+ENV PATH="/root/.local/bin:${PATH}"
+
+RUN uv python install ${PYTHON_VERSION}
+
+WORKDIR /workspace
+
+RUN uv venv --no-project --relocatable axolotl-venv
+
+ENV PATH="/workspace/axolotl-venv/bin:${PATH}"
+
+RUN uv pip install packaging setuptools wheel psutil \
+    && uv pip install torch==${PYTORCH_VERSION} \
+    && uv pip install --no-build-isolation "causal_conv1d @ git+https://github.com/Dao-AILab/causal-conv1d.git@main" \
+    && uv pip install "mamba_ssm @ git+https://github.com/state-spaces/mamba.git@main" \
+    && uv pip install awscli pydantic
+
+RUN if [ "$PYTORCH_VERSION" = "2.7.1" ] ; then \
+        uv pip install --no-build-isolation flash-attn==2.7.4.post1; \
+    fi

docs/.gitignore

@@ -2,3 +2,4 @@
 _site/
 /api/*.qmd
 /api/*.html
+config-reference.qmd

docs/cli.qmd

@@ -209,6 +209,16 @@ axolotl delinearize-llama4 --model path/to/model_dir --output path/to/output_dir
 
 This would be necessary to use with other frameworks. If you have an adapter, merge it with the non-quantized linearized model before delinearizing.
 
+### quantize
+
+Quantizes a model using the quantization configuration specified in your YAML file.
+
+```bash
+axolotl quantize config.yml
+```
+
+See [Quantization](./quantize.qmd) for more details.
+
 
 ## Legacy CLI Usage
 

docs/config.qmd

@@ -1,743 +0,0 @@
----
-title: Config Reference
-description: A complete list of all configuration options.
----
-
-```yaml
-# This is the huggingface model that contains *.pt, *.safetensors, or *.bin files
-# This can also be a relative path to a model on disk
-base_model: ./llama-7b-hf
-# You can specify an ignore pattern if the model repo contains more than 1 model type (*.pt, etc)
-base_model_ignore_patterns:
-# If the base_model repo on hf hub doesn't include configuration .json files,
-# You can set that here, or leave this empty to default to base_model
-base_model_config: ./llama-7b-hf
-# You can specify to choose a specific model revision from huggingface hub
-revision_of_model:
-# Optional tokenizer configuration path in case you want to use a different tokenizer
-# than the one defined in the base model
-tokenizer_config:
-# If you want to specify the type of model to load, AutoModelForCausalLM is a good choice too
-model_type: AutoModelForCausalLM
-# Corresponding tokenizer for the model AutoTokenizer is a good choice
-tokenizer_type: AutoTokenizer
-# Trust remote code for untrusted source
-trust_remote_code:
-# use_fast option for tokenizer loading from_pretrained, default to True
-tokenizer_use_fast:
-# Whether to use the legacy tokenizer setting, defaults to True
-tokenizer_legacy:
-# Resize the model embeddings when new tokens are added to multiples of 32
-# This is reported to improve training speed on some models
-resize_token_embeddings_to_32x:
-# Optional[bool] Whether to shrink the embeddings to len(tokenizer). By default, we won't shrink.
-shrink_embeddings:
-# Optional[bool] Don't upcast the embeddings to float32 when using PEFT. Useful for low-VRAM GPUs
-embeddings_skip_upcast:
-# Whether to load the model with randomly initialized weights. Useful for
-# pre-training a model from scratch or debugging purposes.
-random_init_weights:
-
-# (Internal use only)
-# Used to identify which the model is based on
-is_falcon_derived_model:
-is_llama_derived_model:
-is_qwen_derived_model:
-# Please note that if you set this to true, `padding_side` will be set to "left" by default
-is_mistral_derived_model:
-
-# optional overrides to the base model configuration
-overrides_of_model_config:
-  # RoPE Scaling https://github.com/huggingface/transformers/pull/24653
-  rope_scaling:
-    type: # linear | dynamic
-    factor: # float
-
-# optional overrides the base model loading from_pretrained
-overrides_of_model_kwargs:
-  # use_cache: False
-
-# optional overrides to the bnb 4bit quantization configuration
-# https://huggingface.co/docs/transformers/main/main_classes/quantization#transformers.BitsAndBytesConfig
-bnb_config_kwargs:
-  # These are default values
-  llm_int8_has_fp16_weight: false
-  bnb_4bit_quant_type: nf4
-  bnb_4bit_use_double_quant: true
-
-
-# Whether you are training a 4-bit GPTQ quantized model
-gptq: true
-
-# This will attempt to quantize the model down to 8 bits and use adam 8 bit optimizer
-load_in_8bit: true
-# Use bitsandbytes 4 bit
-load_in_4bit:
-
-# Use CUDA bf16
-bf16: true # bool or 'full' for `bf16_full_eval`, or 'auto' for automatic detection. require >=ampere
-# Use CUDA fp16
-fp16: true
-# Use CUDA tf32
-tf32: true # require >=ampere
-# Note: if bf16 is set to 'auto', and fp16 is set to true, we will prefer the explict fp16 setting
-
-# No AMP (automatic mixed precision)
-bfloat16: true # require >=ampere
-float16: true
-
-# Limit the memory for all available GPUs to this amount (if an integer, expressed in gigabytes); default: unset
-gpu_memory_limit: 20GiB
-# Do the LoRA/PEFT loading on CPU -- this is required if the base model is so large it takes up most or all of the available GPU VRAM, e.g. during a model and LoRA merge
-lora_on_cpu: true
-
-# List[str]. Add plugins to extend the pipeline.
-# See `src/axolotl/integrations` for the available plugins or doc below for more details.
-# https://docs.axolotl.ai/docs/custom_integrations.html
-plugins:
-  # - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
-
-# A list of one or more datasets to finetune the model with
-datasets:
-  # HuggingFace dataset repo | s3://,gs:// path | "json" for local dataset, make sure to fill data_files
-  - path: vicgalle/alpaca-gpt4
-    # The type of prompt to use for training. [alpaca, gpteacher, oasst, reflection]
-    type: alpaca # format | format:<prompt_style> (chat/instruct) | <prompt_strategies>.load_<load_fn>
-    ds_type: # Optional[str] (json|arrow|parquet|text|csv) defines the datatype when path is a file
-    data_files: # Optional[str] path to source data files
-
-    shards: # Optional[int] split dataset into N pieces (use with shards_idx)
-    shards_idx: # Optional[int] = 0 the index of sharded dataset to use
-
-    preprocess_shards: # Optional[int] process dataset in N sequential chunks for memory efficiency (exclusive with `shards`)
-
-    name: # Optional[str] name of dataset configuration to load
-    split: train # Optional[str] name of dataset split to load from
-    revision: # Optional[str] The specific revision of the dataset to use when loading from the Hugging Face Hub. This can be a commit hash, tag, or branch name. If not specified, the latest version will be used. This parameter is ignored for local datasets.
-    trust_remote_code: # Optional[bool] Trust remote code for untrusted source
-
-  # Custom user instruction prompt
-  - path: repo
-    type:
-      # The below are defaults. only set what's needed if you use a different column name.
-      system_prompt: ""
-      system_format: "{system}"
-      field_system: system
-      field_instruction: instruction
-      field_input: input
-      field_output: output
-
-      # Customizable to be single line or multi-line
-      # Use {instruction}/{input} as key to be replaced
-      # 'format' can include {input}
-      format: |-
-        User: {instruction} {input}
-        Assistant:
-      # 'no_input_format' cannot include {input}
-      no_input_format: "{instruction} "
-
-      # For `completion` datsets only, uses the provided field instead of `text` column
-      field:
-
-  # Using chat template
-  - path: ...
-    # Set type to `chat_template` to use this strategy
-    type: chat_template
-    # Specify the name of the chat template to use
-    # The name of the chat template to use for training, following values are supported:
-    # - tokenizer_default: Uses the chat template that is available in the tokenizer_config.json. If the chat template is not available in the tokenizer, it will raise an error. This is the default.
-    # - alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates are available in the axolotl codebase at src/axolotl/utils/chat_templates.py
-    # - tokenizer_default_fallback_*: where * is the name of the chat template to fallback to if the tokenizer does not have a chat template else default to tokenizer. E.g. tokenizer_default_fallback_chatml.
-    # - jinja: Uses a custom jinja template for the chat template. The custom jinja template should be provided in the chat_template_jinja field.
-    chat_template: tokenizer_default
-
-    # Custom jinja chat template. Used only if `chat_template: jinja` or empty.
-    chat_template_jinja:
-
-    # Key containing the messages (default: "messages")
-    field_messages: messages
-
-    # Key containing the system message (default: "system")
-    # If the system message is not present in the dataset sample, it will be loaded from the field_system property.
-    field_system: system
-
-    # Mapping of properties from the input dataset to the chat template.
-    # (default: message_property_mappings={'role':'role', 'content':'content'})
-    # If a property exists in the template but not in this mapping, the system will attempt
-    # to load it directly from the message using the property name as the key.
-    # Example: In the mapping below, 'from' is loaded from input dataset and used as 'role',
-    # while 'value' is loaded and used as 'content' in the chat template.
-    message_property_mappings:
-      role: from
-      content: value
-      # ...
-
-    # Optional[Dict[str, List]]. Roles mapping in the messages.
-    # The format is {target_role: [source_roles]}. All source roles will be mapped to the target role.
-    # The default is:
-    roles:
-      user: ["human", "user"]
-      assistant: ["gpt", "assistant"]
-      system: ["system"]
-      tool: ["tool"]
-
-    # Optional[bool]. Whether to drop the system turn from the dataset. Only works with chat_template.
-    # This does not drop the default system message from chat_template if it exists. If you wish to,
-    # we recommend using a custom jinja template with the default system message removed or
-    # adding a system turn with empty content.
-    drop_system_message:
-
-    # Optional[bool]. (for Qwen3 template only) Whether to split the assistant content based on a reasoning trace inside delimited tags
-    # See example at `docs/dataset-formats/conversation.qmd`
-    split_thinking:
-
-    # IMPORTANT: The following fields determine which parts of the conversation to train on.
-    # Priority order: message_field_training > message_field_training_detail > train_on_inputs or role in roles_to_train
-    # See examples at `docs/dataset-formats/conversation.qmd`
-    # Note: If the below 5 fields are empty, defaults to training only on the last message.
-
-    # Optional[List[str]]. Roles to train on. The tokens from these roles will be considered for the loss.
-    roles_to_train: ["assistant"]  # default
-    # Optional[str]. Which EOS tokens to train on in the conversation. Possible values are:
-    # - all: train on all EOS tokens
-    # - turn (default): train on the EOS token at the end of each trainable turn
-    # - last: train on the last EOS token in the conversation
-    # TIP: Please make sure that your `tokenizer.eos_token` is same as EOS/EOT token in template. Otherwise, set `eos_token` under `special_tokens`.
-    train_on_eos: turn
-    # Optional[str]. Which EOT (End-of-Turn) tokens to train on in the conversation. Possible values are:
-    # - all: train on all EOT tokens
-    # - turn: train on the EOT token at the end of each trainable turn
-    # - last: train on the last EOT token in the conversation
-    # If not specified, defaults to the value of train_on_eos for backward compatibility.
-    train_on_eot:
-    # The key in the message turn that indicates via boolean whether tokens of a turn should be considered for training. Useful to selectively train on certain turns besides the `roles_to_train`.
-    message_field_training: training
-    # The key in the message turn that contains the training details. Useful to selectively train on certain tokens in a turn.
-    # The value of the key is a List[Dict] containing `begin_offset` (start character index in content), `end_offset` (end character index in content), and `train` (boolean whether to train).
-    message_field_training_detail: train_detail
-
-
-# If false, the datasets will not be shuffled and will keep their original order in `datasets`.
-# The same applies to the `test_datasets` option and the `pretraining_dataset` option. Default is true.
-shuffle_merged_datasets: true
-
-Deduplicates datasets and test_datasets with identical entries.
-dataset_exact_deduplication: true
-
-# A list of one or more datasets to eval the model with.
-# You can use either test_datasets, or val_set_size, but not both.
-test_datasets:
-  - path: /workspace/data/eval.jsonl
-    ds_type: json
-    # You need to specify a split. For "json" datasets the default split is called "train".
-    split: train
-    type: completion
-    data_files:
-      - /workspace/data/eval.jsonl
-
-# use RL training: 'dpo', 'ipo', 'kto', 'simpo', 'orpo', 'grpo'
-rl:
-rl_beta:  # Optional[float]. The beta parameter for the RL training.
-
-# dpo
-dpo_use_weighting:  # Optional[bool]. Whether to perform weighting.
-rpo_alpha: # Optional[float]. Weighting of NLL term in loss from RPO paper.
-
-# orpo
-orpo_alpha: 0.1  # Parameter controlling the relative ratio loss weight in the ORPO loss. Passed to `beta` in `ORPOConfig` due to trl mapping.
-
-# kto
-kto_desirable_weight: # Optional[float]. Factor for desirable loss term in KTO loss.
-kto_undesirable_weight: # Optional[float]. Factor for undesirable loss term in KTO loss.
-
-# simpo
-cpo_alpha: 1.0  # Weight of the BC regularizer
-simpo_gamma: 0.5  # Target reward margin for the SimPO loss
-
-# grpo
-trl:
-  use_vllm: # Optional[bool]. Whether to use VLLM for RL training.
-  vllm_server_host: # Optional[str]. Host of the vLLM server to connect to.
-  vllm_server_port: # Optional[int]. Port of the vLLM server to connect to.
-  vllm_server_timeout: # Optional[int]. Total timeout (in seconds) to wait for the vLLM server to respond.
-  vllm_guided_decoding_regex: # Optional[str]. Regex for vLLM guided decoding.
-
-  beta: # Optional[float]. Beta parameter for the RL training. Same as `rl_beta`. Use
-  max_completion_length: # Optional[int]. Maximum length of the completion for RL training.
-
-  reward_funcs: # Optional[list[str]]. List of reward functions to load. Paths must be importable from current dir.
-  reward_weights: # Optional[list[float]]. List of reward weights for the reward functions.
-
-  num_generations: # Optional[int]. Number of generations to sample.
-  log_completions: # Optional[bool]. Whether to log completions.
-
-  sync_ref_model: # Optional[bool]. Whether to sync the reference model.
-  ref_model_mixup_alpha: # Optional[float]. Mixup alpha for the reference model.
-  ref_model_sync_steps: # Optional[int]. Sync steps for the reference model.
-
-
-# reward modelling: `True` or `False`
-reward_model:
-
-# process reward modelling: `True` or `False`
-process_reward_model:
-
-# The name of the chat template to use for training, following values are supported:
-# - tokenizer_default: Uses the chat template that is available in the tokenizer_config.json. If the chat template is not available in the tokenizer, it will raise an error. This is the default value.
-# - alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates are available in the axolotl codebase at src/axolotl/utils/chat_templates.py
-# - tokenizer_default_fallback_*: where * is the name of the chat template to fallback to. E.g. tokenizer_default_fallback_chatml. This is useful when the chat template is not available in the tokenizer.
-# - jinja: Uses a custom jinja template for the chat template. The custom jinja template should be provided in the chat_template_jinja field.
-# The selected chat template will be saved to the tokenizer_config.json for easier inferencing
-# Note: It is recommended to set train_on_inputs to true when using a chat template that is different from the model's default chat template.
-chat_template: tokenizer_default
-# custom jinja template for chat template. This will be only used if chat_template is set to `jinja` or `null` (in which case chat_template is automatically set to `jinja`). Default is null.
-chat_template_jinja: null
-# Optional[List[str]]. Custom EOT (End-of-Turn) tokens to mask/unmask during training.
-# These tokens mark the boundaries between conversation turns.
-# For example: ["/INST", "</s>", "[/SYSTEM_PROMPT]"]
-# If not specified, defaults to just the model's eos_token.
-# This is useful for templates that use multiple delimiter tokens.
-eot_tokens:
-  # - "</s>"
-  # - "[/INST]"
-  # - "[/SYSTEM_PROMPT]"
-# Changes the default system message
-default_system_message: You are a helpful assistant. Please give a long and detailed answer. # Currently only supports chatml.
-# Axolotl attempts to save the dataset as an arrow after packing the data together so
-# subsequent training attempts load faster, relative path
-dataset_prepared_path: data/last_run_prepared
-# Push prepared dataset to hub
-push_dataset_to_hub: # Optional[str] repo_org/repo_name
-# The maximum number of processes to use while preprocessing your input dataset. This defaults to `os.cpu_count()`
-# if not set.
-dataset_processes: # defaults to os.cpu_count() if not set
-# Keep dataset in memory while preprocessing
-# Only needed if cached dataset is taking too much storage
-dataset_keep_in_memory:
-# push checkpoints to hub
-hub_model_id: # private repo path to push finetuned model
-# how to push checkpoints to hub
-# https://huggingface.co/docs/transformers/v4.31.0/en/main_classes/trainer#transformers.TrainingArguments.hub_strategy
-hub_strategy:
-# Whether to use hf `use_auth_token` for loading datasets. Useful for fetching private datasets
-# Required to be true when used in combination with `push_dataset_to_hub`
-hf_use_auth_token: # boolean
-# How much of the dataset to set aside as evaluation. 1 = 100%, 0.50 = 50%, etc. 0 for no eval.
-val_set_size: 0.04
-# Num shards for whole dataset
-dataset_shard_num:
-# Index of shard to use for whole dataset
-dataset_shard_idx:
-
-# The maximum length of an input to train with, this should typically be less than 2048
-# as most models have a token/context limit of 2048
-sequence_len: 2048
-# Pad inputs so each step uses constant sized buffers
-# This will reduce memory fragmentation and may prevent OOMs, by re-using memory more efficiently
-pad_to_sequence_len:
-# Use efficient multi-packing with block diagonal attention and per sequence position_ids. Recommend set to 'true'
-sample_packing:
-# Set to 'false' if getting errors during eval with sample_packing on.
-eval_sample_packing:
-# You can set these packing optimizations AFTER starting a training at least once.
-# The trainer will provide recommended values for these values.
-sample_packing_eff_est:
-total_num_tokens:
-# Increasing the following values helps with packing, but usually only slightly (<%1.)
-# The number of samples packed at a time.
-sample_packing_group_size: 100000
-# The number of samples which can be packed into one sequence. Increase if using a large sequence_len with many short samples.
-sample_packing_bin_size: 200
-sample_pack_sequentially: # Optional[bool]. Whether to pack samples sequentially.
-
-# whether to concatenate samples during pretraining
-pretraining_sample_concatenation:
-
-curriculum_sampling: # Optional[bool]. Whether to use sequential sampling for curriculum learning
-
-# Use batch flattening for speedups when not using sample_packing
-batch_flattening:
-
-# Passed through to transformers when loading the model when launched without accelerate
-# Use `sequential` when training w/ model parallelism to limit memory
-device_map:
-# Defines the max memory usage per gpu on the system. Passed through to transformers when loading the model.
-max_memory:
-
-# If you want to use 'lora' or 'qlora' or leave blank to train all parameters in original model
-adapter: lora
-# If you already have a lora model trained that you want to load, put that here.
-# This means after training, if you want to test the model, you should set this to the value of `output_dir`.
-# Note that if you merge an adapter to the base model, a new subdirectory `merged` will be created under the `output_dir`.
-lora_model_dir:
-
-# LoRA hyperparameters
-# For more details about the following options, see:
-# https://www.anyscale.com/blog/fine-tuning-llms-lora-or-full-parameter-an-in-depth-analysis-with-llama-2
-lora_r: 8
-lora_alpha: 16
-lora_dropout: 0.05
-lora_target_modules:
-  - q_proj
-  - v_proj
-#  - k_proj
-#  - o_proj
-#  - gate_proj
-#  - down_proj
-#  - up_proj
-lora_target_linear: # If true, will target all linear modules
-
-# List[int] | int. # The layer indices to transform, otherwise, apply to all layers
-# https://huggingface.co/docs/peft/v0.15.0/en/package_reference/lora#peft.LoraConfig.layers_to_transform
-peft_layers_to_transform:
-
-# Optional[bool]. Whether to use DoRA.
-# https://huggingface.co/docs/peft/v0.15.0/en/developer_guides/lora#weight-decomposed-low-rank-adaptation-dora
-peft_use_dora:
-
-# Optional[bool]. Whether to use RSLoRA.
-# https://huggingface.co/docs/peft/v0.15.0/en/developer_guides/lora#rank-stabilized-lora
-peft_use_rslora:
-
-# Optional[list[tuple[int, int]]]. List of layer indices to replicate.
-# https://huggingface.co/docs/peft/v0.15.0/en/developer_guides/lora#memory-efficient-layer-replication-with-lora
-peft_layer_replication:
-
-# bool | Literal["gaussian", "eva", "olora", "pissa", "pissa_niter_[number of iters]", "corda", "loftq"]
-# How to initialize LoRA weights. Default to True which is MS original implementation.
-# https://huggingface.co/docs/peft/v0.15.0/en/developer_guides/lora#initialization
-peft_init_lora_weights:
-
-# If you added new tokens to the tokenizer, you may need to save some LoRA modules because they need to know the new tokens.
-# For LLaMA and Mistral, you need to save `embed_tokens` and `lm_head`. It may vary for other models.
-# `embed_tokens` converts tokens to embeddings, and `lm_head` converts embeddings to token probabilities.
-# https://github.com/huggingface/peft/issues/334#issuecomment-1561727994
-lora_modules_to_save:
-#  - embed_tokens
-#  - lm_head
-
-lora_fan_in_fan_out: false
-
-# Apply custom LoRA autograd functions and activation function Triton kernels for
-# speed and memory savings
-# See: https://docs.axolotl.ai/docs/lora_optims.html
-lora_mlp_kernel: true
-lora_qkv_kernel: true
-lora_o_kernel: true
-
-# LoRA+ hyperparameters
-# For more details about the following options, see:
-# https://arxiv.org/abs/2402.12354  and `src/axolotl/core/train_builder.py`
-loraplus_lr_ratio: # loraplus learning rate ratio lr_B / lr_A. Recommended value is 2^4.
-loraplus_lr_embedding: #  loraplus learning rate for lora embedding layers. Default value is 1e-6.
-
-peft:
-  # Configuration options for loftq initialization for LoRA
-  # https://huggingface.co/docs/peft/developer_guides/quantization#loftq-initialization
-  loftq_config:
-    loftq_bits:  # typically 4 bits
-
-# ReLoRA configuration
-# Must use either 'lora' or 'qlora' adapter, and does not support fsdp or deepspeed
-relora_steps: # Number of steps per ReLoRA restart
-relora_warmup_steps: # Number of per-restart warmup steps
-relora_anneal_steps: # Number of anneal steps for each relora cycle
-relora_prune_ratio: # threshold for optimizer magnitude when pruning
-relora_cpu_offload: # True to perform lora weight merges on cpu during restarts, for modest gpu memory savings
-
-# wandb configuration if you're using it
-# Make sure your `WANDB_API_KEY` environment variable is set (recommended) or you login to wandb with `wandb login`.
-wandb_mode: # "offline" to save run metadata locally and not sync to the server, "disabled" to turn off wandb
-wandb_project: # Your wandb project name
-wandb_entity: # A wandb Team name if using a Team
-wandb_watch:
-wandb_name: # Set the name of your wandb run
-wandb_run_id: # Set the ID of your wandb run
-wandb_log_model: # "checkpoint" to log model to wandb Artifacts every `save_steps` or "end" to log only at the end of training
-
-# mlflow configuration if you're using it
-mlflow_tracking_uri: # URI to mlflow
-mlflow_experiment_name: # Your experiment name
-mlflow_run_name: # Your run name
-hf_mlflow_log_artifacts:  # set to true to copy each saved checkpoint on each save to mlflow artifact registry
-
-# Comet configuration if you're using it
-# Make sure your `COMET_API_KEY` environment variable is set (recommended) or you login to Comet with `comet login`.
-# Check out our documentation for more details https://www.comet.com/docs/v2/api-and-sdk/python-sdk/reference/Experiment-Creation/#comet_ml.start
-use_comet: # Enable or disable Comet integration.
-comet_api_key: # API key for Comet. Recommended to set via `comet login`.
-comet_workspace: # Workspace name in Comet. Defaults to the user's default workspace.
-comet_project_name: # Project name in Comet. Defaults to Uncategorized.
-comet_experiment_key: # Identifier for the experiment. Used to append data to an existing experiment or control the key of new experiments. Default to a random key.
-comet_mode: # Create a new experiment ("create") or log to an existing one ("get"). Default ("get_or_create") auto-selects based on configuration.
-comet_online: # Set to True to log data to Comet server, or False for offline storage. Default is True.
-comet_experiment_config: # Dictionary for additional configuration settings, see the doc for more details.
-
-# Tensorboard
-use_tensorboard: # Optional[bool]
-
-# Where to save the full-finetuned model to
-output_dir: ./completed-model
-
-# Whether to use torch.compile and which backend to use
-# setting to `auto` will enable torch compile when torch>=2.5.1
-torch_compile:  # Optional[Union[Literal["auto"], bool]]
-torch_compile_backend:  # Optional[str]
-
-# Training hyperparameters
-
-# If greater than 1, backpropagation will be skipped and the gradients will be accumulated for the given number of steps.
-gradient_accumulation_steps: 1
-# The number of samples to include in each batch. This is the number of samples sent to each GPU.
-# Batch size per gpu = micro_batch_size * gradient_accumulation_steps
-micro_batch_size: 2
-eval_batch_size:
-num_epochs: 4
-warmup_steps: 100  # cannot use with warmup_ratio
-warmup_ratio: 0.05  # cannot use with warmup_steps
-learning_rate: 0.00003
-lr_quadratic_warmup:
-logging_steps:
-eval_steps: # Leave empty to eval at each epoch, integer for every N steps. float for fraction of total steps
-evals_per_epoch: # number of times per epoch to run evals, mutually exclusive with eval_steps
-eval_strategy: # Set to `"no"` to skip evaluation, `"epoch"` at end of each epoch, leave empty to infer from `eval_steps`.
-save_strategy: # Set to `"no"` to skip checkpoint saves, `"epoch"` at end of each epoch, `"best"` when better result is achieved, leave empty to infer from `save_steps`.
-save_steps: # Leave empty to save at each epoch, integer for every N steps. float for fraction of total steps
-saves_per_epoch: # number of times per epoch to save a checkpoint, mutually exclusive with save_steps
-save_total_limit: # Checkpoints saved at a time
-# Maximum number of iterations to train for. It precedes num_epochs which means that
-# if both are set, num_epochs will not be guaranteed.
-# e.g., when 1 epoch is 1000 steps => `num_epochs: 2` and `max_steps: 100` will train for 100 steps
-max_steps:
-
-# bool of whether to include tokens trainer per second in the training metrics. This iterates over the entire dataset once, so it takes some time.
-include_tokens_per_second: # Optional[bool]
-
-# whether to find batch size that fits in memory. Passed to underlying transformers Trainer
-auto_find_batch_size: # Optional[bool]
-
-eval_table_size: # Approximate number of predictions sent to wandb depending on batch size. Enabled above 0. Default is 0
-eval_max_new_tokens: # Total number of tokens generated for predictions sent to wandb. Default is 128
-do_causal_lm_eval: # Whether to run causal language model evaluation for metrics in `eval_causal_lm_metrics`.
-eval_causal_lm_metrics: # HF evaluate metrics used during evaluation. Default is ["sacrebleu", "comet", "ter", "chrf", "perplexity"]
-
-profiler_steps: # enable the pytorch profiler to capture the first N steps of training to the output_dir.
-                # see https://pytorch.org/blog/understanding-gpu-memory-1/ for more information
-                # snapshots can be visualized @ https://pytorch.org/memory_viz
-
-loss_watchdog_threshold: # High loss value, indicating the learning has broken down (a good estimate is ~2 times the loss at the start of training)
-loss_watchdog_patience: # Number of high-loss steps in a row before the trainer aborts (default: 3)
-
-# Save model as safetensors (require safetensors package)
-save_safetensors:
-
-# Whether to mask out or include the human's prompt from the training labels
-train_on_inputs: false
-# Group similarly sized data to minimize padding.
-# May be slower to start, as it must download and sort the entire dataset.
-# Note that training loss may have an oscillating pattern with this enabled.
-group_by_length: false
-
-# Whether to use gradient checkpointing. Available options are: true, false, "offload".
-# https://huggingface.co/docs/transformers/v4.18.0/en/performance#gradient-checkpointing
-gradient_checkpointing: false
-# additional kwargs to pass to the trainer for gradient checkpointing
-# gradient_checkpointing_kwargs:
-#   use_reentrant: true
-
-# Stop training after this many evaluation losses have increased in a row
-# https://huggingface.co/transformers/v4.2.2/_modules/transformers/trainer_callback.html#EarlyStoppingCallback
-early_stopping_patience: 3
-
-# Specify a scheduler and kwargs to use with the optimizer
-lr_scheduler: # 'one_cycle' | 'rex' | 'log_sweep' | 'linear' | 'cosine_with_restarts' | 'polynomial' | 'constant' | 'constant_with_warmup' | 'inverse_sqrt' | 'reduce_lr_on_plateau' | 'cosine_with_min_lr' | 'warmup_stable_decay' | empty for cosine
-lr_scheduler_kwargs:
-cosine_min_lr_ratio: # decay lr to some percentage of the peak lr, e.g. cosine_min_lr_ratio=0.1 for 10% of peak lr
-cosine_constant_lr_ratio: # freeze lr at some percentage of the step, e.g. cosine_constant_lr_ratio=0.8 means start cosine_min_lr at 80% of training step (https://arxiv.org/pdf/2308.04014.pdf)
-
-# For one_cycle optim
-lr_div_factor: # Learning rate div factor
-
-# Specify optimizer
-# Valid values are driven by the Transformers OptimizerNames class, see:
-# https://github.com/huggingface/transformers/blob/cbf924b76c03828101a34069a96d209314114fd5/src/transformers/training_args.py#L144-L189
-#
-# Note that not all optimizers may be available in your environment, ex: 'adamw_anyprecision' is part of
-# torchdistx, 'adamw_bnb_8bit' is part of bnb.optim.Adam8bit, etc. When in doubt, it is recommended to start with the optimizer used
-# in the examples/ for your model and fine-tuning use case.
-#
-# Valid values for 'optimizer' include:
-# - adamw_torch
-# - adamw_torch_fused
-# - adamw_torch_xla
-# - adamw_torch_npu_fused
-# - adamw_apex_fused
-# - adopt_adamw  (an EXPERIMENTAL optimizer, only for torch version >= 2.5.1)
-# - adafactor
-# - adamw_anyprecision
-# - adamw_torch_4bit
-# - ademamix
-# - sgd
-# - adagrad
-# - adamw_bnb_8bit
-# - adamw_8bit   # alias for adamw_bnb_8bit
-# - ademamix_8bit
-# - lion_8bit
-# - lion_32bit
-# - paged_adamw_32bit
-# - paged_adamw_8bit
-# - paged_ademamix_32bit
-# - paged_ademamix_8bit
-# - paged_lion_32bit
-# - paged_lion_8bit
-# - rmsprop
-# - rmsprop_bnb
-# - rmsprop_bnb_8bit
-# - rmsprop_bnb_32bit
-# - galore_adamw
-# - galore_adamw_8bit
-# - galore_adafactor
-# - galore_adamw_layerwise
-# - galore_adamw_8bit_layerwise
-# - galore_adafactor_layerwise
-# - lomo
-# - adalomo
-# - grokadamw
-# - schedule_free_adamw
-# - schedule_free_sgd
-# - apollo_adamw
-# - apollo_adamw_layerwise
-#
-# Additional custom optimizers include:
-# - optimi_adamw
-# - ao_adamw_8bit
-# - ao_adamw_fp8
-# - came_pytorch
-optimizer:
-# Dictionary of arguments to pass to the optimizer
-optim_args:
-# For Galore Optimizers the following optim_args are available
-# rank:  # type: int
-# update_proj_gap  # type: int
-# scale  # type: float
-# proj_type:  # type: str, default = std
-
-# The target modules to optimize, i.e. the module names that you would like to train, right now this is used only for GaLore algorithm
-optim_target_modules:
-# - self_attn  # for llama
-# - mlp
-
-# Specify weight decay
-weight_decay:
-# adamw hyperparams
-adam_beta1:
-adam_beta2:
-adam_epsilon:
-# Gradient clipping max norm
-max_grad_norm:
-
-# Augmentation techniques
-# NEFT https://arxiv.org/abs/2310.05914, set this to a number (paper default is 5) to add noise to embeddings
-# currently only supported on Llama and Mistral
-neftune_noise_alpha:
-
-# Optional[bool]. Whether to bettertransformers
-flash_optimum:
-
-# Note: Only one of the following attention patches can be used at a time.
-# For example, if you set `xformers_attention` to `true`, do not set `flash_attention` to `true`.
-
-# Optional[bool]. Whether to use xformers attention patch https://github.com/facebookresearch/xformers:
-xformers_attention:
-# Optional[bool]. Whether to use flash attention patch https://github.com/Dao-AILab/flash-attention:
-flash_attention:
-flash_attn_cross_entropy:  # Optional[bool]. Whether to use flash-attention cross entropy implementation - advanced use only
-flash_attn_rms_norm:  # Optional[bool]. Whether to use flash-attention rms norm implementation - advanced use only
-flash_attn_fuse_qkv: # Optional[bool]. Whether to fuse QKV into a single operation
-flash_attn_fuse_mlp: # Optional[bool]. Whether to fuse part of the MLP into a single operation
-# Optional[bool]. Whether to use scaled-dot-product attention
-# https://pytorch.org/docs/stable/generated/torch.nn.functional.scaled_dot_product_attention.html
-sdp_attention:
-# Optional[bool]. Shifted-sparse attention (only llama) - https://arxiv.org/pdf/2309.12307.pdf
-s2_attention:
-
-# Optional[bool]. Whether to use low_cpu_mem_usage
-low_cpu_mem_usage:
-# Optional[str]. Resume from a specific checkpoint dir
-resume_from_checkpoint:
-# Optional[bool]. If resume_from_checkpoint isn't set and you simply want it to start where it left off.
-# Be careful with this being turned on between different models.
-auto_resume_from_checkpoints: false
-
-## Multimodal section
-# int | tuple[int, int] | None . Size to resize images to, width x height.
-# Will read from model/processor config if not set.
-image_size:
-# str. Algorithm to use for image resizing. "bilinear", "bicubic", "lanczos". Default is "bilinear".
-image_resize_algorithm: 'bilinear'
-## End of multimodal section
-
-# Don't mess with this, it's here for accelerate and torchrun
-local_rank:
-
-# Add or change special tokens.
-# If you add tokens here, you don't need to add them to the `tokens` list.
-special_tokens:
-  # bos_token: "<s>"
-  # eos_token: "</s>"
-  # unk_token: "<unk>"
-  # pad_token: "[PAD]"
-
-# Optional[list[str]]. Add extra tokens to the tokenizer.
-tokens:
-  # - "<|startoftext|>"
-  # - "<|endoftext|>"
-
-# Mapping token_id to new_token_string to override reserved added_tokens in the tokenizer.
-# Only works for tokens that are not part of the base vocab (aka are added_tokens).
-# Can be checked if they exist in tokenizer.json added_tokens.
-added_tokens_overrides:  # Dict[int, str]
-#  128041: "<|im_start|>"
-#  128042: "<|im_end|>"
-
-# FSDP
-fsdp:
-fsdp_config:
-
-# Deepspeed config path. e.g., deepspeed_configs/zero3.json
-deepspeed:
-
-# Advanced DDP Arguments
-ddp_timeout:
-ddp_bucket_cap_mb:
-ddp_broadcast_buffers:
-
-# Sequence parallelism
-# Set to a divisor of the number of GPUs available to split sequences into chunks of equal size.
-# Use in long context training to prevent OOM when sequences cannot fit into a single GPU's VRAM.
-# E.g., if 4 GPUs are available, set this value to 2 to split each sequence into two equal-sized
-# subsequences, or set to 4 to split into four equal-sized subsequences.
-# See https://docs.axolotl.ai/docs/sequence_parallelism.html for more details.
-sequence_parallel_degree:
-# Optional; strides across the key dimension. Larger values use more memory but should make training faster.
-# Must evenly divide the number of KV heads in your model.
-heads_k_stride: 1
-# One of "varlen_llama3", "batch_ring", "batch_zigzag", "batch_stripe". Defaults to "varlen_llama3"
-# in the sample packing case, and "batch_ring" in the non-sample packing case.
-ring_attn_func:
-
-# Path to torch distx for optim 'adamw_anyprecision'
-torchdistx_path:
-
-# Set to HF dataset for type: 'completion' for streaming instead of pre-tokenize
-pretraining_dataset:
-
-# Debug mode
-debug:
-
-# Seed
-seed:
-
-# Allow overwrite yml config using from cli
-strict:
-```

docs/dataset-formats/conversation.qmd

@@ -9,10 +9,10 @@ order: 3
 Chat Template strategy uses a jinja2 template that converts a list of messages into a prompt. Support using tokenizer's template, a supported template, or custom jinja2.
 
 ```{.json filename="data.jsonl"}
-{"conversations": [{"role": "...", "content": "..."}]}
+{"messages": [{"role": "...", "content": "..."}, {"role": "...", "content": "..."}, ...]}
 ```
 
-See [configs](../config.qmd) for full configs and supported templates.
+See [configs](../config-reference.qmd) for full configs and supported templates.
 
 ### Migrating from sharegpt
 
@@ -52,7 +52,9 @@ We recommend checking the below examples for other usecases.
 
 ### Examples
 
-1. (Legacy) Using the default chat template in the tokenizer_config.json on OpenAI messages format, training on only last message.
+#### Training on last message
+
+(Legacy) Using the default chat template in the tokenizer_config.json on OpenAI messages format, training on only last message.
 
 ```yaml
 datasets:
@@ -66,7 +68,9 @@ datasets:
 If you receive an error like "`chat_template` choice is `tokenizer_default` but tokenizer's `chat_template` is null.", it means the tokenizer does not have a default `chat_template`. Follow the examples below instead to set a custom `chat_template`.
 :::
 
-2. Using the `gemma` chat template to override the tokenizer_config.json's chat template on OpenAI messages format, training on all assistant messages.
+#### Overriding default chat template
+
+Using the `gemma` chat template to override the tokenizer_config.json's chat template on OpenAI messages format, training on all assistant messages.
 
 ```yaml
 chat_template: gemma # this overwrites the tokenizer's chat_template
@@ -76,7 +80,13 @@ datasets:
     roles_to_train: ["assistant"]  # default value
 ```
 
-3. Using the tokenizer_config.json's chat template or `chatml` as fallback if the former's chat template does not exist, on OpenAI messages format, training on all assistant messages.
+::: {.callout-note}
+If you want to use built-in chat_template, use `chat_template: tokenizer_default` (this is set by default).
+:::
+
+#### Using default chat template with fallback
+
+Using the tokenizer_config.json's chat template or `chatml` as fallback if the former's chat template does not exist, on OpenAI messages format, training on all assistant messages.
 
 ```yaml
 chat_template: tokenizer_default_fallback_chatml # this overwrites the tokenizer's chat_template
@@ -85,7 +95,9 @@ datasets:
     type: chat_template
 ```
 
-4. Using a custom jinja template on OpenAI messages format, training on all assistant messages.
+#### Custom Jinja template
+
+Using a custom jinja template on OpenAI messages format, training on all assistant messages.
 
 ```yaml
 # chat_template: jinja # `jinja` will be implied if the `chat_template_jinja` is set and this field is empty
@@ -100,7 +112,9 @@ datasets:
 Please make sure that your `tokenizer.eos_token` is same as EOS (End-of-Sequence) token in template. Otherwise, set `eos_token` under `special_tokens: `.
 :::
 
-5. If you are using a template that has a different EOT (End-of-Turn) token from EOS token or multiple EOT tokens (like Mistral V7 Tekken), set the `eot_tokens: ` config. The handling of EOT tokens follows `train_on_eos: ` which defaults to turn.
+#### Using template with different token for EOT and EOS
+
+- If you are using a template that has a different EOT (End-of-Turn) token from EOS token or multiple EOT tokens (like Mistral V7 Tekken), set the `eot_tokens: ` config. The handling of EOT tokens follows `train_on_eos: ` which defaults to turn.
 
 ```yaml
 eot_tokens:
@@ -116,16 +130,16 @@ datasets:
 ```
 
 ::: {.callout-tip}
-See [config documentation](../config.qmd) for detailed explanations of "turn", "last", and "all" options for training on tokens.
+See [config documentation](../config-reference.qmd) for detailed explanations of "turn", "last", and "all" options for training on tokens.
 :::
 
 ::: {.callout-note}
 Using `eot_tokens` requires each token that exists in `chat_template` to be a single token in the tokenizer. Otherwise, the tokenizer will split the token and cause unexpected behavior.
 
-You can add those tokens as new tokens under `tokens: ` or (recommended) override unused added_tokens via `added_tokens_overrides: `. See [config](../config.qmd) for more details.
+You can add those tokens as new tokens under `tokens: ` or (recommended) override unused added_tokens via `added_tokens_overrides: `. See [config](../config-reference.qmd) for more details.
 :::
 
-6. Continuing from the previous example, if you want to train on all EOT token trainable turns but only last EOS token, set `train_on_eos: last`.
+- Continuing from the previous example, if you want to train on all EOT token trainable turns but only last EOS token, set `train_on_eos: last`.
 
 ```yaml
 eot_tokens:
@@ -145,7 +159,73 @@ If EOS token only appears at the end of a prompt, `train_on_eos: last` is equiva
 :::
 
 
-7. (Advanced) Using fine-grained control over tokens and turns to train in a conversation
+#### Using tool use
+
+Instead of passing `tools` via the system prompt, an alternative method would be to have the `tools` in a separate column and loaded via `chat_template` to let the template dynamically build it.
+
+```json
+{
+    "tools": [
+        {
+            "type": "...",
+            "function": {
+                "name": "...",
+                "description": "...",
+                "parameters": {
+                    "type": "...",
+                    "properties": {
+                        // ...
+                    },
+                    "required": ["..."],
+                },
+            },
+        },
+    ],
+    "messages": [
+        // ...
+        {
+            "role": "assistant", // call the function via assistant
+            "tool_calls": [
+                {
+                    "type": "function",
+                    "function": {
+                        "name": "...",
+                        "arguments": {
+                            "...": "...",
+                        }
+                    }
+                }
+            ]
+        },
+        {
+            "role": "tool",
+            "name": "...",
+            "content": "..."
+        },
+    ],
+}
+```
+
+::: {.callout-note}
+Tools need to follow [JSON schema](https://json-schema.org/learn/getting-started-step-by-step).
+:::
+
+```yaml
+chat_template: llama4
+datasets:
+  - path: ...
+    type: chat_template
+    # field_tools: tools # default is `tools`
+```
+
+::: {.callout-tip}
+Look into the `chat_template` you are using to see if it supports `tools` and what the expected role is for the tool answer. In the example above, the tool answer is expected to be in the `tool` or `ipython` role for `llama4` template.
+:::
+
+
+#### Using fine-grained control over token masking
+
+(Advanced) Using fine-grained control over tokens and turns to train in a conversation
 
 For a data sample that looks like:
 
@@ -196,7 +276,9 @@ datasets:
 It is not necessary to set both `message_field_training` and `message_field_training_detail` at once.
 :::
 
-8. (For Qwen3 template only) Enable reasoning split, where the reasoning is split from the content and passed as a separate field into the template.
+#### Reasoning split
+
+(For Qwen3 template only) Enable reasoning split, where the reasoning is split from the content and passed as a separate field into the template.
 
 ```yaml
 datasets:

docs/dataset-formats/index.qmd

@@ -36,10 +36,6 @@ It is typically recommended to save your dataset as `.jsonl` due to its flexibil
 
 Axolotl supports loading from a Hugging Face hub repo or from local files.
 
-::: {.callout-important}
-For pre-training only, Axolotl would split texts if it exceeds the context length into multiple smaller prompts.
-:::
-
 ### Pre-training from Hugging Face hub datasets
 
 As an example, to train using a Hugging Face dataset `hf_org/name`, you can pass the following config:
@@ -77,18 +73,21 @@ datasets:
     type: completion
 ```
 
-From local files (either example works):
+From local files:
 
 ```yaml
 datasets:
   - path: A.jsonl
     type: completion
 
-  - path: json
-    data_files: ["A.jsonl", "B.jsonl", "C.jsonl"]
+  - path: B.jsonl
     type: completion
 ```
 
+::: {.callout-important}
+For `completion` only, Axolotl would split texts if it exceeds the context length into multiple smaller prompts. If you are interested in having this for `pretraining_dataset` too, please let us know or help make a PR!
+:::
+
 ### Pre-training dataset configuration tips
 
 #### Setting max_steps

docs/dataset-formats/inst_tune.qmd

@@ -186,4 +186,4 @@ datasets:
       no_input_format: "[INST] {instruction} [/INST]"
 ```
 
-See full config options under [here](../config.qmd).
+See full config options under [here](../config-reference.qmd).

docs/dataset_loading.qmd

@@ -36,7 +36,7 @@ This matches the API of [`datasets.load_dataset`](https://github.com/huggingface
 
 For HuggingFace's guide to load different dataset types, see [here](https://huggingface.co/docs/datasets/loading).
 
-For full details on the config, see [config.qmd](config.qmd).
+For full details on the config, see [config-reference.qmd](config-reference.qmd).
 
 ::: {.callout-note}
 
@@ -54,7 +54,7 @@ datasets:
 
 #### Files
 
-Usually, to load a JSON file, you would do something like this:
+To load a JSON file, you would do something like this:
 
 ```python
 from datasets import load_dataset
@@ -66,20 +66,12 @@ Which translates to the following config:
 
 ```yaml
 datasets:
-  - path: json
-    data_files: /path/to/your/file.jsonl
-```
-
-However, to make things easier, we have added a few shortcuts for loading local dataset files.
-
-You can just point the `path` to the file or directory along with the `ds_type` to load the dataset. The below example shows for a JSON file:
-
-```yaml
-datasets:
-  - path: /path/to/your/file.jsonl
+  - path: data.json
     ds_type: json
 ```
 
+In the example above, it can be seen that we can just point the `path` to the file or directory along with the `ds_type` to load the dataset.
+
 This works for CSV, JSON, Parquet, and Arrow files.
 
 ::: {.callout-tip}

docs/docker.qmd

@@ -8,6 +8,10 @@ format:
 
 This section describes the different Docker images that are released by AxolotlAI at [Docker Hub](https://hub.docker.com/u/axolotlai).
 
+::: {.callout-important}
+For Blackwell GPUs, please use the tags with Pytorch 2.7.1 and CUDA 12.8.
+:::
+
 ## Base
 
 The base image is the most minimal image that can install Axolotl. It is based on the `nvidia/cuda` image. It includes python, torch, git, git-lfs, awscli, pydantic, and more.
@@ -28,11 +32,10 @@ main-base-py{python_version}-cu{cuda_version}-{pytorch_version}
 
 Tags examples:
 
-- `main-base-py3.11-cu128-2.7.0`
-- `main-base-py3.11-cu126-2.7.0`
+- `main-base-py3.11-cu128-2.7.1`
+- `main-base-py3.11-cu126-2.7.1`
 - `main-base-py3.11-cu124-2.6.0`
 - `main-base-py3.11-cu124-2.5.1`
-- `main-base-py3.11-cu124-2.4.1`
 
 ## Main
 
@@ -73,12 +76,10 @@ Tags examples:
 - `main-py3.11-cu126-2.7.0`
 - `main-py3.11-cu124-2.6.0`
 - `main-py3.11-cu124-2.5.1`
-- `main-py3.11-cu124-2.4.1`
 - `main-latest`
 - `main-20250303-py3.11-cu124-2.6.0`
 - `main-20250303-py3.11-cu124-2.5.1`
-- `main-20250303-py3.11-cu124-2.4.1`
-- `0.7.1`
+- `0.9.2`
 
 ## Cloud
 

docs/faq.qmd

@@ -9,11 +9,11 @@ description: Frequently asked questions
 
 > A: Usually an issue with the GPUs communicating with each other. See the [NCCL doc](nccl.qmd)
 
-**Q: Exitcode -9**
+**Q: exitcode: -9**
 
 > A: This usually happens when you run out of system RAM.
 
-**Q: Exitcode -7 while using deepspeed**
+**Q: exitcode: -7 while using deepspeed**
 
 > A: Try upgrading deepspeed w: `pip install -U deepspeed`
 
@@ -110,3 +110,17 @@ description: Frequently asked questions
 > A: If `eot_tokens: ` is not provided, the default behavior is the same as before. EOS tokens used to delimit turns are masked/unmasked depending on whether the turn is trainable.
 
 > Internally, `eot_tokens: tokenizer.eos_token` and `train_on_eot: train_on_eos` (which defaults to `turn`). This transition helps clarify the naming and behavior of EOT/EOS tokens.
+
+**Q: `Data processing error: CAS service error`**
+
+> A: Try disabling XET with `export HF_HUB_DISABLE_XET=1`
+
+**Q: `torch._inductor.exc.LoweringException: NoValidChoicesError: No choices to select, please consider adding ATEN into max_autotune_gemm_backends config (defined in torch/_inductor/config.py) to allow at least one choice. `**
+
+> A: Depending on the version of torch, you may need to include this in your YAML:
+
+> ```yaml
+> flex_attn_compile_kwargs:
+>   dynamic: false
+>   mode: max-autotune-no-cudagraphs
+> ```

docs/getting-started.qmd

@@ -55,7 +55,7 @@ output_dir: ./outputs/lora-out
 - To perform QLoRA finetuning, replace with `load_in_4bit: true` and `adapter: qlora`.
 :::
 
-See our [Config options](config.qmd) for more details.
+See our [config options](config-reference.qmd) for more details.
 
 ### Training {#sec-training}
 
@@ -104,7 +104,7 @@ the `alpaca` dataset format, which has the following format:
 Please see our [Dataset Formats](dataset-formats) for more dataset formats and how to
 format them.
 
-2. Prepare your JSONL data in the specified format (in this case, the expected `alpaca
+2. Prepare your JSONL data in the specified format (in this case, the expected `alpaca`
 format):
 
 ```json
@@ -120,6 +120,12 @@ axolotl train my_training.yml
 
 ## Common Tasks {#sec-common-tasks}
 
+::: {.callout-tip}
+
+The same yaml file is used for training, inference, and merging.
+
+:::
+
 ### Testing Your Model {#sec-testing}
 
 After training, test your model:
@@ -128,6 +134,16 @@ After training, test your model:
 axolotl inference my_training.yml --lora-model-dir="./outputs/lora-out"
 ```
 
+More details can be found in [Inference](inference.qmd).
+
+### Using a UI {#sec-ui}
+
+Launch a Gradio interface:
+
+```bash
+axolotl inference my_training.yml --lora-model-dir="./outputs/lora-out" --gradio
+```
+
 ### Preprocessing Data {#sec-preprocessing}
 
 For large datasets, preprocess first:
@@ -136,14 +152,22 @@ For large datasets, preprocess first:
 axolotl preprocess my_training.yml
 ```
 
-### Using a UI {#sec-ui}
+Please make sure to set `dataset_prepared_path: ` in your config to set the path to save the prepared dataset.
 
-Launch a Gradio interface:
+More details can be found in [Dataset Preprocessing](dataset_preprocessing.qmd).
+
+### Merging LoRA weights {#sec-merging-lora}
+
+To merge the LoRA weights back into the base model, run:
 
 ```bash
-axolotl inference my_training.yml --lora-model-dir="./outputs/lora-out" --gradio
+axolotl merge-lora my_training.yml --lora-model-dir="./outputs/lora-out"
 ```
 
+The merged model will be saved in the `{output_dir}/merged` directory.
+
+More details can be found in [Merging LoRA weights](inference.qmd#sec-merging).
+
 ## Next Steps {#sec-next-steps}
 
 Now that you have the basics, you might want to:
@@ -155,7 +179,8 @@ Now that you have the basics, you might want to:
 
 Check our other guides for details on these topics:
 
-- [Configuration Guide](config.qmd) - Full configuration options
+- [Configuration Guide](config-reference.qmd) - Full configuration options
+- [Dataset Loading](dataset_loading.qmd) - Loading datasets from various sources
 - [Dataset Formats](dataset-formats) - Working with different data formats
 - [Multi-GPU Training](multi-gpu.qmd)
 - [Multi-Node Training](multi-node.qmd)

docs/installation.qmd

@@ -14,8 +14,8 @@ This guide covers all the ways you can install and set up Axolotl for your envir
 ## Requirements {#sec-requirements}
 
 - NVIDIA GPU (Ampere architecture or newer for `bf16` and Flash Attention) or AMD GPU
-- Python ≥3.10
-- PyTorch ≥2.4.1
+- Python ≥3.11
+- PyTorch ≥2.5.1
 
 ## Installation Methods {#sec-installation-methods}
 
@@ -25,6 +25,10 @@ Please make sure to have Pytorch installed before installing Axolotl in your loc
 Follow the instructions at: [https://pytorch.org/get-started/locally/](https://pytorch.org/get-started/locally/)
 :::
 
+::: {.callout-important}
+For Blackwell GPUs, please use Pytorch 2.7.0 and CUDA 12.8.
+:::
+
 ### PyPI Installation (Recommended) {#sec-pypi}
 
 ```{.bash}
@@ -37,6 +41,40 @@ installed) in order not to clobber it, and so that we set the correct version of
 dependencies that are specific to the PyTorch version or other installed
 co-dependencies.
 
+### uv Installation {#sec-uv}
+
+uv is a fast, reliable Python package installer and resolver built in Rust. It offers significant performance improvements over pip and provides better dependency resolution, making it an excellent choice for complex environments.
+
+Install uv if not already installed
+```{.bash}
+curl -LsSf https://astral.sh/uv/install.sh | sh
+source $HOME/.local/bin/env
+```
+
+Choose your CUDA version to use with PyTorch; e.g. `cu124`, `cu126`, `cu128`,
+then create the venv and activate
+```{.bash}
+export UV_TORCH_BACKEND=cu126
+uv venv --no-project --relocatable
+source .venv/bin/activate
+```
+
+Install PyTorch
+- PyTorch 2.6.0 recommended
+```{.bash}
+uv pip install packaging setuptools wheel
+uv pip install torch==2.6.0
+uv pip install awscli pydantic
+```
+
+Install axolotl from PyPi
+```{.bash}
+uv pip install --no-build-isolation axolotl[deepspeed,flash-attn]
+
+# optionally install with vLLM if you're using torch==2.6.0 and want to train w/ GRPO
+uv pip install --no-build-isolation axolotl[deepspeed,flash-attn,vllm]
+```
+
 ### Edge/Development Build {#sec-edge-build}
 
 For the latest features between releases:
@@ -72,6 +110,10 @@ docker run --privileged --gpus '"all"' --shm-size 10g --rm -it \
 ```
 :::
 
+::: {.callout-important}
+For Blackwell GPUs, please use `axolotlai/axolotl:main-py3.11-cu128-2.7.0` or the cloud variant `axolotlai/axolotl-cloud:main-py3.11-cu128-2.7.0`.
+:::
+
 Please refer to the [Docker documentation](docker.qmd) for more information on the different Docker images that are available.
 
 ## Cloud Environments {#sec-cloud}
@@ -111,7 +153,7 @@ We recommend using WSL2 (Windows Subsystem for Linux) or Docker.
 
 ### Conda/Pip venv {#sec-conda}
 
-1. Install Python ≥3.10
+1. Install Python ≥3.11
 2. Install PyTorch: https://pytorch.org/get-started/locally/
 3. Install Axolotl:
    ```{.bash}

docs/lora_optims.qmd

@@ -84,6 +84,10 @@ lora_qkv_kernel: true
 lora_o_kernel: true
 ```
 
+::: {.callout-note}
+Currently, LoRA kernels are not supported for RLHF training, only SFT.
+:::
+
 ## Requirements
 
 - One or more NVIDIA or AMD GPUs (in order to use the Triton kernels)

docs/multi-gpu.qmd

@@ -87,20 +87,7 @@ We support sequence parallelism (SP) via the
 allows one to split up sequences across GPUs, which is useful in the event that a
 single sequence causes OOM errors during model training.
 
-First, install `ring-flash-attn`, recommended via `pip install axolotl[ring-flash-attn]`,
-or from source with `pip install .[ring-flash-attn]`.
-
-Your Axolotl YAML config should contain the following lines:
-
-```{.yaml}
-sequence_parallel_degree: 4  # Split each sequence into 4 parts, one per GPU
-flash_attention: true  # Required with sequence parallelism
-
-# Optional; strides across the key dimension. Larger values use more memory but will make training faster.
-heads_k_stride: 1
-```
-
-See our [dedicated guide](sequence_parallelism.qmd) for more details.
+See our [dedicated guide](sequence_parallelism.qmd) for more information.
 
 ### FSDP + QLoRA {#sec-fsdp-qlora}
 

docs/multimodal.qmd

@@ -43,7 +43,7 @@ datasets:
 # leave the vision model and vision tower frozen
 # load_in_8bit: true
 adapter: lora
-lora_target_modules: 'language_model.model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
+lora_target_modules: 'model.language_model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
 
 # (optional) if you want to resize images to a set size
 image_size: 512

docs/qat.qmd

@@ -0,0 +1,32 @@
+---
+title: "Quantization Aware Training (QAT)"
+back-to-top-navigation: true
+toc: true
+toc-expand: 2
+toc-depth: 4
+---
+
+## Overview
+
+[Quantization Aware Training](https://pytorch.org/blog/introduction-to-quantization-on-pytorch/#quantization-aware-training) (QAT) is a technique for improving the accuracy of models which are quantized
+by applying "fake" quantizations to the model's weights (and optionally, activations) during training. This fake
+quantization allows for the model to adjust for noise introduced by the quantization, so when the model is eventually
+quantized, the accuracy loss is minimized. We use the quantization techniques implemented in [torchao](https://github.com/pytorch/ao) to provide
+support for QAT and post-training quantization (PTQ) in axolotl.
+
+We recommend reviewing the excellent QAT tutorial in the [torchtune library](https://pytorch.org/torchtune/main/tutorials/qat_finetune.html#quantizing-the-qat-model),
+and the QAT documentation in the [torchao library](https://github.com/pytorch/ao/tree/main/torchao/quantization/qat), for more details.
+
+## Configuring QAT in Axolotl
+
+To enable QAT in axolotl, add the following to your configuration file:
+
+```yaml
+qat:
+  activation_dtype: # Optional[str] = "int8". Fake quantization layout to use for activation quantization. Valid options are "int4" and "int8"
+  weight_dtype: # Optional[str] = "int8". Fake quantization layout to use for weight quantization. Valid options are "int4" and "int8"
+  group_size: # Optional[int] = 32. The number of elements in each group for per-group fake quantization
+  fake_quant_after_n_steps: # Optional[int] = None. The number of steps to apply fake quantization after
+```
+
+Once you have finished training, you must quantize your model by using the same quantization configuration which you used to train the model with. You can use the [`quantize`](./quantize.qmd) command to do this.

docs/quantize.qmd

@@ -0,0 +1,53 @@
+---
+title: "Quantization with torchao"
+back-to-top-navigation: true
+toc: true
+toc-expand: 2
+toc-depth: 4
+---
+
+Quantization is a technique to lower the memory footprint of your model, potentially at the cost of accuracy or model performance. We support quantizing your model using the [torchao](https://github.com/pytorch/ao) library. Quantization is supported for both post-training quantization (PTQ) and quantization-aware training (QAT).
+
+
+::: {.callout-note}
+
+We do not currently support quantization techniques such as GGUF/GPTQ,EXL2 at the moment.
+
+:::
+
+## Configuring Quantization in Axolotl
+
+Quantization is configured using the `quantization` key in your configuration file.
+
+```yaml
+base_model: # The path to the model to quantize.
+quantization:
+  weight_dtype: # Optional[str] = "int8". Fake quantization layout to use for weight quantization. Valid options are uintX for X in [1, 2, 3, 4, 5, 6, 7], or int4, or int8
+  activation_dtype: # Optional[str] = "int8". Fake quantization layout to use for activation quantization. Valid options are "int4" and "int8"
+  group_size: # Optional[int] = 32. The number of elements in each group for per-group fake quantization
+  quantize_embedding: # Optional[bool] = False. Whether to quantize the embedding layer.
+
+output_dir:  # The path to the output directory.
+```
+
+Once quantization is complete, your quantized model will be saved in the `{output_dir}/quantized` directory.
+
+You may also use the `quantize` command to quantize a model which has been trained with [QAT](./qat.qmd) - you can do this by using the existing QAT configuration file which
+you used to train the model:
+
+```yaml
+# qat.yml
+qat:
+  activation_dtype: int8
+  weight_dtype: int8
+  group_size: 256
+  quantize_embedding: true
+
+output_dir: # The path to the output directory used during training where the final checkpoint has been saved.
+```
+
+```bash
+axolotl quantize qat.yml
+```
+
+This ensures that an identical quantization configuration is used to quantize the model as was used to train it.

docs/rlhf.qmd

@@ -16,7 +16,8 @@ feedback. Various methods include, but not limited to:
 - [Identity Preference Optimization (IPO)](#ipo)
 - [Kahneman-Tversky Optimization (KTO)](#kto)
 - [Odds Ratio Preference Optimization (ORPO)](#orpo)
-- Proximal Policy Optimization (PPO) (not yet supported in axolotl)
+- [Group Relative Policy Optimization (GRPO)](#grpo)
+- Proximal Policy Optimization (PPO) (not yet supported in axolotl, if you're interested in contributing, please reach out!)
 
 
 ## RLHF using Axolotl
@@ -499,7 +500,7 @@ The input format is a simple JSON input with customizable fields based on the ab
 ### GRPO
 
 ::: {.callout-tip}
-Check out our [GRPO cookbook](https://github.com/axolotl-ai-cloud/axolotl-cookbook/tree/main/grpo#training-an-r1-style-large-language-model-using-grpo).
+Check out our [GRPO cookbook](https://github.com/axolotl-ai-cloud/grpo_code).
 :::
 
 In the latest GRPO implementation, `vLLM` is used to significantly speedup trajectory generation during training. In this example, we're using 4 GPUs - 2 for training, and 2 for vLLM:
@@ -582,7 +583,20 @@ datasets:
 
 To see other examples of custom reward functions, please see [TRL GRPO Docs](https://github.com/huggingface/trl/blob/main/docs/source/grpo_trainer.md#using-a-custom-reward-function).
 
-To see description of the configs, please see [TRLConfig](https://github.com/axolotl-ai-cloud/axolotl/blob/main/src/axolotl/utils/config/models/input/v0_4_1/trl.py).
+To see all configs, please see [TRLConfig](https://github.com/axolotl-ai-cloud/axolotl/blob/v0.9.2/src/axolotl/utils/schemas/trl.py).
+
+#### GRPO with DAPO/Dr. GRPO loss
+
+The DAPO paper and subsequently Dr. GRPO paper proposed an alternative loss function for GRPO to remediate the penalty in longer responses.
+
+```yaml
+trl:
+  loss_type: dr_grpo
+  # Normalizes loss based on max completion length (default: 256)
+  max_completion_length:
+```
+
+For more information, see [GRPO docs](https://huggingface.co/docs/trl/v0.17.0/en/grpo_trainer#loss-types).
 
 ### SimPO
 

docs/scripts/generate_config_docs.py

@@ -0,0 +1,752 @@
+# type: ignore
+
+"""
+Quarto documentation generation from Pydantic models. Uses Pydantic model source code
+to automatically group fields, including inherited fields from parent classes.
+"""
+
+import ast
+import inspect
+import textwrap
+import types
+import typing
+from typing import Any, FrozenSet, Type, Union
+
+from pydantic import BaseModel
+
+from axolotl.utils.schemas.config import AxolotlInputConfig
+
+
+class QuartoGenerator:
+    """Generate Quarto documentation from Pydantic models."""
+
+    def __init__(self):
+        self._class_fields_cache = {}
+        self._inheritance_map_cache = {}
+        self._nested_models_cache = {}
+
+    def _get_direct_fields(self, cls: Type[BaseModel]) -> FrozenSet[str]:
+        """Get fields defined directly in a single class (not inherited)."""
+        if cls in self._class_fields_cache:
+            return self._class_fields_cache[cls]
+
+        fields = set()
+
+        # Get annotated fields
+        if hasattr(cls, "__annotations__"):
+            fields.update(cls.__annotations__.keys())
+
+        # Filter out private/special methods
+        fields = {f for f in fields if not f.startswith("_")}
+
+        result = frozenset(fields)
+        self._class_fields_cache[cls] = result
+        return result
+
+    def _is_pydantic_model(self, type_obj) -> bool:
+        """Check if a type is a Pydantic BaseModel."""
+        return inspect.isclass(type_obj) and issubclass(type_obj, BaseModel)
+
+    # pylint: disable=too-many-return-statements
+    def _extract_nested_type(self, field_type) -> Any:
+        """Extract the actual type from complex type annotations."""
+        # Handle Annotated types (Python 3.9+)
+        if hasattr(typing, "get_origin") and hasattr(typing, "get_args"):
+            origin = typing.get_origin(field_type)
+            args = typing.get_args(field_type)
+
+            if origin is not None:
+                # Handle Annotated[SomeType, ...] - extract the first argument
+                if hasattr(typing, "Annotated") and origin is typing.Annotated:
+                    if args:
+                        return self._extract_nested_type(
+                            args[0]
+                        )  # Recursively process the actual type
+
+                # Handle list[SomeType], List[SomeType], etc.
+                elif origin in (list, typing.List):
+                    if args:
+                        return self._extract_nested_type(
+                            args[0]
+                        )  # Extract element type
+
+                # Handle Union types (including | syntax)
+                elif origin is typing.Union:
+                    # Get non-None types from the Union
+                    non_none_types = [arg for arg in args if arg is not type(None)]
+                    if len(non_none_types) >= 1:
+                        # Prioritize Pydantic models over primitive types
+                        pydantic_models = [
+                            arg
+                            for arg in non_none_types
+                            if self._is_pydantic_model(arg)
+                        ]
+                        if pydantic_models:
+                            # Return the first Pydantic model found
+                            return self._extract_nested_type(pydantic_models[0])
+
+                        # No Pydantic models, return the first non-None type
+                        return self._extract_nested_type(non_none_types[0])
+
+        # Handle new Python 3.10+ union syntax (PeftConfig | None)
+        if hasattr(field_type, "__class__") and field_type.__class__ is types.UnionType:
+            # Get non-None types from the Union
+            non_none_types = [
+                arg for arg in field_type.__args__ if arg is not type(None)
+            ]
+            if len(non_none_types) >= 1:
+                # Prioritize Pydantic models over primitive types
+                pydantic_models = [
+                    arg for arg in non_none_types if self._is_pydantic_model(arg)
+                ]
+                if pydantic_models:
+                    return self._extract_nested_type(pydantic_models[0])
+                return self._extract_nested_type(non_none_types[0])
+
+        # Handle old typing.Union syntax (fallback)
+        if hasattr(field_type, "__origin__"):
+            if field_type.__origin__ is Union:
+                # Get non-None types from the Union
+                non_none_types = [
+                    arg for arg in field_type.__args__ if arg is not type(None)
+                ]
+                if len(non_none_types) >= 1:
+                    # Prioritize Pydantic models over primitive types
+                    pydantic_models = [
+                        arg for arg in non_none_types if self._is_pydantic_model(arg)
+                    ]
+                    if pydantic_models:
+                        return self._extract_nested_type(pydantic_models[0])
+                    return self._extract_nested_type(non_none_types[0])
+            # Handle other generic types like dict[str, Any], etc.
+            elif hasattr(field_type, "__args__"):
+                return field_type
+
+        return field_type
+
+    # pylint: disable=too-many-return-statements
+    def _extract_all_pydantic_models_from_type(
+        self, field_type
+    ) -> list[type[BaseModel]]:
+        """Extract all Pydantic models from a type annotation, including from Unions."""
+        models = []
+
+        if field_type is None:
+            return models
+
+        # Handle Annotated types
+        if hasattr(typing, "get_origin") and hasattr(typing, "get_args"):
+            origin = typing.get_origin(field_type)
+            args = typing.get_args(field_type)
+
+            if origin is not None:
+                # Handle Annotated[SomeType, ...] - extract from the first argument
+                if hasattr(typing, "Annotated") and origin is typing.Annotated:
+                    if args:
+                        models.extend(
+                            self._extract_all_pydantic_models_from_type(args[0])
+                        )
+                    return models
+
+                # Handle list[SomeType], List[SomeType], etc.
+                if origin in (list, typing.List):
+                    if args:
+                        models.extend(
+                            self._extract_all_pydantic_models_from_type(args[0])
+                        )
+                    return models
+
+                # Handle Union types
+                if origin is typing.Union:
+                    for arg in args:
+                        if arg is not type(None):  # Skip None type
+                            models.extend(
+                                self._extract_all_pydantic_models_from_type(arg)
+                            )
+                    return models
+
+        # Handle new Python 3.10+ union syntax
+        if hasattr(field_type, "__class__") and field_type.__class__ is types.UnionType:
+            for arg in field_type.__args__:
+                if arg is not type(None):  # Skip None type
+                    models.extend(self._extract_all_pydantic_models_from_type(arg))
+            return models
+
+        # Handle old typing.Union syntax (fallback)
+        if hasattr(field_type, "__origin__") and field_type.__origin__ is Union:
+            for arg in field_type.__args__:
+                if arg is not type(None):  # Skip None type
+                    models.extend(self._extract_all_pydantic_models_from_type(arg))
+            return models
+
+        # Check if this type itself is a Pydantic model
+        if self._is_pydantic_model(field_type):
+            models.append(field_type)
+
+        return models
+
+    def _get_nested_models(
+        self, model_class: type[BaseModel], visited=None
+    ) -> dict[str, type[BaseModel]]:
+        """Get all nested Pydantic models from a model class."""
+        if visited is None:
+            visited = set()
+
+        # Avoid infinite recursion
+        if model_class in visited:
+            return {}
+
+        if model_class in self._nested_models_cache:
+            return self._nested_models_cache[model_class]
+
+        visited.add(model_class)
+        nested_models = {}
+
+        # Check all fields in the model
+        for field_info in model_class.model_fields.values():
+            field_type = self._extract_nested_type(field_info.annotation)
+
+            if self._is_pydantic_model(field_type):
+                nested_models[field_type.__name__] = field_type
+                # Recursively get nested models from this nested model
+                deeper_nested = self._get_nested_models(field_type, visited.copy())
+                nested_models.update(deeper_nested)
+
+        self._nested_models_cache[model_class] = nested_models
+        return nested_models
+
+    def _build_inheritance_map(self, child_class: Type[BaseModel]):
+        """Build inheritance map for a class and all its parents."""
+        if child_class in self._inheritance_map_cache:
+            return self._inheritance_map_cache[child_class]
+
+        inheritance_map = {}
+
+        # Get MRO and filter out BaseModel and object
+        mro_classes = [
+            cls
+            for cls in child_class.__mro__
+            if cls not in (BaseModel, object) and hasattr(cls, "__annotations__")
+        ]
+
+        # Process each class in the MRO
+        for cls in mro_classes:
+            inheritance_map[cls] = self._get_direct_fields(cls)
+
+        self._inheritance_map_cache[child_class] = inheritance_map
+        return inheritance_map
+
+    def _wrap_comment(self, text: str, width: int = 88) -> list[str]:
+        """Wrap a comment to specified width, accounting for '# ' prefix."""
+        if not text.strip():
+            return ["#"]
+
+        # Account for "# " prefix (2 characters)
+        content_width = width - 2
+        wrapped_lines = textwrap.wrap(text, width=content_width)
+        return [f"# {line}" for line in wrapped_lines]
+
+    def _extract_type_from_source(
+        self, model_class: type[BaseModel], field_name: str
+    ) -> str:
+        """Extract the actual type annotation text from source code, checking inheritance chain."""
+        # Use inheritance map to check classes efficiently
+        inheritance_map = self._build_inheritance_map(model_class)
+
+        # Check classes in MRO order
+        for cls in model_class.__mro__:
+            if cls in inheritance_map and field_name in inheritance_map[cls]:
+                type_annotation = self._get_type_from_class_source(cls, field_name)
+                if type_annotation != "unknown":
+                    return type_annotation
+
+        return "unknown"
+
+    def _get_type_from_class_source(self, class_obj: type, field_name: str) -> str:
+        """Extract type annotation from a specific class's source code."""
+        try:
+            source = inspect.getsource(class_obj)
+            tree = ast.parse(source)
+        except (OSError, TypeError):
+            return "unknown"
+
+        # Find the class definition
+        for node in tree.body:
+            if isinstance(node, ast.ClassDef) and node.name == class_obj.__name__:
+                # Find the field assignment
+                for body_node in node.body:
+                    if isinstance(body_node, ast.AnnAssign) and isinstance(
+                        body_node.target, ast.Name
+                    ):
+                        if body_node.target.id == field_name and body_node.annotation:
+                            return ast.unparse(body_node.annotation)
+                break
+
+        return "unknown"
+
+    def _extract_field_groups_from_all_classes(
+        self, model_class: type[BaseModel]
+    ) -> list[dict]:
+        """Extract field groups from all classes in the inheritance hierarchy."""
+        all_groups = []
+        inheritance_map = self._build_inheritance_map(model_class)
+
+        # Get all Pydantic base classes in MRO order (most specific first)
+        # This puts AxolotlInputConfig fields first, then parent class fields
+        pydantic_classes = [
+            cls
+            for cls in model_class.__mro__
+            if cls in inheritance_map and inheritance_map[cls]
+        ]
+
+        # Extract groups from each class
+        for cls in pydantic_classes:
+            class_groups = self._extract_field_groups_from_source(cls)
+            for group in class_groups:
+                all_groups.append(group)
+
+        # If no groups found, create a default grouping by class
+        if not all_groups:
+            for cls in pydantic_classes:
+                fields_in_class = inheritance_map[cls]
+                if fields_in_class:
+                    all_groups.append(
+                        {
+                            "fields": list(fields_in_class),
+                        }
+                    )
+
+        return all_groups
+
+    # pylint: disable=too-many-return-statements
+    def _extract_field_groups_from_source(
+        self, model_class: type[BaseModel]
+    ) -> list[dict]:
+        """Extract field groups from source code based on blank lines and comments."""
+        try:
+            source = inspect.getsource(model_class)
+            tree = ast.parse(source)
+        except (OSError, TypeError):
+            # Fallback if we can't get source code
+            fields_in_class = self._get_direct_fields(model_class)
+            if fields_in_class:
+                return [
+                    {
+                        "fields": list(fields_in_class),
+                    }
+                ]
+            return []
+
+        groups = []
+        current_group_fields = []
+        current_group_comment = None
+
+        # Find the class definition
+        class_node = None
+        for node in ast.walk(tree):
+            if isinstance(node, ast.ClassDef) and node.name == model_class.__name__:
+                class_node = node
+                break
+
+        if not class_node:
+            fields_in_class = self._get_direct_fields(model_class)
+            if fields_in_class:
+                return [
+                    {
+                        "fields": list(fields_in_class),
+                    }
+                ]
+            return []
+
+        # Parse the source lines to detect groupings
+        source_lines = source.split("\n")
+
+        # Get fields that are actually defined in this specific class
+        fields_in_class = self._get_direct_fields(model_class)
+
+        # Find assignments that correspond to model fields for THIS class only
+        field_assignments = []
+        for node in class_node.body:
+            if isinstance(node, ast.AnnAssign) and isinstance(node.target, ast.Name):
+                field_name = node.target.id
+                if field_name in fields_in_class:
+                    field_assignments.append(
+                        {
+                            "name": field_name,
+                            "lineno": node.lineno,
+                            "end_lineno": getattr(node, "end_lineno", node.lineno),
+                        }
+                    )
+
+        if not field_assignments:
+            if fields_in_class:
+                return [
+                    {
+                        "fields": list(fields_in_class),
+                    }
+                ]
+            return []
+
+        # Sort by line number
+        field_assignments.sort(key=lambda x: x["lineno"])
+
+        # Group fields based on blank lines and comments
+        for i, field_info in enumerate(field_assignments):
+            field_name = field_info["name"]
+            current_line = field_info["lineno"]
+
+            # Check if this starts a new group (blank line before or significant gap)
+            is_new_group = False
+
+            if i == 0:
+                is_new_group = True
+            else:
+                prev_end_line = field_assignments[i - 1]["end_lineno"]
+
+                # Check for blank lines or comments between fields
+                lines_between = source_lines[prev_end_line : current_line - 1]
+                has_blank_line = any(line.strip() == "" for line in lines_between)
+                has_comment = any(
+                    line.strip().startswith("#") for line in lines_between
+                )
+
+                # Start new group if there's a blank line or comment, or significant gap
+                if has_blank_line or has_comment or (current_line - prev_end_line > 3):
+                    is_new_group = True
+
+            if is_new_group and current_group_fields:
+                # Save the previous group
+                groups.append(
+                    {
+                        "fields": current_group_fields.copy(),
+                        "description": current_group_comment,
+                    }
+                )
+                current_group_fields = []
+                current_group_comment = None
+
+            current_group_fields.append(field_name)
+
+        # Add the final group
+        if current_group_fields:
+            groups.append(
+                {
+                    "fields": current_group_fields,
+                    "description": current_group_comment,
+                }
+            )
+
+        return groups
+
+    def _generate_field_documentation(
+        self,
+        model_class: type[BaseModel],
+        field_name: str,
+        field_info: dict,
+        field_type_str: str,
+        is_required: bool,
+        indent_level: int = 0,
+        visited_models: set = None,
+    ) -> list[str]:
+        """Generate documentation for a single field, expanding nested models inline."""
+        if visited_models is None:
+            visited_models = set()
+
+        lines = []
+        indent = "  " * indent_level
+
+        # Get the actual field type for nested model detection
+        if field_name in model_class.model_fields:
+            pydantic_field_info = model_class.model_fields[field_name]
+            actual_field_type = pydantic_field_info.annotation
+        else:
+            actual_field_type = None
+
+        # Add description comment if available
+        description = field_info.get("description", "")
+        if description:
+            wrapped_lines = self._wrap_comment(description, width=88 - len(indent))
+            for line in wrapped_lines:
+                lines.append(f"{indent}{line}")
+
+        # Extract nested Pydantic models from the type annotation
+        nested_models = self._extract_all_pydantic_models_from_type(actual_field_type)
+
+        # Filter out already visited models to prevent infinite recursion
+        expandable_models = [
+            model for model in nested_models if model not in visited_models
+        ]
+
+        if expandable_models:
+            # This field contains Pydantic models that can be expanded
+
+            # Show the field with its full type annotation
+            field_line = f"{indent}{field_name}: {field_type_str}"
+            if field_info.get("default") is not None:
+                field_line += f" = {field_info['default']}"
+            if is_required:
+                field_line += " (required)"
+            lines.append(field_line)
+
+            # Add to visited to prevent infinite recursion
+            new_visited = visited_models.copy()
+            new_visited.update(expandable_models)
+
+            # Expand each nested Pydantic model
+            for i, nested_model in enumerate(expandable_models):
+                if i > 0:
+                    lines.append("\n")
+                lines.append(f"{indent}  # For {nested_model.__name__}:")
+
+                # Get nested model schema
+                try:
+                    nested_schema = nested_model.model_json_schema()
+                    nested_properties = nested_schema.get("properties", {})
+                    nested_required = nested_schema.get("required", [])
+                except Exception:  # pylint: disable=broad-exception-caught
+                    # Fallback: use model fields directly
+                    nested_properties = {}
+                    nested_required = []
+                    for (
+                        nested_field_name,
+                        nested_field_info,
+                    ) in nested_model.model_fields.items():
+                        nested_description = ""
+                        if (
+                            hasattr(nested_field_info, "json_schema_extra")
+                            and nested_field_info.json_schema_extra
+                        ):
+                            nested_description = (
+                                nested_field_info.json_schema_extra.get(
+                                    "description", ""
+                                )
+                            )
+                        elif (
+                            hasattr(nested_field_info, "description")
+                            and nested_field_info.description
+                        ):
+                            nested_description = nested_field_info.description
+
+                        nested_default_val = None
+                        if (
+                            hasattr(nested_field_info, "default")
+                            and nested_field_info.default is not None
+                        ):
+                            if str(nested_field_info.default) != "PydanticUndefined":
+                                nested_default_val = nested_field_info.default
+
+                        nested_properties[nested_field_name] = {
+                            "type": "unknown",
+                            "description": nested_description,
+                            "default": nested_default_val,
+                        }
+
+                        if nested_field_info.is_required():
+                            nested_required.append(nested_field_name)
+
+                # Get field groups for the nested model
+                nested_field_groups = self._extract_field_groups_from_all_classes(
+                    nested_model
+                )
+
+                # Generate nested fields with increased indentation
+                for i, group in enumerate(nested_field_groups):
+                    if not group["fields"]:
+                        continue
+
+                    # Add blank line between groups (except before first group)
+                    if i > 0:
+                        lines.append("")
+
+                    # Process nested fields
+                    for nested_field_name in group["fields"]:
+                        if nested_field_name not in nested_properties:
+                            continue
+
+                        nested_field_info = nested_properties[nested_field_name]
+                        nested_field_type = self._extract_type_from_source(
+                            nested_model, nested_field_name
+                        )
+                        nested_is_required = nested_field_name in nested_required
+
+                        # Recursively generate documentation for nested field
+                        nested_lines = self._generate_field_documentation(
+                            nested_model,
+                            nested_field_name,
+                            nested_field_info,
+                            nested_field_type,
+                            nested_is_required,
+                            indent_level + 1,
+                            new_visited,
+                        )
+                        lines.extend(nested_lines)
+        else:
+            # Regular field (no expandable nested models)
+            field_line = f"{indent}{field_name}: {field_type_str}"
+            if field_info.get("default") is not None:
+                field_line += f" = {field_info['default']}"
+            if is_required:
+                field_line += " (required)"
+            lines.append(field_line)
+
+        return lines
+
+    def generate_qmd(
+        self,
+        model_class: type[BaseModel],
+        title: str | None = None,
+        expand_nested: bool = True,
+    ) -> str:
+        """Auto-generate config reference documentation including inherited fields."""
+
+        if title is None:
+            title = f"{model_class.__name__} Reference"
+
+        # Try to get JSON schema, with fallback for serialization issues
+        try:
+            schema = model_class.model_json_schema()
+            properties = schema.get("properties", {})
+            required = schema.get("required", [])
+        except Exception as e:  # pylint: disable=broad-exception-caught
+            print(
+                f"Warning: Could not generate JSON schema ({e}). Using model fields instead."
+            )
+            # Fallback: use model fields directly
+            properties = {}
+            required = []
+            for field_name, field_info in model_class.model_fields.items():
+                # Extract description from json_schema_extra or field info
+                description = ""
+                if (
+                    hasattr(field_info, "json_schema_extra")
+                    and field_info.json_schema_extra
+                ):
+                    description = field_info.json_schema_extra.get("description", "")
+                elif hasattr(field_info, "description") and field_info.description:
+                    description = field_info.description
+
+                # Get default value
+                default_val = None
+                if hasattr(field_info, "default") and field_info.default is not None:
+                    # Handle special Pydantic default markers
+                    if str(field_info.default) != "PydanticUndefined":
+                        default_val = field_info.default
+
+                properties[field_name] = {
+                    "type": "unknown",
+                    "description": description,
+                    "default": default_val,
+                }
+
+                if field_info.is_required():
+                    required.append(field_name)
+
+        # Extract field groups from all classes in inheritance hierarchy
+        field_groups = self._extract_field_groups_from_all_classes(model_class)
+
+        # Start building QMD content
+        qmd_lines = [
+            "---",
+            f"title: {title}",
+            "description: A complete list of all configuration options.",
+            "---",
+            "",
+        ]
+
+        # Generate one big code block with all fields (inline nested expansion)
+        qmd_lines.append("```yaml")
+
+        for i, group in enumerate(field_groups):
+            if not group["fields"]:
+                continue
+
+            # Add blank line between groups (except before first group)
+            if i > 0:
+                qmd_lines.append("")
+
+            # Process fields in the order they appear in source
+            for field_name in group["fields"]:
+                if field_name not in properties:
+                    continue
+
+                field_info = properties[field_name]
+                field_type = self._extract_type_from_source(model_class, field_name)
+                is_required = field_name in required
+
+                if expand_nested:
+                    # Check if this field has nested models
+                    if field_name in model_class.model_fields:
+                        pydantic_field_info = model_class.model_fields[field_name]
+                        nested_models = self._extract_all_pydantic_models_from_type(
+                            pydantic_field_info.annotation
+                        )
+                        has_nested = bool(nested_models)
+                    else:
+                        has_nested = False
+
+                    # Add blank line before nested config
+                    if has_nested:
+                        qmd_lines.append("")
+
+                    # Use the new inline generation method
+                    field_lines = self._generate_field_documentation(
+                        model_class,
+                        field_name,
+                        field_info,
+                        field_type,
+                        is_required,
+                        indent_level=0,
+                        visited_models=set(),
+                    )
+                    qmd_lines.extend(field_lines)
+
+                    # Add blank line after nested config
+                    if has_nested:
+                        qmd_lines.append("")
+                else:
+                    # Original simple approach
+                    description = field_info.get("description", "")
+                    default = field_info.get("default")
+
+                    # Add wrapped comment for description
+                    if description:
+                        wrapped_lines = self._wrap_comment(description)
+                        qmd_lines.extend(wrapped_lines)
+
+                    line = f"{field_name}: {field_type}"
+                    if default is not None:
+                        line += f" = {default}"
+                    if is_required:
+                        line += " (required)"
+                    qmd_lines.append(line)
+
+        qmd_lines.append("```")
+
+        # Join all lines and clean up any double newlines
+        content = "\n".join(qmd_lines)
+
+        # Replace multiple consecutive newlines with just two newlines (one blank line)
+        import re
+
+        content = re.sub(r"\n{3,}", "\n\n", content)
+
+        # Ensure single newline at the very end
+        content = content.rstrip("\n") + "\n"
+
+        return content
+
+
+def main():
+    generator = QuartoGenerator()
+
+    print("Generating config reference content...")
+    qmd_content = generator.generate_qmd(AxolotlInputConfig, "Config Reference", True)
+
+    print("Writing to file...")
+    with open("docs/config-reference.qmd", "w", encoding="utf-8") as f:
+        f.write(qmd_content)
+    print("Done!")
+
+
+if __name__ == "__main__":
+    main()

docs/sequence_parallelism.qmd

@@ -3,8 +3,6 @@ title: Sequence Parallelism
 description: Train with long sequences split across multiple GPUs.
 ---
 
-# Sequence Parallelism
-
 Sequence parallelism is a technique that splits sequences across multiple GPUs,
 allowing you to train with very long sequences that wouldn't fit on a single GPU. Each
 GPU processes a different portion of the sequence, and the results are aggregated
@@ -27,7 +25,7 @@ To enable sequence parallelism, add the following to your configuration file:
 sequence_parallel_degree: 4  # Split sequences across 4 GPUs
 # Optional; strides across the key dimension. Larger values use more memory but should make training faster.
 heads_k_stride: 1
-# Optional; one of "varlen_llama3", "batch_ring", "batch_zigzag", "batch_stripe". Defaults to
+# Optional; one of "varlen_llama3" or "batch_ring". Defaults to
 # "varlen_llama3" when `sample_packing: true`, and "batch_ring" otherwise.
 ring_attn_func:
 ```
@@ -43,7 +41,7 @@ When sequence parallelism is enabled:
 
 1. Each sequence is divided into equal chunks across the GPUs in a sequence parallel group
 2. The data collator handles the chunking of input_ids, attention_mask, labels, and position_ids
-3. Position IDs are adjusted to maintain proper relative positions, especially for packed sequences
+3. Position IDs are adjusted to maintain proper relative positions
 4. The trainer uses special ring communication patterns for attention operations
 
 ## Requirements
@@ -69,9 +67,11 @@ sequence_len: 8192
 ...
 
 sequence_parallel_degree: 4  # Split each sequence into 4 parts, one per GPU
-flash_attention: true  # Required with sequence parallelism
 # Optional; strides across the key dimension. Larger values use more memory but should make training faster.
 heads_k_stride: 1
+# Optional; one of "varlen_llama3" or "batch_ring". Defaults to
+# "varlen_llama3" when `sample_packing: true`, and "batch_ring" otherwise.
+ring_attn_func:
 
 ...
 ```

examples/falcon-h1/falcon-h1-1b-deep-qlora.yaml

@@ -0,0 +1,71 @@
+base_model: tiiuae/Falcon-H1-1.5B-Deep-Base
+# optionally might have model_type or tokenizer_type
+model_type: AutoModelForCausalLM
+tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: true
+
+# huggingface repo
+chat_template: falcon_h1
+datasets:
+  - path: cgato/SlimOrcaDedupCleaned
+    type: chat_template
+    field_messages: conversations
+    message_property_mappings:
+      role: from
+      content: value
+
+val_set_size: 0.0
+output_dir: ./outputs/out
+
+adapter: qlora
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_modules:
+  - q_proj
+  - k_proj
+  - v_proj
+  - o_proj
+  - in_proj
+  - gate_proj
+  - up_proj
+  - down_proj
+
+sequence_len: 2048
+sample_packing: false
+eval_sample_packing: false
+pad_to_sequence_len: true
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 4
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+bf16: auto
+tf32: true
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+resume_from_checkpoint:
+logging_steps: 1
+flash_attention: true
+
+warmup_ratio: 0.1
+evals_per_epoch:
+saves_per_epoch: 1
+weight_decay: 0.0
+special_tokens:

examples/falcon-h1/falcon-h1-1b-qlora.yaml

@@ -0,0 +1,71 @@
+base_model: tiiuae/Falcon-H1-1.5B-Base
+# optionally might have model_type or tokenizer_type
+model_type: AutoModelForCausalLM
+tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: true
+
+# huggingface repo
+chat_template: falcon_h1
+datasets:
+  - path: cgato/SlimOrcaDedupCleaned
+    type: chat_template
+    field_messages: conversations
+    message_property_mappings:
+      role: from
+      content: value
+
+val_set_size: 0.0
+output_dir: ./outputs/out
+
+adapter: qlora
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_modules:
+  - q_proj
+  - k_proj
+  - v_proj
+  - o_proj
+  - in_proj
+  - gate_proj
+  - up_proj
+  - down_proj
+
+sequence_len: 2048
+sample_packing: false
+eval_sample_packing: false
+pad_to_sequence_len: true
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 4
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+bf16: auto
+tf32: true
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+resume_from_checkpoint:
+logging_steps: 1
+flash_attention: true
+
+warmup_ratio: 0.1
+evals_per_epoch:
+saves_per_epoch: 1
+weight_decay: 0.0
+special_tokens:

examples/falcon-h1/falcon-h1-34b-qlora.yaml

@@ -0,0 +1,71 @@
+base_model: tiiuae/Falcon-H1-34B-Base
+# optionally might have model_type or tokenizer_type
+model_type: AutoModelForCausalLM
+tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: true
+
+# huggingface repo
+chat_template: falcon_h1
+datasets:
+  - path: cgato/SlimOrcaDedupCleaned
+    type: chat_template
+    field_messages: conversations
+    message_property_mappings:
+      role: from
+      content: value
+
+val_set_size: 0.0
+output_dir: ./outputs/out
+
+adapter: qlora
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_modules:
+  - q_proj
+  - k_proj
+  - v_proj
+  - o_proj
+  - in_proj
+  - gate_proj
+  - up_proj
+  - down_proj
+
+sequence_len: 2048
+sample_packing: false
+eval_sample_packing: false
+pad_to_sequence_len: true
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 4
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+bf16: auto
+tf32: true
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+resume_from_checkpoint:
+logging_steps: 1
+flash_attention: true
+
+warmup_ratio: 0.1
+evals_per_epoch:
+saves_per_epoch: 1
+weight_decay: 0.0
+special_tokens:

examples/falcon-h1/falcon-h1-3b-qlora.yaml

@@ -0,0 +1,71 @@
+base_model: tiiuae/Falcon-H1-3B-Base
+# optionally might have model_type or tokenizer_type
+model_type: AutoModelForCausalLM
+tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: true
+
+# huggingface repo
+chat_template: falcon_h1
+datasets:
+  - path: cgato/SlimOrcaDedupCleaned
+    type: chat_template
+    field_messages: conversations
+    message_property_mappings:
+      role: from
+      content: value
+
+val_set_size: 0.0
+output_dir: ./outputs/out
+
+adapter: qlora
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_modules:
+  - q_proj
+  - k_proj
+  - v_proj
+  - o_proj
+  - in_proj
+  - gate_proj
+  - up_proj
+  - down_proj
+
+sequence_len: 2048
+sample_packing: false
+eval_sample_packing: false
+pad_to_sequence_len: true
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 4
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+bf16: auto
+tf32: true
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+resume_from_checkpoint:
+logging_steps: 1
+flash_attention: true
+
+warmup_ratio: 0.1
+evals_per_epoch: 1
+saves_per_epoch: 1
+weight_decay: 0.0
+special_tokens:

examples/falcon-h1/falcon-h1-500m-qlora.yaml

@@ -0,0 +1,71 @@
+base_model: tiiuae/Falcon-H1-0.5B-Instruct
+# optionally might have model_type or tokenizer_type
+model_type: AutoModelForCausalLM
+tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: true
+
+# huggingface repo
+chat_template: falcon_h1
+datasets:
+  - path: cgato/SlimOrcaDedupCleaned
+    type: chat_template
+    field_messages: conversations
+    message_property_mappings:
+      role: from
+      content: value
+
+val_set_size: 0.0
+output_dir: ./outputs/out
+
+adapter: qlora
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_modules:
+  - q_proj
+  - k_proj
+  - v_proj
+  - o_proj
+  - in_proj
+  - gate_proj
+  - up_proj
+  - down_proj
+
+sequence_len: 2048
+sample_packing: false
+eval_sample_packing: false
+pad_to_sequence_len: true
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 4
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+bf16: auto
+tf32: true
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+resume_from_checkpoint:
+logging_steps: 1
+flash_attention: true
+
+warmup_ratio: 0.1
+evals_per_epoch:
+saves_per_epoch: 1
+weight_decay: 0.0
+special_tokens:

examples/falcon-h1/falcon-h1-7b-qlora.yaml

@@ -0,0 +1,71 @@
+base_model: tiiuae/Falcon-H1-7B-Base
+# optionally might have model_type or tokenizer_type
+model_type: AutoModelForCausalLM
+tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: true
+
+# huggingface repo
+chat_template: falcon_h1
+datasets:
+  - path: cgato/SlimOrcaDedupCleaned
+    type: chat_template
+    field_messages: conversations
+    message_property_mappings:
+      role: from
+      content: value
+
+val_set_size: 0.0
+output_dir: ./outputs/out
+
+adapter: qlora
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_modules:
+  - q_proj
+  - k_proj
+  - v_proj
+  - o_proj
+  - in_proj
+  - gate_proj
+  - up_proj
+  - down_proj
+
+sequence_len: 2048
+sample_packing: false
+eval_sample_packing: false
+pad_to_sequence_len: true
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 4
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+bf16: auto
+tf32: true
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+resume_from_checkpoint:
+logging_steps: 1
+flash_attention: true
+
+warmup_ratio: 0.1
+evals_per_epoch: 1
+saves_per_epoch: 1
+weight_decay: 0.0
+special_tokens:

examples/gemma3/gemma-3-1b-qlora.yml

@@ -13,6 +13,8 @@ load_in_4bit: true
 
 # huggingface repo
 chat_template: gemma3
+eot_tokens:
+  - <end_of_turn>
 datasets:
   - path: cgato/SlimOrcaDedupCleaned
     type: chat_template

examples/gemma3/gemma-3-4b-qlora.yml

@@ -6,6 +6,8 @@ load_in_4bit: true
 ddp_find_unused_parameters: true
 
 chat_template: gemma3
+eot_tokens:
+  - <end_of_turn>
 datasets:
   - path: cgato/SlimOrcaDedupCleaned
     type: chat_template
@@ -28,7 +30,7 @@ pad_to_sequence_len: true
 lora_r: 32
 lora_alpha: 16
 lora_dropout: 0.05
-lora_target_modules: 'language_model.model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
+lora_target_modules: 'model.language_model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
 
 wandb_project:
 wandb_entity:

examples/gemma3/gemma-3-4b-vision-qlora.yml

@@ -12,6 +12,8 @@ sample_packing: false
 ddp_find_unused_parameters: true
 
 chat_template: gemma3
+eot_tokens:
+  - <end_of_turn>
 datasets:
   - path: HuggingFaceH4/llava-instruct-mix-vsft
     type: chat_template
@@ -30,7 +32,7 @@ pad_to_sequence_len: false
 lora_r: 32
 lora_alpha: 16
 lora_dropout: 0.05
-lora_target_modules: 'language_model.model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
+lora_target_modules: 'model.language_model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
 
 wandb_project:
 wandb_entity:

examples/llama-3-vision/lora-11b.yaml

@@ -29,7 +29,7 @@ pad_to_sequence_len: false
 lora_r: 32
 lora_alpha: 16
 lora_dropout: 0.05
-lora_target_modules: 'language_model.model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
+lora_target_modules: 'model.language_model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
 
 wandb_project:
 wandb_entity:

examples/llama-3/3b-qat-fsdp2.yaml

@@ -0,0 +1,79 @@
+base_model: meta-llama/Llama-3.2-3B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+plugins:
+  - axolotl.integrations.liger.LigerPlugin
+
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+liger_layer_norm: true
+liger_fused_linear_cross_entropy: true
+
+datasets:
+  - path: yahma/alpaca-cleaned
+    type: alpaca
+
+output_dir: ./outputs/qat_out/
+
+sample_packing: true
+pad_to_sequence_len: true
+sequence_len: 512
+
+flex_attention: true
+flex_attn_compile_kwargs:
+  dynamic: false
+  mode: max-autotune-no-cudagraphs
+
+qat:
+  activation_dtype: int8
+  weight_dtype: int4
+  group_size: 32
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 1
+micro_batch_size: 16
+num_epochs: 1
+optimizer: adamw_torch_fused
+
+cosine_constant_lr_ratio: 0
+cosine_min_lr_ratio: 1.0
+learning_rate: 2e-5
+save_only_model: true
+bf16: true
+
+resume_from_checkpoint:
+logging_steps: 1
+
+evals_per_epoch: 1
+saves_per_epoch: 1
+
+warmup_steps: 10
+weight_decay: 0.0
+fsdp:
+  - full_shard
+  - auto_wrap
+
+fsdp_config:
+  fsdp_version: 2
+  fsdp_offload_params: false
+  fsdp_cpu_ram_efficient_loading: true
+  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
+  fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer
+  fsdp_state_dict_type: FULL_STATE_DICT
+  fsdp_sharding_strategy: FULL_SHARD
+  fsdp_reshard_after_forward: true
+  fsdp_activation_checkpointing: true
+
+special_tokens:
+  pad_token: <|end_of_text|>

examples/llama-3/instruct-dpo-lora-8b.yml

@@ -5,6 +5,10 @@ tokenizer_type: AutoTokenizer
 # Automatically upload checkpoint and final model to HF
 # hub_model_id: username/custom_model_name
 
+special_tokens:
+  pad_token: <|finetune_right_pad_id|>
+  eos_token: <|eot_id|>
+
 load_in_8bit: true
 load_in_4bit: false
 

examples/llama-3/lora-1b.yml

@@ -5,7 +5,7 @@ base_model: NousResearch/Llama-3.2-1B
 datasets:
   - path: teknium/GPT4-LLM-Cleaned
     type: alpaca
-dataset_prepared_path: last_run_prepared
+
 val_set_size: 0.1
 output_dir: ./outputs/lora-out
 
@@ -38,6 +38,7 @@ wandb_log_model:
 gradient_accumulation_steps: 2
 micro_batch_size: 2
 num_epochs: 1
+
 optimizer: adamw_8bit
 lr_scheduler: cosine
 learning_rate: 0.0002

examples/llava/lora-7b.yaml

@@ -25,7 +25,7 @@ pad_to_sequence_len: false
 lora_r: 32
 lora_alpha: 16
 lora_dropout: 0.05
-lora_target_modules: 'language_model.model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
+lora_target_modules: 'model.language_model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
 
 wandb_project:
 wandb_entity:

examples/magistral/README.md

@@ -0,0 +1,71 @@
+# Finetune Magistral Small with Axolotl
+
+Magistral Small is a 24B parameter opensource model from MistralAI found on [HuggingFace](https://huggingface.co/mistralai/Magistral-Small-2506). This guide shows how to fine-tune it with Axolotl with multi-turn conversations with proper masking.
+
+MistralAI has also released a proprietary medium-sized version called Magistral Medium.
+
+Thanks to the team at MistralAI for giving us early access to prepare for this release.
+
+## Getting started
+
+1. Install Axolotl following the [installation guide](https://docs.axolotl.ai/docs/installation.html). You need to install from main as Magistral is only on nightly or use our latest [Docker images](https://docs.axolotl.ai/docs/docker.html).
+
+    Here is an example of how to install from main for pip:
+
+```bash
+# Ensure you have Pytorch installed (Pytorch 2.6.0 recommended)
+git clone https://github.com/axolotl-ai-cloud/axolotl.git
+cd axolotl
+
+pip3 install packaging==23.2 setuptools==75.8.0 wheel ninja
+pip3 install --no-build-isolation -e '.[flash-attn,mistral]'
+```
+
+2. Download the example config:
+
+```bash
+axolotl fetch examples
+```
+
+3. Run the finetuning example:
+
+```bash
+axolotl train examples/magistral/magistral-small-qlora.yaml
+```
+
+This config uses about 24GB VRAM.
+
+Let us know how it goes. Happy finetuning! 🚀
+
+### TIPS
+
+- For inference, the official MistralAI team recommends `top_p: 0.95` and `temperature: 0.7` with `max_tokens: 40960`.
+- You can run a full finetuning by removing the `adapter: qlora` and `load_in_4bit: true` from the config.
+- Read more on how to load your own dataset at [docs](https://docs.axolotl.ai/docs/dataset_loading.html).
+- The dataset format is the OpenAI Messages format as seen [here](https://docs.axolotl.ai/docs/dataset-formats/conversation.html#chat_template).
+
+## Optimization Guides
+
+- [Multi-GPU Training](https://docs.axolotl.ai/docs/multi-gpu.html)
+- [Multi-Node Training](https://docs.axolotl.ai/docs/multi-node.html)
+- [LoRA Optimizations](https://docs.axolotl.ai/docs/lora_optims.html)
+
+## Limitations
+
+We only support the `mistral-common` tokenizer for Supervised Fine-tuning at the moment and for `type: chat_template` only.
+
+The tokenizer does not work with `dataset.map` with multiprocessing, so we had to disable it. In addition, we do not support overriding tokens yet.
+
+## Related Resources
+
+- [MistralAI Magistral Blog](https://mistral.ai/news/magistral/)
+- [Axolotl Docs](https://docs.axolotl.ai)
+- [Axolotl Website](https://axolotl.ai)
+- [Axolotl GitHub](https://github.com/axolotl-ai-cloud/axolotl)
+- [Axolotl Discord](https://discord.gg/7m9sfhzaf3)
+
+
+## Future Work
+
+- Add parity to Preference Tuning, RL, Multi-modal, etc.
+- Add parity to other tokenizer configs like overriding tokens.

examples/magistral/magistral-small-fsdp-qlora.yaml

@@ -0,0 +1,72 @@
+base_model: mistralai/Magistral-Small-2506
+
+# Enable to use mistral-common tokenizer
+tokenizer_use_mistral_common: true
+
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: true
+
+datasets:
+  - path: fozziethebeat/alpaca_messages_2k_test
+    type: chat_template
+
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.1
+output_dir: ./outputs/lora-out
+
+adapter: qlora
+lora_model_dir:
+
+sequence_len: 2048
+sample_packing: true
+eval_sample_packing: false
+pad_to_sequence_len: true
+
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_target_modules:
+  - gate_proj
+  - down_proj
+  - up_proj
+  - q_proj
+  - v_proj
+  - k_proj
+  - o_proj
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 2
+num_epochs: 1
+optimizer: adamw_torch_fused
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+bf16: auto
+tf32: false
+
+gradient_checkpointing:
+resume_from_checkpoint:
+logging_steps: 1
+flash_attention: true
+
+warmup_ratio: 0.1
+evals_per_epoch: 1
+saves_per_epoch: 1
+
+fsdp:
+  - full_shard
+  - auto_wrap
+fsdp_config:
+  fsdp_state_dict_type: FULL_STATE_DICT
+  fsdp_transformer_layer_cls_to_wrap: MistralDecoderLayer
+  fsdp_activation_checkpointing: true

examples/magistral/magistral-small-qlora.yaml

@@ -0,0 +1,63 @@
+base_model: mistralai/Magistral-Small-2506
+
+# Enable to use mistral-common tokenizer
+tokenizer_use_mistral_common: true
+
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: true
+
+datasets:
+  - path: fozziethebeat/alpaca_messages_2k_test
+    type: chat_template
+
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.1
+output_dir: ./outputs/lora-out
+
+adapter: qlora
+lora_model_dir:
+
+sequence_len: 2048
+sample_packing: true
+pad_to_sequence_len: true
+
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_target_modules:
+  - gate_proj
+  - down_proj
+  - up_proj
+  - q_proj
+  - v_proj
+  - k_proj
+  - o_proj
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 2
+num_epochs: 1
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+bf16: auto
+tf32: false
+
+gradient_checkpointing: true
+resume_from_checkpoint:
+logging_steps: 1
+flash_attention: true
+
+warmup_ratio: 0.1
+evals_per_epoch: 1
+saves_per_epoch: 1

examples/mistral/mistral-small-3.1-24B-lora.yml

@@ -27,7 +27,7 @@ pad_to_sequence_len: false
 lora_r: 32
 lora_alpha: 16
 lora_dropout: 0.05
-lora_target_modules: 'language_model.model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
+lora_target_modules: 'model.language_model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
 
 wandb_project:
 wandb_entity:

examples/pixtral/lora-12b.yml

@@ -25,7 +25,7 @@ pad_to_sequence_len: false
 lora_r: 32
 lora_alpha: 16
 lora_dropout: 0.05
-lora_target_modules: 'language_model.model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
+lora_target_modules: 'model.language_model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
 
 wandb_project:
 wandb_entity:

examples/qwen2-vl/lora-7b.yaml

@@ -25,7 +25,7 @@ pad_to_sequence_len: false
 lora_r: 32
 lora_alpha: 16
 lora_dropout: 0.05
-lora_target_modules: 'model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
+lora_target_modules: 'model.language_model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
 
 wandb_project:
 wandb_entity:

examples/qwen2/dpo.yaml

@@ -2,7 +2,6 @@ base_model: Qwen/Qwen2.5-0.5B
 # Automatically upload checkpoint and final model to HF
 # hub_model_id: username/custom_model_name
 
-
 chat_template: qwen_25
 rl: dpo
 datasets:

examples/qwen2_5-vl/lora-7b.yaml

@@ -0,0 +1,55 @@
+base_model: Qwen/Qwen2.5-VL-7B-Instruct
+processor_type: AutoProcessor
+
+# these 3 lines are needed for now to handle vision chat templates w images
+skip_prepare_dataset: true
+remove_unused_columns: false
+sample_packing: false
+
+chat_template: qwen2_vl
+datasets:
+  - path: HuggingFaceH4/llava-instruct-mix-vsft
+    type: chat_template
+    split: train[:1%]
+    field_messages: messages
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.0
+output_dir: ./outputs/out
+
+adapter: lora
+lora_model_dir:
+
+sequence_len: 8192
+pad_to_sequence_len: false
+
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_modules: 'model.language_model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 1
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+bf16: true
+fp16:
+tf32: true
+
+gradient_checkpointing: true
+logging_steps: 1
+flash_attention: true
+eager_attention:
+
+warmup_ratio: 0.1
+evals_per_epoch: 1
+saves_per_epoch: 1
+weight_decay: 0.0

examples/qwen3/8b-qat-fsdp2.yml

@@ -0,0 +1,78 @@
+base_model: Qwen/Qwen3-8B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+plugins:
+  - axolotl.integrations.liger.LigerPlugin
+
+liger_rope: true
+liger_rms_norm: true
+liger_glu_activation: true
+liger_layer_norm: true
+liger_fused_linear_cross_entropy: true
+
+datasets:
+  - path: tatsu-lab/alpaca
+    type: alpaca
+
+output_dir: ./outputs/qat_out/
+
+sequence_len: 2048
+sample_packing: true
+flex_attention: true
+pad_to_sequence_len: true
+
+flex_attn_compile_kwargs:
+  dynamic: false
+  mode: max-autotune-no-cudagraphs
+
+qat:
+  activation_dtype: int8
+  weight_dtype: int4
+  group_size: 256
+  fake_quant_after_n_steps: 1000
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 1
+micro_batch_size: 2
+max_steps: 2000
+optimizer: adamw_torch_fused
+lr_scheduler: cosine
+learning_rate: 2e-5
+
+bf16: true
+tf32: true
+
+resume_from_checkpoint:
+logging_steps: 1
+
+evals_per_epoch: 1
+saves_per_epoch: 1
+
+warmup_steps: 10
+weight_decay: 0.0
+fsdp:
+  - full_shard
+  - auto_wrap
+
+fsdp_config:
+  fsdp_version: 2
+  fsdp_offload_params: false
+  fsdp_cpu_ram_efficient_loading: true
+  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
+  fsdp_transformer_layer_cls_to_wrap: Qwen3DecoderLayer
+  fsdp_state_dict_type: FULL_STATE_DICT
+  fsdp_sharding_strategy: FULL_SHARD
+  fsdp_reshard_after_forward: true
+  fsdp_activation_checkpointing: true
+
+special_tokens:

requirements.txt

@@ -6,21 +6,20 @@ triton>=3.0.0
 mamba-ssm==1.2.0.post1
 xformers>=0.0.23.post1
 autoawq==0.2.7.post3
-liger-kernel==0.5.9
+liger-kernel==0.5.10
 # END section
 
 packaging==23.2
 
-huggingface_hub==0.31.0
+huggingface_hub==0.32.2
 peft==0.15.2
-transformers==4.51.3
+transformers==4.52.4
 tokenizers>=0.21.1
-accelerate==1.6.0
-datasets==3.5.1
-deepspeed>=0.15.4
-trl==0.17.0
-hf_xet==1.1.0
-hqq==0.2.5
+accelerate==1.7.0
+datasets==3.6.0
+deepspeed>=0.17.0
+trl==0.18.2
+hf_xet==1.1.2
 
 optimum==1.16.2
 hf_transfer
@@ -63,8 +62,10 @@ langdetect==1.0.9
 immutabledict==4.2.0
 antlr4-python3-runtime==4.13.2
 
-torchao==0.9.0
+torchao==0.10.0
 schedulefree==1.4.1
 
 axolotl-contribs-lgpl==0.0.6
 axolotl-contribs-mit==0.0.3
+
+mistral-common==1.6.0

scripts/cutcrossentropy_install.py

@@ -9,6 +9,8 @@ except ImportError as exc:
     raise ImportError("Install torch via `pip install torch`") from exc
 from packaging.version import Version as V
 
+USE_UV = "--uv" in sys.argv[1:]
+
 v = V(torch.__version__)
 
 # no cut-cross-entropy support for torch < 2.4.0
@@ -23,7 +25,9 @@ if cce_spec:
     if not importlib.util.find_spec("cut_cross_entropy.transformers"):
         UNINSTALL_PREFIX = "pip uninstall -y cut-cross-entropy && "
 
+UV_PREFIX = "uv " if USE_UV else ""
+
 print(
     UNINSTALL_PREFIX
-    + 'pip install "cut-cross-entropy[transformers] @ git+https://github.com/apple/ml-cross-entropy.git@bad6f7b49c75fdec69471abb71b4cddd0f0c6438"'
+    + f'{UV_PREFIX}pip install "cut-cross-entropy[transformers] @ git+https://github.com/axolotl-ai-cloud/ml-cross-entropy.git@78b2a45713a54c9bedf8b33f5e31cf07a1a57154"'
 )

scripts/motd

@@ -11,7 +11,7 @@
                                  =@#       @#  #@=     #@   =#@@@@#=    +#@@=  +#@@@@#=    .##@@+   @@
     @@@@  @@@@@@@@@@@@@@@@
 
-Welcome to the axolotl cloud image! If the you've mounted a disk to /workspace and the axolotl directory ie empty, run the following commands:
+Welcome to the axolotl cloud image! If the you've mounted a disk to /workspace and the axolotl directory is empty, run the following commands:
 
 ```
 cd /workspace

scripts/unsloth_install.py

@@ -1,11 +1,15 @@
 # noqa
 # pylint: skip-file
+import sys
+
 try:
     import torch
 except ImportError:
     raise ImportError("Install torch via `pip install torch`")
 from packaging.version import Version as V
 
+use_uv = "--uv" in sys.argv[1:]
+
 v = V(torch.__version__)
 cuda = str(torch.version.cuda)
 try:
@@ -31,6 +35,7 @@ elif v < V("2.6.0"):
 else:
     raise RuntimeError(f"Torch = {v} too new!")
 x = x.format(cuda.replace(".", ""), "-ampere" if is_ampere else "")
+uv_prefix = "uv " if use_uv else ""
 print(
-    f'pip install unsloth-zoo==2024.12.1 && pip install --no-deps "unsloth[{x}]==2024.12.4"'
+    f'{uv_prefix}pip install unsloth-zoo==2024.12.1 && {uv_prefix}pip install --no-deps "unsloth[{x}]==2024.12.4"'
 )

setup.py

@@ -118,7 +118,7 @@ extras_require = {
         "yunchang==0.6.0",
     ],
     "deepspeed": [
-        "deepspeed==0.15.4",
+        "deepspeed==0.17.1",
         "deepspeed-kernels",
     ],
     "mamba-ssm": [

src/axolotl/__init__.py

@@ -4,4 +4,4 @@ import pkgutil
 
 __path__ = pkgutil.extend_path(__path__, __name__)  # Make this a namespace package
 
-__version__ = "0.10.0.dev0"
+__version__ = "0.11.0.dev"

src/axolotl/cli/args.py

@@ -28,7 +28,6 @@ class TrainerCliArgs:
     debug: bool = field(default=False)
     debug_text_only: bool = field(default=False)
     debug_num_examples: int = field(default=0)
-    merge_lora: bool = field(default=False)
     prompter: Optional[str] = field(default=None)
     shard: bool = field(default=False)
     main_process_port: Optional[int] = field(default=None)
@@ -82,6 +81,32 @@ class VllmServeCliArgs:
             "hardware support this feature."
         },
     )
+    serve_module: Optional[str] = field(
+        default=None,
+        metadata={
+            "help": "Module to serve. If not set, the default module will be used."
+        },
+    )
+
+    enable_reasoning: Optional[bool] = field(
+        default=None,
+    )
+
+    reasoning_parser: Optional[str] = field(
+        default=None,
+    )
+
+
+@dataclass
+class QuantizeCliArgs:
+    """Dataclass with CLI arguments for `axolotl quantize` command."""
+
+    base_model: Optional[str] = field(default=None)
+    weight_dtype: Optional[str] = field(default=None)
+    activation_dtype: Optional[str] = field(default=None)
+    quantize_embedding: Optional[bool] = field(default=None)
+    group_size: Optional[int] = field(default=None)
+    output_dir: Optional[str] = field(default=None)
 
 
 @dataclass

src/axolotl/cli/checks.py

@@ -1,14 +1,16 @@
 """Various checks for Axolotl CLI."""
 
-import logging
 import os
 from pathlib import Path
 
 from accelerate.commands.config import config_args
 from huggingface_hub import HfApi
 from huggingface_hub.utils import LocalTokenNotFoundError
+from requests import HTTPError
 
-LOG = logging.getLogger(__name__)
+from axolotl.utils.logging import get_logger
+
+LOG = get_logger(__name__)
 
 
 def check_accelerate_default_config() -> None:
@@ -45,3 +47,8 @@ def check_user_token() -> bool:
             "Error verifying HuggingFace token. Remember to log in using `huggingface-cli login` and get your access token from https://huggingface.co/settings/tokens if you want to use gated models or datasets."
         )
         return False
+    except HTTPError:
+        LOG.warning(
+            "Error accessing HuggingFace. This may be due to a network issue or rate limiting."
+        )
+        return False

src/axolotl/cli/cloud/__init__.py

@@ -7,7 +7,6 @@ from typing import Union
 
 import yaml
 
-from axolotl.cli.art import print_axolotl_text_art
 from axolotl.cli.cloud.modal_ import ModalCloud
 from axolotl.utils.dict import DictDefault
 
@@ -24,7 +23,6 @@ def do_cli_preprocess(
     cloud_config: Union[Path, str],
     config: Union[Path, str],
 ) -> None:
-    print_axolotl_text_art()
     cloud_cfg = load_cloud_cfg(cloud_config)
     cloud = ModalCloud(cloud_cfg)
     with open(config, "r", encoding="utf-8") as file:
@@ -39,7 +37,6 @@ def do_cli_train(
     cwd=None,
     **kwargs,
 ) -> None:
-    print_axolotl_text_art()
     cloud_cfg = load_cloud_cfg(cloud_config)
     cloud = ModalCloud(cloud_cfg)
     with open(config, "r", encoding="utf-8") as file:
@@ -54,7 +51,6 @@ def do_cli_lm_eval(
     cloud_config: Union[Path, str],
     config: Union[Path, str],
 ) -> None:
-    print_axolotl_text_art()
     cloud_cfg = load_cloud_cfg(cloud_config)
     cloud = ModalCloud(cloud_cfg)
     with open(config, "r", encoding="utf-8") as file:

src/axolotl/cli/cloud/modal_.py

@@ -82,7 +82,7 @@ class ModalCloud(Cloud):
         return res
 
     def get_image(self):
-        docker_tag = "main-py3.11-cu124-2.5.1"
+        docker_tag = "main-py3.11-cu124-2.6.0"
         if self.config.docker_tag:
             docker_tag = self.config.docker_tag
         docker_image = f"axolotlai/axolotl:{docker_tag}"

src/axolotl/cli/config.py

@@ -1,7 +1,6 @@
 """Configuration loading and processing."""
 
 import json
-import logging
 import os
 import tempfile
 from pathlib import Path
@@ -22,11 +21,14 @@ from axolotl.utils.config import (
     validate_config,
 )
 from axolotl.utils.dict import DictDefault
+from axolotl.utils.logging import get_logger
 from axolotl.utils.mlflow_ import setup_mlflow_env_vars
 from axolotl.utils.trainer import prepare_opinionated_env, prepare_optim_env
 from axolotl.utils.wandb_ import setup_wandb_env_vars
 
-LOG = logging.getLogger(__name__)
+LOG = get_logger(__name__)
+
+API_KEY_FIELDS = {"comet_api_key"}
 
 
 def check_remote_config(config: Union[str, Path]) -> Union[str, Path]:
@@ -119,12 +121,12 @@ def choose_config(path: Path) -> str:
         )
 
     if len(yaml_files) == 1:
-        print(f"Using default YAML file '{yaml_files[0]}'")
+        LOG.info(f"Using default YAML file '{yaml_files[0]}'")
         return str(yaml_files[0])
 
-    print("Choose a YAML file:")
+    LOG.info("Choose a YAML file:")
     for idx, file in enumerate(yaml_files):
-        print(f"{idx + 1}. {file}")
+        LOG.info(f"{idx + 1}. {file}")
 
     chosen_file = None
     while chosen_file is None:
@@ -133,9 +135,9 @@ def choose_config(path: Path) -> str:
             if 1 <= choice <= len(yaml_files):
                 chosen_file = str(yaml_files[choice - 1])
             else:
-                print("Invalid choice. Please choose a number from the list.")
+                LOG.info("Invalid choice. Please choose a number from the list.")
         except ValueError:
-            print("Invalid input. Please enter a number.")
+            LOG.info("Invalid input. Please enter a number.")
 
     return chosen_file
 
@@ -233,4 +235,15 @@ def load_cfg(
     setup_comet_env_vars(cfg)
     plugin_set_cfg(cfg)
 
+    cfg_to_log = {
+        k: "[REDACTED]" if k in API_KEY_FIELDS else v
+        for k, v in cfg.items()
+        if v is not None
+    }
+
+    LOG.info(
+        "config:\n%s",
+        json.dumps(cfg_to_log, indent=2, default=str, sort_keys=True),
+    )
+
     return cfg

src/axolotl/cli/evaluate.py

@@ -1,6 +1,5 @@
 """CLI to run evaluation on a model."""
 
-import logging
 import os
 from pathlib import Path
 from typing import Union
@@ -10,15 +9,15 @@ from dotenv import load_dotenv
 from transformers.hf_argparser import HfArgumentParser
 
 from axolotl.cli.args import TrainerCliArgs
-from axolotl.cli.art import print_axolotl_text_art
 from axolotl.cli.checks import check_accelerate_default_config, check_user_token
 from axolotl.cli.config import load_cfg
 from axolotl.common.datasets import load_datasets, load_preference_datasets
 from axolotl.evaluate import evaluate
 from axolotl.utils import patch_optimized_env
 from axolotl.utils.dict import DictDefault
+from axolotl.utils.logging import get_logger
 
-LOG = logging.getLogger(__name__)
+LOG = get_logger(__name__)
 
 
 def do_evaluate(cfg: DictDefault, cli_args: TrainerCliArgs) -> None:
@@ -35,7 +34,6 @@ def do_evaluate(cfg: DictDefault, cli_args: TrainerCliArgs) -> None:
     patch_optimized_env()
 
     # pylint: disable=duplicate-code
-    print_axolotl_text_art()
     check_accelerate_default_config()
     if int(os.getenv("LOCAL_RANK", "0")) == 0:
         check_user_token()

src/axolotl/cli/inference.py

@@ -1,7 +1,6 @@
 """CLI to run inference on a trained model."""
 
 import importlib
-import logging
 import sys
 from pathlib import Path
 from threading import Thread
@@ -14,7 +13,6 @@ from dotenv import load_dotenv
 from transformers import GenerationConfig, TextIteratorStreamer, TextStreamer
 
 from axolotl.cli.args import InferenceCliArgs
-from axolotl.cli.art import print_axolotl_text_art
 from axolotl.cli.config import load_cfg
 from axolotl.cli.utils import load_model_and_tokenizer
 from axolotl.utils.chat_templates import (
@@ -22,8 +20,9 @@ from axolotl.utils.chat_templates import (
     get_chat_template_from_config,
 )
 from axolotl.utils.dict import DictDefault
+from axolotl.utils.logging import get_logger
 
-LOG = logging.getLogger(__name__)
+LOG = get_logger(__name__)
 
 
 def get_multi_line_input() -> str:
@@ -255,7 +254,6 @@ def do_cli(
         kwargs: Additional keyword arguments to override config file values.
     """
     # pylint: disable=duplicate-code
-    print_axolotl_text_art()
     parsed_cfg = load_cfg(config, inference=True, rl=None, **kwargs)
     parsed_cfg.sample_packing = False
     parser = transformers.HfArgumentParser(InferenceCliArgs)

src/axolotl/cli/main.py

@@ -2,7 +2,6 @@
 
 # pylint: disable=redefined-outer-name
 
-import logging
 import os
 import subprocess  # nosec B404
 import tempfile
@@ -17,9 +16,11 @@ import axolotl
 from axolotl.cli.args import (
     EvaluateCliArgs,
     PreprocessCliArgs,
+    QuantizeCliArgs,
     TrainerCliArgs,
     VllmServeCliArgs,
 )
+from axolotl.cli.art import print_axolotl_text_art
 from axolotl.cli.sweeps import generate_sweep_configs
 from axolotl.cli.utils import (
     add_options_from_config,
@@ -30,13 +31,17 @@ from axolotl.cli.utils import (
 )
 from axolotl.integrations.lm_eval.cli import lm_eval
 from axolotl.utils import patch_optimized_env
+from axolotl.utils.logging import get_logger
 from axolotl.utils.schemas.config import AxolotlInputConfig
 
+LOG = get_logger(__name__)
+
 
 @click.group()
 @click.version_option(version=axolotl.__version__, prog_name="axolotl")
 def cli():
     """Axolotl CLI - Train and fine-tune large language models"""
+    print_axolotl_text_art()
 
 
 @cli.command()
@@ -176,7 +181,7 @@ def train(
 
                     do_cli(config=cfg_file, **kwargs)
         except subprocess.CalledProcessError as exc:
-            logging.error(f"Failed to train/fine-tune config '{cfg_file}': {exc}")
+            LOG.error(f"Failed to train/fine-tune config '{cfg_file}': {exc}")
             if not sweep:
                 raise exc
 
@@ -333,6 +338,16 @@ def vllm_serve(config: str, **cli_args: VllmServeCliArgs):
     do_vllm_serve(config, cli_args)
 
 
+@cli.command()
+@click.argument("config", type=click.Path(exists=True, path_type=str))
+@add_options_from_dataclass(QuantizeCliArgs)
+@filter_none_kwargs
+def quantize(config: str, **cli_args: QuantizeCliArgs):
+    from axolotl.cli.quantize import do_quantize
+
+    do_quantize(config, cli_args)
+
+
 @cli.command()
 @click.argument("model", type=click.Path(exists=True, path_type=str))
 @click.argument("output", type=click.Path(exists=False, path_type=str))

src/axolotl/cli/merge_lora.py

@@ -1,20 +1,17 @@
 """CLI to merge a trained LoRA into a base model."""
 
-import logging
 from pathlib import Path
 from typing import Union
 
 import fire
-import transformers
 from dotenv import load_dotenv
 
-from axolotl.cli.args import TrainerCliArgs
-from axolotl.cli.art import print_axolotl_text_art
 from axolotl.cli.config import load_cfg
 from axolotl.cli.utils import load_model_and_tokenizer
 from axolotl.utils.dict import DictDefault
+from axolotl.utils.logging import get_logger
 
-LOG = logging.getLogger(__name__)
+LOG = get_logger(__name__)
 
 
 def do_merge_lora(*, cfg: DictDefault) -> None:
@@ -25,8 +22,6 @@ def do_merge_lora(*, cfg: DictDefault) -> None:
     Args:
         cfg: Dictionary mapping `axolotl` config keys to values.
     """
-    print_axolotl_text_art()
-
     model, tokenizer, processor = load_model_and_tokenizer(cfg=cfg)
     safe_serialization = cfg.save_safetensors is True
 
@@ -68,12 +63,6 @@ def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs) -> None:
     Raises:
         ValueError: If target directory for LoRA merged model does not exist.
     """
-    # pylint: disable=duplicate-code
-    parser = transformers.HfArgumentParser(TrainerCliArgs)
-    parsed_cli_args, _ = parser.parse_args_into_dataclasses(
-        return_remaining_strings=True
-    )
-    parsed_cli_args.merge_lora = True
 
     parsed_cfg = load_cfg(
         config,

src/axolotl/cli/merge_sharded_fsdp_weights.py

@@ -1,7 +1,6 @@
 """CLI to merge sharded FSDP model checkpoints into a single combined checkpoint."""
 
 import json
-import logging
 import os
 import shutil
 from pathlib import Path
@@ -11,7 +10,6 @@ import fire
 import torch
 import torch.distributed.checkpoint as dist_cp
 import torch.distributed.checkpoint.format_utils as dist_cp_format_utils
-import transformers
 from accelerate.utils import (
     SAFE_WEIGHTS_INDEX_NAME,
     SAFE_WEIGHTS_NAME,
@@ -24,11 +22,10 @@ from huggingface_hub import split_torch_state_dict_into_shards
 from safetensors.torch import save_file as safe_save_file
 from torch.distributed.checkpoint.format_utils import _EmptyStateDictLoadPlanner
 
-from axolotl.cli.args import TrainerCliArgs
-from axolotl.cli.art import print_axolotl_text_art
 from axolotl.cli.config import load_cfg
+from axolotl.utils.logging import get_logger
 
-LOG = logging.getLogger(__name__)
+LOG = get_logger(__name__)
 
 
 class BFloat16CastPlanner(_EmptyStateDictLoadPlanner):
@@ -196,12 +193,6 @@ def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs):
         kwargs: Additional keyword arguments to override config file values.
     """
     # pylint: disable=duplicate-code
-    print_axolotl_text_art()
-    parser = transformers.HfArgumentParser(TrainerCliArgs)
-    parsed_cli_args, _ = parser.parse_args_into_dataclasses(
-        return_remaining_strings=True
-    )
-    parsed_cli_args.merge_lora = True
     parsed_cfg = load_cfg(config, **kwargs)
 
     fsdp_dir = Path(parsed_cfg.output_dir) / "pytorch_model_fsdp_0"

src/axolotl/cli/preprocess.py

@@ -1,6 +1,5 @@
 """CLI to run preprocessing of a dataset."""
 
-import logging
 import warnings
 from pathlib import Path
 from typing import Union
@@ -13,16 +12,16 @@ from dotenv import load_dotenv
 from transformers import AutoModelForCausalLM
 
 from axolotl.cli.args import PreprocessCliArgs
-from axolotl.cli.art import print_axolotl_text_art
 from axolotl.cli.checks import check_accelerate_default_config, check_user_token
 from axolotl.cli.config import load_cfg
 from axolotl.common.const import DEFAULT_DATASET_PREPARED_PATH
 from axolotl.common.datasets import load_datasets, load_preference_datasets
 from axolotl.integrations.base import PluginManager
 from axolotl.utils.dict import DictDefault
+from axolotl.utils.logging import get_logger
 from axolotl.utils.trainer import disable_datasets_caching
 
-LOG = logging.getLogger(__name__)
+LOG = get_logger(__name__)
 
 
 def do_preprocess(cfg: DictDefault, cli_args: PreprocessCliArgs) -> None:
@@ -33,7 +32,6 @@ def do_preprocess(cfg: DictDefault, cli_args: PreprocessCliArgs) -> None:
         cfg: Dictionary mapping `axolotl` config keys to values.
         cli_args: Preprocessing-specific CLI arguments.
     """
-    print_axolotl_text_art()
     check_accelerate_default_config()
     check_user_token()
 

src/axolotl/cli/quantize.py

@@ -0,0 +1,88 @@
+"""
+CLI to post-training quantize a model using torchao
+"""
+
+from pathlib import Path
+from typing import Union
+
+from transformers import AutoModelForCausalLM
+
+from axolotl.cli.config import load_cfg
+from axolotl.loaders import load_tokenizer
+from axolotl.utils.logging import get_logger
+from axolotl.utils.quantization import TorchIntDType, quantize_model_for_ptq
+
+LOG = get_logger(__name__)
+
+
+def do_quantize(
+    config: Union[Path, str],
+    cli_args: dict,
+):
+    """
+    Quantizes a model's model's weights
+
+    Args:
+        config (Union[Path, str]): The path to the config file
+        cli_args (dict): Additional command-line arguments
+    """
+
+    cfg = load_cfg(config)
+
+    if cfg.qat and cfg.quantization:
+        raise ValueError(
+            "QAT and quantization cannot be used together. Please specify only one of qat or quantization in your config file."
+        )
+
+    if cfg.qat:
+        quantize_cfg = cfg.qat
+    elif cfg.quantization:
+        quantize_cfg = cfg.quantization
+    else:
+        raise ValueError(
+            "No quantization configuration found. Please specify either qat or quantization in your config file."
+        )
+
+    model_path = cli_args.get("model_path") or cfg.output_dir
+    if weight_dtype := cli_args.get("weight_dtype"):
+        weight_dtype = TorchIntDType[weight_dtype]
+    else:
+        weight_dtype = quantize_cfg.weight_dtype
+    if activation_dtype := cli_args.get("activation_dtype"):
+        activation_dtype = TorchIntDType[activation_dtype]
+    else:
+        activation_dtype = quantize_cfg.activation_dtype
+    group_size = cli_args.get("group_size") or quantize_cfg.group_size
+    quantize_embedding = (
+        cli_args.get("quantize_embedding") or quantize_cfg.quantize_embedding
+    )
+    output_dir = cli_args.get("output_dir") or cfg.output_dir
+
+    LOG.info(f"Loading model from {model_path}...")
+    tokenizer = load_tokenizer(cfg)
+    model = AutoModelForCausalLM.from_pretrained(model_path, device_map="auto")
+
+    LOG.info(
+        f"Quantizing model with configuration: \n"
+        f"\tweight_dtype: {weight_dtype}\n"
+        f"\tactivation_dtype: {activation_dtype}\n"
+        f"\tgroup_size: {group_size}\n"
+        f"\tquantize_embedding: {quantize_embedding}"
+    )
+
+    quantize_model_for_ptq(
+        model, weight_dtype, group_size, activation_dtype, quantize_embedding
+    )
+
+    LOG.info(f"Saving quantized model to: {str(Path(output_dir) / 'quantized')}...")
+    model.save_pretrained(
+        str(Path(output_dir) / "quantized"),
+        safe_serialization=False,
+        progressbar=True,
+    )
+    tokenizer.save_pretrained(
+        str(Path(output_dir) / "quantized"),
+        safe_serialization=False,
+        progressbar=True,
+    )
+    LOG.info(f"Quantized model saved to: {str(Path(output_dir) / 'quantized')}...")

src/axolotl/cli/train.py

@@ -1,7 +1,6 @@
 """CLI to run training on a model."""
 
 import gc
-import logging
 import os
 from pathlib import Path
 from typing import Union
@@ -12,7 +11,6 @@ from dotenv import load_dotenv
 from transformers.hf_argparser import HfArgumentParser
 
 from axolotl.cli.args import TrainerCliArgs
-from axolotl.cli.art import print_axolotl_text_art
 from axolotl.cli.checks import check_accelerate_default_config, check_user_token
 from axolotl.cli.config import load_cfg
 from axolotl.common.datasets import load_datasets, load_preference_datasets
@@ -22,8 +20,6 @@ from axolotl.utils import patch_optimized_env
 from axolotl.utils.config import normalize_config, resolve_dtype
 from axolotl.utils.dict import DictDefault
 
-LOG = logging.getLogger(__name__)
-
 
 def do_train(cfg: DictDefault, cli_args: TrainerCliArgs):
     """
@@ -38,7 +34,6 @@ def do_train(cfg: DictDefault, cli_args: TrainerCliArgs):
     # Enable expandable segments for cuda allocation to improve VRAM usage
     patch_optimized_env()
 
-    print_axolotl_text_art()
     check_accelerate_default_config()
     if int(os.getenv("LOCAL_RANK", "0")) == 0:
         check_user_token()

src/axolotl/cli/utils.py

@@ -4,7 +4,6 @@ import concurrent.futures
 import dataclasses
 import hashlib
 import json
-import logging
 from functools import wraps
 from pathlib import Path
 from types import NoneType
@@ -20,10 +19,12 @@ from transformers import (
     ProcessorMixin,
 )
 
+from axolotl.loaders import load_processor, load_tokenizer
+from axolotl.loaders.model import ModelLoader
 from axolotl.utils.dict import DictDefault
-from axolotl.utils.models import load_model, load_processor, load_tokenizer
+from axolotl.utils.logging import get_logger
 
-LOG = logging.getLogger(__name__)
+LOG = get_logger(__name__)
 
 
 def strip_optional_type(field_type: type | str | None):
@@ -318,7 +319,8 @@ def load_model_and_tokenizer(
     tokenizer = load_tokenizer(cfg)
 
     LOG.info("loading model...")
-    model, _ = load_model(cfg, tokenizer, inference=inference)
+    model_loader = ModelLoader(cfg, tokenizer, inference=inference)
+    model, _ = model_loader.load()
 
     processor = None
     if cfg.is_multimodal:

src/axolotl/cli/vllm_serve.py

@@ -2,15 +2,27 @@
 CLI to start the vllm server for online RL
 """
 
+import os
+from dataclasses import dataclass, field
 from pathlib import Path
 from typing import Union
 
+import trl
 from trl.scripts.vllm_serve import ScriptArguments
-from trl.scripts.vllm_serve import main as vllm_serve_main
 
 from axolotl.cli.config import load_cfg
 
 
+@dataclass
+class AxolotlScriptArguments(ScriptArguments):
+    """
+    Additional arguments for the VLLM server
+    """
+
+    reasoning_parser: str = field(default="", kw_only=True)
+    enable_reasoning: bool | None = field(default=None, kw_only=True)
+
+
 def do_vllm_serve(
     config: Union[Path, str],
     cli_args: dict,
@@ -25,9 +37,13 @@ def do_vllm_serve(
     Returns:
         process_id: the process id of the started VLLM server
     """
+    patch_vllm_worker()
     cfg = load_cfg(config)
     model = cfg.base_model
 
+    serve_module = cli_args.get("serve_module", "trl.scripts.vllm_serve")
+    vllm_serve_main = getattr(__import__(serve_module, fromlist=["main"]), "main")
+
     tensor_parallel_size = (
         cli_args.get("tensor_parallel_size") or cfg.vllm.tensor_parallel_size
     )
@@ -41,9 +57,16 @@ def do_vllm_serve(
     enable_prefix_caching = (
         cli_args.get("enable_prefix_caching") or cfg.vllm.enable_prefix_caching
     )
+    reasoning_parser = (
+        cli_args.get("reasoning_parser") or cfg.vllm.reasoning_parser or ""
+    )
+    enable_reasoning = (
+        cli_args.get("enable_reasoning") or cfg.vllm.enable_reasoning or False
+    )
 
-    vllm_script_args = ScriptArguments(
-        model,
+    # pylint: disable=unexpected-keyword-arg
+    vllm_script_args = AxolotlScriptArguments(
+        model=model,
         tensor_parallel_size=tensor_parallel_size,
         host=host,
         port=port,
@@ -51,5 +74,67 @@ def do_vllm_serve(
         dtype=dtype,
         max_model_len=max_model_len,
         enable_prefix_caching=enable_prefix_caching,
+        reasoning_parser=reasoning_parser,
+        enable_reasoning=enable_reasoning,
     )
     vllm_serve_main(vllm_script_args)
+
+
+def patch_vllm_worker():
+    from multiprocessing.connection import Connection
+
+    from vllm import LLM
+
+    def llm_worker(
+        script_args: AxolotlScriptArguments,
+        data_parallel_rank: int,
+        master_port: int,
+        connection: Connection,
+    ) -> None:
+        # Set required environment variables for DP to work with vLLM
+        os.environ["VLLM_DP_RANK"] = str(data_parallel_rank)
+        os.environ["VLLM_DP_RANK_LOCAL"] = str(data_parallel_rank)
+        os.environ["VLLM_DP_SIZE"] = str(script_args.data_parallel_size)
+        os.environ["VLLM_DP_MASTER_PORT"] = str(master_port)
+
+        llm = LLM(
+            model=script_args.model,
+            revision=script_args.revision,
+            tensor_parallel_size=script_args.tensor_parallel_size,
+            gpu_memory_utilization=script_args.gpu_memory_utilization,
+            enforce_eager=script_args.enforce_eager,
+            dtype=script_args.dtype,
+            # Automatic Prefix Caching caches the KV cache of existing queries, so that a new query can
+            # directly reuse the KV cache if it shares the same prefix with one of the existing queries.
+            # This is particularly useful here because we generate completions from the same prompts.
+            enable_prefix_caching=script_args.enable_prefix_caching,
+            kv_cache_dtype=script_args.kv_cache_dtype,
+            max_model_len=script_args.max_model_len,
+            worker_extension_cls="trl.scripts.vllm_serve.WeightSyncWorkerExtension",
+            enable_reasoning=script_args.enable_reasoning,
+            reasoning_parser=script_args.reasoning_parser,
+        )
+
+        # Send ready signal to parent process
+        connection.send({"status": "ready"})
+
+        while True:
+            # Wait for commands from the parent process
+            try:
+                command = connection.recv()
+            except KeyboardInterrupt:
+                llm.collective_rpc(method="close_communicator")
+                break
+
+            # Handle commands
+            if command["type"] in ["call", "fire_and_forget"]:
+                method_name = command["method"]
+                args, kwargs = command.get("args", ()), command.get("kwargs", {})
+                method = getattr(llm, method_name)
+                result = method(*args, **kwargs)
+                if command["type"] == "call":
+                    connection.send(result)
+            elif command["type"] == "shutdown":
+                break
+
+    trl.scripts.vllm_serve.llm_worker = llm_worker

src/axolotl/common/const.py

@@ -1,5 +1,3 @@
-"""
-Various shared constants
-"""
+"""Various shared constants"""
 
 DEFAULT_DATASET_PREPARED_PATH = "last_run_prepared"

src/axolotl/common/datasets.py

@@ -1,22 +1,21 @@
 """Dataset loading utilities."""
 
-import logging
 import math
 import random
 from dataclasses import dataclass
-from typing import Optional, Union
 
 from datasets import Dataset
 
 import axolotl.monkeypatch.data.batch_dataset_fetcher  # pylint: disable=unused-import  # noqa: F401
 from axolotl.cli.args import PreprocessCliArgs, TrainerCliArgs
-from axolotl.utils.data import prepare_dataset
-from axolotl.utils.data.rl import load_prepare_preference_datasets
+from axolotl.loaders import load_processor, load_tokenizer
+from axolotl.utils.data import prepare_datasets, prepare_preference_datasets
 from axolotl.utils.dict import DictDefault
-from axolotl.utils.models import load_processor, load_tokenizer
+from axolotl.utils.logging import get_logger
+from axolotl.utils.schemas.enums import RLType
 from axolotl.utils.tokenization import check_dataset_labels
 
-LOG = logging.getLogger(__name__)
+LOG = get_logger(__name__)
 
 
 @dataclass
@@ -29,16 +28,7 @@ class TrainDatasetMeta:
 
 
 def sample_dataset(dataset: Dataset, num_samples: int) -> Dataset:
-    """
-    Randomly sample `num_samples` samples from `dataset`.
-
-    Args:
-        dataset: Dataset.
-        num_samples: Number of samples to return.
-
-    Returns:
-        Random sample (with replacement) of examples in `dataset`.
-    """
+    """Randomly sample `num_samples` samples with replacement from `dataset`."""
     return dataset.select(
         [random.randrange(0, len(dataset) - 1) for _ in range(num_samples)]  # nosec
     )
@@ -50,55 +40,52 @@ def load_datasets(
     cli_args: PreprocessCliArgs | TrainerCliArgs | None = None,
     debug: bool = False,
 ) -> TrainDatasetMeta:
-    """
-    Loads one or more training or evaluation datasets, calling
-    `axolotl.utils.data.prepare_dataset`. Optionally, logs out debug information.
+    """Loads one or more training or evaluation datasets, calling
+    `axolotl.utils.data.prepare_datasets`. Optionally, logs out debug information.
 
     Args:
         cfg: Dictionary mapping `axolotl` config keys to values.
         cli_args: Command-specific CLI arguments.
-        debug: Whether to print out tokenization of sample
+        debug: Whether to print out tokenization of sample. This is duplicated in
+            `cfg` and `cli_args`, but is kept due to use in our Colab notebooks.
 
     Returns:
         Dataclass with fields for training and evaluation datasets and the computed
-        `total_num_steps`.
+            `total_num_steps`.
     """
     tokenizer = load_tokenizer(cfg)
     processor = load_processor(cfg, tokenizer=tokenizer) if cfg.processor_type else None
-    preprocess_iterable = (
-        cli_args
-        and hasattr(cli_args, "iterable")
-        and cli_args.iterable is not None
-        and cli_args.iterable
-    )
+    preprocess_iterable = getattr(cli_args, "iterable", False)
 
-    train_dataset, eval_dataset, total_num_steps, prompters = prepare_dataset(
+    train_dataset, eval_dataset, total_num_steps, prompters = prepare_datasets(
         cfg,
         tokenizer,
         processor=processor,
         preprocess_iterable=preprocess_iterable,
     )
 
-    if (  # pylint: disable=too-many-boolean-expressions
-        cli_args
-        and (
-            cli_args.debug
-            or cfg.debug
-            or cli_args.debug_text_only
-            or int(cli_args.debug_num_examples) > 0
-        )
-    ) or debug:
+    if (
+        cfg.debug
+        or getattr(cli_args, "debug", False)
+        or getattr(cli_args, "debug_text_only", False)
+        or getattr(cli_args, "debug_num_examples", 0) > 0
+        or debug
+    ):
         LOG.info("check_dataset_labels...")
 
         num_examples = cli_args.debug_num_examples if cli_args else 1
         text_only = cli_args.debug_text_only if cli_args else False
-        train_samples = sample_dataset(train_dataset, num_examples)
-        check_dataset_labels(
-            train_samples,
-            tokenizer,
-            num_examples=num_examples,
-            text_only=text_only,
-        )
+        try:
+            train_samples = sample_dataset(train_dataset, num_examples)
+            check_dataset_labels(
+                train_samples,
+                tokenizer,
+                num_examples=num_examples,
+                text_only=text_only,
+            )
+        except AttributeError:
+            # can't sample iterable datasets
+            pass
 
         LOG.info("printing prompters...")
         for prompter in prompters:
@@ -112,13 +99,10 @@ def load_datasets(
 
 
 def load_preference_datasets(
-    *,
-    cfg: DictDefault,
-    cli_args: Union[PreprocessCliArgs, TrainerCliArgs],
+    *, cfg: DictDefault, cli_args: PreprocessCliArgs | TrainerCliArgs | None = None
 ) -> TrainDatasetMeta:
-    """
-    Loads one or more training or evaluation datasets for RL training using paired
-    preference data, calling `axolotl.utils.data.rl.load_prepare_preference_datasets`.
+    """Loads one or more training or evaluation datasets for RL training using paired
+    preference data, calling `axolotl.utils.data.rl.prepare_preference_datasets`.
     Optionally, logs out debug information.
 
     Args:
@@ -129,23 +113,28 @@ def load_preference_datasets(
         Dataclass with fields for training and evaluation datasets and the computed
         `total_num_steps`.
     """
-    train_dataset, eval_dataset = load_prepare_preference_datasets(cfg)
-    total_num_steps: Optional[int] = int(
-        math.ceil(len(train_dataset) * cfg.num_epochs / cfg.batch_size)
-    )
-    if cfg.rl == "grpo":
-        total_num_steps = None
+    tokenizer = load_tokenizer(cfg)
+    train_dataset, eval_dataset = prepare_preference_datasets(cfg, tokenizer)
 
-    if cli_args.debug or cfg.debug:
+    total_num_steps: int | None = None
+    if cfg.rl is not RLType.GRPO:
+        total_num_steps = int(
+            math.ceil(len(train_dataset) * cfg.num_epochs / cfg.batch_size)
+        )
+
+    if (cli_args and cli_args.debug) or cfg.debug:
         LOG.info("check_dataset_labels...")
 
+        num_examples = cli_args.debug_num_examples if cli_args else 1
+        text_only = cli_args.debug_text_only if cli_args else False
+
         tokenizer = load_tokenizer(cfg)
-        train_samples = sample_dataset(train_dataset, cli_args.debug_num_examples)
+        train_samples = sample_dataset(train_dataset, num_examples)
         check_dataset_labels(
-            train_samples,
-            tokenizer,
-            num_examples=cli_args.debug_num_examples,
-            text_only=cli_args.debug_text_only,
+            dataset=train_samples,
+            tokenizer=tokenizer,
+            num_examples=num_examples,
+            text_only=text_only,
             rl_mode=True,
         )
 

src/axolotl/core/builders/__init__.py

@@ -0,0 +1,6 @@
+"""Trainer builder classes"""
+
+from .causal import HFCausalTrainerBuilder
+from .rl import HFRLTrainerBuilder
+
+__all__ = ["HFCausalTrainerBuilder", "HFRLTrainerBuilder"]

src/axolotl/core/builders/base.py

@@ -0,0 +1,508 @@
+# Copyright 2024 Axolotl AI. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""Base class for trainer builder"""
+
+import abc
+import importlib
+import logging
+import sys
+from abc import abstractmethod
+from contextlib import suppress
+from pathlib import Path
+from typing import Any
+
+import torch
+from transformers import (
+    TrainerCallback,
+)
+from transformers.training_args import OptimizerNames
+
+from axolotl.integrations.base import PluginManager
+from axolotl.monkeypatch.trainer.lr import patch_trainer_get_lr
+from axolotl.utils import is_comet_available, is_mlflow_available
+from axolotl.utils.callbacks import (
+    GCCallback,
+    GPUStatsCallback,
+    SaveAxolotlConfigtoWandBCallback,
+)
+from axolotl.utils.callbacks.profiler import PytorchProfilerCallback
+from axolotl.utils.schemas.enums import CustomSupportedOptimizers
+
+LOG = logging.getLogger(__name__)
+
+with suppress(ImportError):
+    import torch._dynamo  # pylint: disable=ungrouped-imports
+
+
+class TrainerBuilderBase(abc.ABC):
+    """Base class for trainer builder."""
+
+    def __init__(self, cfg, model, tokenizer, processor=None):
+        self.cfg = cfg
+        self.model = model
+        self.tokenizer = tokenizer
+        self.processor = processor
+
+        self._train_dataset = None
+        self._eval_dataset = None
+        self._model_ref = None
+        self._peft_config = None
+
+        # If the model supports tagging, add the axolotl tag.
+        # This makes sure the tag is correctly pushed even if a user calls
+        # model.push_to_hub instead of trainer.push_to_hub.
+        if hasattr(model, "add_model_tags"):
+            model.add_model_tags(["axolotl"])
+
+        patch_trainer_get_lr()
+
+    @property
+    def model_ref(self):
+        return self._model_ref
+
+    @model_ref.setter
+    def model_ref(self, model):
+        self._model_ref = model
+
+    @property
+    def train_dataset(self):
+        return self._train_dataset
+
+    @train_dataset.setter
+    def train_dataset(self, dataset):
+        self._train_dataset = dataset
+
+    @property
+    def eval_dataset(self):
+        return self._eval_dataset
+
+    @eval_dataset.setter
+    def eval_dataset(self, dataset):
+        self._eval_dataset = dataset
+
+    @property
+    def peft_config(self):
+        return self._peft_config
+
+    @peft_config.setter
+    def peft_config(self, peft_config):
+        self._peft_config = peft_config
+
+    @abstractmethod
+    def build(self, total_num_steps):
+        pass
+
+    def get_callbacks(self) -> list[TrainerCallback]:
+        callbacks = []
+
+        plugin_manager = PluginManager.get_instance()
+        callbacks.extend(
+            plugin_manager.add_callbacks_pre_trainer(cfg=self.cfg, model=self.model)
+        )
+
+        if self.cfg.profiler_steps:
+            callbacks.append(
+                PytorchProfilerCallback(
+                    steps_to_profile=self.cfg.profiler_steps,
+                )
+            )
+
+        if self.cfg.gc_steps:
+            callbacks.append(GCCallback(gc_steps=self.cfg.gc_steps))
+
+        if self.cfg.use_wandb:
+            callbacks.append(
+                SaveAxolotlConfigtoWandBCallback(self.cfg.axolotl_config_path)
+            )
+        if self.cfg.use_mlflow and is_mlflow_available():
+            from axolotl.utils.callbacks.mlflow_ import (
+                SaveAxolotlConfigtoMlflowCallback,
+            )
+
+            callbacks.extend(
+                [
+                    SaveAxolotlConfigtoMlflowCallback(self.cfg.axolotl_config_path),
+                ]
+            )
+        if self.cfg.use_comet and is_comet_available():
+            from axolotl.utils.callbacks.comet_ import SaveAxolotlConfigtoCometCallback
+
+            callbacks.append(
+                SaveAxolotlConfigtoCometCallback(self.cfg.axolotl_config_path)
+            )
+
+        callbacks.append(GPUStatsCallback(cfg=self.cfg))
+
+        return callbacks
+
+    def get_post_trainer_create_callbacks(self, trainer):
+        """
+        Callbacks added after the trainer is created, usually b/c these need access to the trainer
+        """
+        callbacks = []
+        if self.cfg.plugins:
+            plugin_manager = PluginManager.get_instance()
+            callbacks.extend(
+                [
+                    cb
+                    for cb in plugin_manager.add_callbacks_post_trainer(
+                        self.cfg, trainer
+                    )
+                    if cb
+                ]
+            )
+        return callbacks
+
+    def hook_pre_create_training_args(self, training_arguments_kwargs):
+        # TODO
+        return training_arguments_kwargs
+
+    def hook_post_create_training_args(self, training_arguments):
+        # TODO
+        return training_arguments
+
+    def hook_pre_create_trainer(self, trainer_kwargs, trainer_cls):
+        # TODO
+        return trainer_kwargs, trainer_cls
+
+    def hook_post_create_trainer(self, trainer):
+        # TODO
+        return trainer
+
+    def _configure_warmup_and_logging(
+        self, total_num_steps: int, training_args_kwargs: dict
+    ):
+        warmup_steps = 0
+        warmup_ratio = 0.0
+        if self.cfg.warmup_steps:
+            warmup_steps = self.cfg.warmup_steps
+        elif self.cfg.warmup_ratio:
+            if total_num_steps:
+                warmup_steps = max(int(self.cfg.warmup_ratio * total_num_steps), 0)
+            else:
+                warmup_ratio = self.cfg.warmup_ratio
+        elif total_num_steps:
+            warmup_steps = min(int(0.03 * total_num_steps), 100)
+        else:
+            warmup_ratio = 0.03
+
+        if warmup_steps == 1:
+            warmup_steps = 2
+
+        if self.cfg.logging_steps is not None:
+            training_args_kwargs["logging_steps"] = self.cfg.logging_steps
+        else:
+            training_args_kwargs["logging_steps"] = (
+                500  # transformers defaults to 500
+                if not total_num_steps
+                else max(min(int(0.005 * total_num_steps), 10), 1)
+            )
+
+        training_args_kwargs["warmup_ratio"] = warmup_ratio
+        training_args_kwargs["warmup_steps"] = warmup_steps
+
+    def _configure_precision_settings(self, training_args_kwargs: dict):
+        training_args_kwargs["fp16"] = (self.cfg.fp16 and not self.cfg.bf16) or False
+        training_args_kwargs["tf32"] = self.cfg.tf32
+        if self.cfg.bf16 == "full":
+            training_args_kwargs["bf16_full_eval"] = True
+        else:
+            training_args_kwargs["bf16"] = self.cfg.bf16 or self.cfg.bfloat16
+
+    def _configure_scheduler(self, training_args_kwargs: dict):
+        if self.cfg.lr_scheduler in ["one_cycle", "rex"]:
+            training_args_kwargs["lr_scheduler_type"] = "cosine"
+            training_args_kwargs["alternate_lr_scheduler_type"] = self.cfg.lr_scheduler
+        else:
+            training_args_kwargs["lr_scheduler_type"] = (
+                self.cfg.lr_scheduler if self.cfg.lr_scheduler else "cosine"
+            )
+        training_args_kwargs["lr_scheduler_kwargs"] = (
+            self.cfg.lr_scheduler_kwargs if self.cfg.lr_scheduler_kwargs else {}
+        )
+
+    def _configure_optimizer(self, training_args_kwargs: dict, trainer_kwargs: dict):
+        def _configure_custom_optimizer(
+            training_args_kwargs: dict, trainer_kwargs: dict
+        ):
+            # Common optimizer kwargs
+            optimizer_kwargs = {
+                "lr": training_args_kwargs["learning_rate"],
+                "weight_decay": training_args_kwargs["weight_decay"],
+            }
+
+            # Adam-specific kwargs
+            adam_kwargs: dict = {}
+            if training_args_kwargs.get("adam_beta1") and training_args_kwargs.get(
+                "adam_beta2"
+            ):
+                adam_kwargs["betas"] = (
+                    training_args_kwargs.get("adam_beta1"),
+                    training_args_kwargs.get("adam_beta2"),
+                )
+            if training_args_kwargs.get("adam_epsilon"):
+                adam_kwargs["eps"] = training_args_kwargs.get("adam_epsilon")
+
+            if self.cfg.optimizer == "muon":
+                from axolotl.contribs.mit.muon import (  # pylint: disable=no-name-in-module
+                    MuonOptimizerFactory,
+                )
+
+                optimizer_cls = MuonOptimizerFactory
+                optimizer_kwargs.update(adam_kwargs)
+            elif self.cfg.optimizer == "optimi_adamw":
+                from optimi import AdamW
+
+                optimizer_kwargs["foreach"] = False
+                optimizer_cls = AdamW
+                optimizer_kwargs.update(adam_kwargs)
+            elif self.cfg.optimizer == "ao_adamw_4bit":
+                # TODO remove 20250401
+                from torchao.prototype.low_bit_optim import AdamW4bit
+
+                optimizer_cls = AdamW4bit
+                optimizer_kwargs.update(adam_kwargs)
+
+                LOG.warning(
+                    f"`ao_adamw_4bit` will be deprecated soon. Please use `{OptimizerNames.ADAMW_TORCH_4BIT}` instead."
+                )
+            elif self.cfg.optimizer == "ao_adamw_8bit":
+                from torchao.prototype.low_bit_optim import AdamW8bit
+
+                optimizer_cls = AdamW8bit
+                optimizer_kwargs.update(adam_kwargs)
+            elif self.cfg.optimizer == "ao_adamw_fp8":
+                from torchao.prototype.low_bit_optim import AdamWFp8
+
+                optimizer_cls = AdamWFp8
+                optimizer_kwargs.update(adam_kwargs)
+            elif self.cfg.optimizer == "adopt_adamw":
+                from axolotl.utils.optimizers.adopt import ADOPT
+
+                optimizer_cls = ADOPT
+                adam_kwargs["decouple"] = True
+                optimizer_kwargs.update(adam_kwargs)
+            elif self.cfg.optimizer == "came_pytorch":
+                from came_pytorch import CAME
+
+                optimizer_cls = CAME
+
+                beta1 = training_args_kwargs.get("adam_beta1", 0.9)
+                beta2 = training_args_kwargs.get("adam_beta2", 0.999)
+                beta3 = training_args_kwargs.get("adam_beta3", 0.9999)
+                eps1 = training_args_kwargs.get("adam_epsilon", 1e-30)
+                eps2 = training_args_kwargs.get("adam_epsilon2", 1e-16)
+                adam_kwargs["betas"] = (beta1, beta2, beta3)
+                adam_kwargs["eps"] = (eps1, eps2)
+
+                optimizer_kwargs.update(adam_kwargs)
+            else:
+                raise ValueError(
+                    f"Unhandled optimizer: {self.cfg.optimizer}. Please raise an Issue."
+                )
+
+            # Parse any additional optimizer args from config
+            if self.cfg.optim_args:
+                if isinstance(self.cfg.optim_args, dict):
+                    optimizer_kwargs.update(self.cfg.optim_args)
+                else:
+                    # Parse string format "key1=value1,key2=value2"
+                    for mapping in self.cfg.optim_args.replace(" ", "").split(","):
+                        key, value = mapping.split("=")
+                        optimizer_kwargs[key] = value
+
+            # Note: This is not used in training_args_kwargs, but in trainer_kwargs
+            trainer_kwargs["optimizer_cls_and_kwargs"] = (
+                optimizer_cls,
+                optimizer_kwargs,
+            )
+
+        # Handle custom optimizer
+        custom_supported_optimizers = [opt.value for opt in CustomSupportedOptimizers]
+        if self.cfg.optimizer in custom_supported_optimizers:
+            _configure_custom_optimizer(training_args_kwargs, trainer_kwargs)
+        else:
+            # Use transformers' optimizer
+            training_args_kwargs["optim"] = self.cfg.optimizer
+
+            # Parse any additional optimizer args from config
+            if self.cfg.optim_args:
+                if isinstance(self.cfg.optim_args, dict):
+                    optim_args = ",".join(
+                        [f"{key}={value}" for key, value in self.cfg.optim_args.items()]
+                    )
+                else:
+                    optim_args = self.cfg.optim_args
+                training_args_kwargs["optim_args"] = optim_args
+
+            if (
+                self.cfg.optimizer == "adamw_anyprecision"
+                and Path(self.cfg.torchdistx_path).exists()
+            ):
+                sys.path.append(self.cfg.torchdistx_path)
+                importlib.import_module("torchdistx")
+
+    def _configure_hub_parameters(self, training_args_kwargs: dict):
+        if self.cfg.hub_model_id:
+            training_args_kwargs["hub_model_id"] = self.cfg.hub_model_id
+            training_args_kwargs["push_to_hub"] = True
+            training_args_kwargs["hub_private_repo"] = True
+            training_args_kwargs["hub_always_push"] = True
+
+            if self.cfg.hub_strategy:
+                training_args_kwargs["hub_strategy"] = self.cfg.hub_strategy
+
+    def _configure_save_and_eval_strategy(self, training_args_kwargs: dict):
+        # save_strategy and save_steps
+        if self.cfg.save_steps:
+            training_args_kwargs["save_strategy"] = "steps"
+            training_args_kwargs["save_steps"] = self.cfg.save_steps
+        elif self.cfg.save_strategy:
+            training_args_kwargs["save_strategy"] = self.cfg.save_strategy
+        else:
+            # default to saving each epoch if not defined
+            training_args_kwargs["save_strategy"] = "epoch"
+
+        training_args_kwargs["save_total_limit"] = (
+            self.cfg.save_total_limit if self.cfg.save_total_limit else 4
+        )
+
+        # eval_strategy and eval_steps
+        if not self.eval_dataset and self.cfg.val_set_size == 0:
+            # do not eval if no eval_dataset and val_set_size=0
+            training_args_kwargs["eval_strategy"] = "no"
+        elif self.cfg.eval_steps:
+            training_args_kwargs["eval_strategy"] = "steps"
+            training_args_kwargs["eval_steps"] = self.cfg.eval_steps
+            training_args_kwargs["eval_on_start"] = True
+        elif self.cfg.eval_strategy:
+            training_args_kwargs["eval_strategy"] = self.cfg.eval_strategy
+            training_args_kwargs["eval_on_start"] = True
+
+    def _configure_reporting(self, training_args_kwargs: dict):
+        report_to = []
+        if self.cfg.use_wandb:
+            report_to.append("wandb")
+        if self.cfg.use_mlflow:
+            report_to.append("mlflow")
+        if self.cfg.use_tensorboard:
+            report_to.append("tensorboard")
+        if self.cfg.use_comet:
+            report_to.append("comet_ml")
+
+        training_args_kwargs["report_to"] = report_to
+
+        if self.cfg.use_wandb:
+            training_args_kwargs["run_name"] = self.cfg.wandb_name
+        elif self.cfg.use_mlflow:
+            training_args_kwargs["run_name"] = self.cfg.mlflow_run_name
+        else:
+            training_args_kwargs["run_name"] = None
+
+    def _configure_torch_compile(self, training_args_kwargs: dict):
+        if self.cfg.torch_compile and getattr(torch, "_dynamo", None):
+            torch._dynamo.config.suppress_errors = (  # pylint: disable=protected-access
+                True
+            )
+            training_args_kwargs["torch_compile"] = self.cfg.torch_compile
+            if self.cfg.torch_compile_backend:
+                training_args_kwargs["torch_compile_backend"] = (
+                    self.cfg.torch_compile_backend
+                )
+            if self.cfg.torch_compile_mode:
+                training_args_kwargs["torch_compile_mode"] = self.cfg.torch_compile_mode
+
+    def _configure_gradient_checkpointing(self, training_args_kwargs: dict):
+        if self.cfg.gradient_checkpointing:
+            training_args_kwargs["gradient_checkpointing"] = (
+                self.cfg.gradient_checkpointing
+            )
+            if self.cfg.gradient_checkpointing_kwargs is not None:
+                training_args_kwargs["gradient_checkpointing_kwargs"] = (
+                    self.cfg.gradient_checkpointing_kwargs
+                )
+            else:
+                training_args_kwargs["gradient_checkpointing_kwargs"] = {
+                    "use_reentrant": False
+                }
+
+    def _set_base_training_args(
+        self, total_num_steps
+    ) -> tuple[dict[str, Any], dict[str, Any]]:
+        training_args_kwargs: dict[str, Any] = {}
+        trainer_kwargs: dict[str, Any] = {}
+
+        self._configure_warmup_and_logging(total_num_steps, training_args_kwargs)
+        self._configure_precision_settings(training_args_kwargs)
+        self._configure_save_and_eval_strategy(training_args_kwargs)
+        self._configure_gradient_checkpointing(training_args_kwargs)
+
+        # set arg into trainer_args_kwargs with same name if value not None
+        for arg in [
+            # optim/scheduler
+            "adam_beta1",
+            "adam_beta2",
+            "adam_beta3",
+            "adam_epsilon",
+            "adam_epsilon2",
+            "cosine_min_lr_ratio",
+            "cosine_constant_lr_ratio",
+            "optim_target_modules",
+            # trainer
+            "max_grad_norm",
+            "dataloader_num_workers",
+            "dataloader_pin_memory",
+            "dataloader_prefetch_factor",
+            "gradient_accumulation_steps",
+            "learning_rate",
+            "embedding_lr",
+            "embedding_lr_scale",
+            "lr_groups",
+            "loraplus_lr_ratio",
+            "loraplus_lr_embedding",
+            "output_dir",
+            "save_safetensors",
+            "save_only_model",
+            "include_tokens_per_second",
+            "weight_decay",
+            "seed",
+        ]:
+            if hasattr(self.cfg, arg) and getattr(self.cfg, arg) is not None:
+                training_args_kwargs[arg] = getattr(self.cfg, arg)
+
+        training_args_kwargs["per_device_train_batch_size"] = self.cfg.micro_batch_size
+
+        if self.cfg.eval_batch_size:
+            training_args_kwargs["per_device_eval_batch_size"] = (
+                self.cfg.eval_batch_size
+            )
+
+        training_args_kwargs["max_steps"] = self.cfg.max_steps or total_num_steps or -1
+        training_args_kwargs["num_train_epochs"] = self.cfg.num_epochs
+
+        if self.cfg.dataset_processes:
+            training_args_kwargs["dataset_num_proc"] = self.cfg.dataset_processes
+
+        # max_length is not used in CausalTrainer
+        if self.cfg.reward_model or self.cfg.rl:
+            training_args_kwargs["max_length"] = self.cfg.sequence_len
+
+        self._configure_reporting(training_args_kwargs)
+        self._configure_hub_parameters(training_args_kwargs)
+        self._configure_scheduler(training_args_kwargs)
+        self._configure_optimizer(training_args_kwargs, trainer_kwargs)
+        self._configure_torch_compile(training_args_kwargs)
+
+        return training_args_kwargs, trainer_kwargs

src/axolotl/core/builders/causal.py

@@ -0,0 +1,489 @@
+"""Builder for causal trainers"""
+
+import inspect
+import math
+import os
+from pathlib import Path
+from typing import Type, Union
+
+import transformers
+from transformers import (
+    DataCollatorWithFlattening,
+    EarlyStoppingCallback,
+)
+from trl.trainer.utils import RewardDataCollatorWithPadding
+
+from axolotl.core.builders.base import TrainerBuilderBase
+from axolotl.core.trainers import (
+    AxolotlMambaTrainer,
+    AxolotlPRMTrainer,
+    AxolotlRewardTrainer,
+    AxolotlTrainer,
+    ReLoRATrainer,
+)
+from axolotl.integrations.base import PluginManager
+from axolotl.monkeypatch.multipack import SUPPORTED_MULTIPACK_MODEL_TYPES
+from axolotl.monkeypatch.relora import ReLoRACallback
+from axolotl.processing_strategies import get_processing_strategy
+from axolotl.utils import is_comet_available, is_mlflow_available
+from axolotl.utils.callbacks import (
+    LossWatchDogCallback,
+    SaveBetterTransformerModelCallback,
+    bench_eval_callback_factory,
+    causal_lm_bench_eval_callback_factory,
+    colab_inference_post_train_callback,
+    log_prediction_callback_factory,
+)
+from axolotl.utils.callbacks.lisa import lisa_callback_factory
+from axolotl.utils.callbacks.qat import QATCallback
+from axolotl.utils.chat_templates import get_chat_template_from_config
+from axolotl.utils.collators import (
+    BatchSamplerDataCollatorForSeq2Seq,
+    DataCollatorForSeq2Seq,
+    MambaDataCollator,
+    V2BatchSamplerDataCollatorForSeq2Seq,
+)
+from axolotl.utils.collators.mm_chat import MultiModalChatDataCollator
+from axolotl.utils.logging import get_logger
+
+LOG = get_logger(__name__)
+
+
+class HFCausalTrainerBuilder(TrainerBuilderBase):
+    """
+    Build the HuggingFace training args/trainer for causal models and reward modeling
+    using TRL.
+    """
+
+    def get_callbacks(self):
+        callbacks = super().get_callbacks()
+
+        if self.cfg.relora_steps:
+            callbacks.append(ReLoRACallback(self.cfg))
+
+        if (
+            hasattr(self.model, "use_bettertransformer")
+            and self.model.use_bettertransformer is True
+        ):
+            callbacks.append(SaveBetterTransformerModelCallback())
+
+        # TODO: check if can move to base class
+        if self.cfg.loss_watchdog_threshold is not None:
+            callbacks.append(LossWatchDogCallback(self.cfg))
+
+        if self.cfg.qat:
+            callbacks.append(QATCallback(self.cfg.qat))
+
+        return callbacks
+
+    def get_post_trainer_create_callbacks(self, trainer):
+        callbacks = []
+        if self.cfg.use_wandb and self.cfg.eval_table_size > 0:
+            LogPredictionCallback = log_prediction_callback_factory(
+                trainer, self.tokenizer, "wandb"
+            )
+            callbacks.append(LogPredictionCallback(self.cfg))
+        if (
+            self.cfg.use_mlflow
+            and is_mlflow_available()
+            and self.cfg.eval_table_size > 0
+        ):
+            LogPredictionCallback = log_prediction_callback_factory(
+                trainer, self.tokenizer, "mlflow"
+            )
+            callbacks.append(LogPredictionCallback(self.cfg))
+        if self.cfg.use_comet and is_comet_available() and self.cfg.eval_table_size > 0:
+            LogPredictionCallback = log_prediction_callback_factory(
+                trainer, self.tokenizer, "comet_ml"
+            )
+            callbacks.append(LogPredictionCallback(self.cfg))
+
+        if self.cfg.do_bench_eval:
+            callbacks.append(bench_eval_callback_factory(trainer, self.tokenizer))
+        if self.cfg.do_causal_lm_eval:
+            CausalLMBenchEvalCallback = causal_lm_bench_eval_callback_factory(
+                trainer, self.tokenizer
+            )
+            callbacks.append(CausalLMBenchEvalCallback(self.cfg))
+
+        if self.cfg.early_stopping_patience:
+            early_stop_cb = EarlyStoppingCallback(
+                self.cfg.early_stopping_patience,
+            )
+            callbacks.append(early_stop_cb)
+
+        if self.cfg.lisa_step_interval and self.cfg.lisa_n_layers:
+            callbacks.append(lisa_callback_factory(trainer))
+
+        if any("COLAB_" in key for key in os.environ):
+            ColabCallback = colab_inference_post_train_callback(trainer)
+            callbacks.append(ColabCallback(self.cfg))
+
+        callbacks.extend(super().get_post_trainer_create_callbacks(trainer=trainer))
+        return callbacks
+
+    def _get_trainer_cls(self):
+        """
+        Gets the trainer class for the given configuration.
+        """
+        if self.cfg.plugins:
+            plugin_manager = PluginManager.get_instance()
+            trainer_cls = plugin_manager.get_trainer_cls(self.cfg)
+            if trainer_cls:
+                return trainer_cls
+        if self.cfg.relora_steps:
+            return ReLoRATrainer
+        if self.cfg.model_config_type == "mamba":
+            return AxolotlMambaTrainer
+        if self.cfg.reward_model:
+            return AxolotlRewardTrainer
+        if self.cfg.process_reward_model:
+            return AxolotlPRMTrainer
+        return AxolotlTrainer
+
+    def build(self, total_num_steps):
+        from axolotl.core.training_args import (
+            AxolotlPRMConfig,
+            AxolotlRewardConfig,
+            AxolotlTrainingArguments,
+        )
+
+        training_arguments_kwargs, trainer_kwargs = self._set_base_training_args(
+            total_num_steps
+        )
+
+        if self.cfg.fsdp:
+            training_arguments_kwargs["fsdp"] = self.cfg.fsdp
+            if self.cfg.fsdp_config:
+                training_arguments_kwargs["fsdp_config"] = {
+                    k.lstrip("fsdp_"): v for k, v in dict(self.cfg.fsdp_config).items()
+                }
+
+        if self.cfg.adapter == "qlora":
+            training_arguments_kwargs["qlora"] = True
+
+        # deepspeed
+        if self.cfg.deepspeed:
+            training_arguments_kwargs["deepspeed"] = self.cfg.deepspeed
+
+        if self.cfg.lr_quadratic_warmup is not None:
+            training_arguments_kwargs["lr_quadratic_warmup"] = (
+                self.cfg.lr_quadratic_warmup
+            )
+
+        if self.cfg.dataloader_drop_last is not None:
+            training_arguments_kwargs["dataloader_drop_last"] = (
+                self.cfg.dataloader_drop_last
+            )
+        elif self.cfg.sample_packing and self.cfg.eval_sample_packing is False:
+            training_arguments_kwargs["dataloader_drop_last"] = True
+
+        if self.cfg.remove_unused_columns is not None:
+            training_arguments_kwargs["remove_unused_columns"] = (
+                self.cfg.remove_unused_columns
+            )
+
+        if self.cfg.do_bench_eval:
+            training_arguments_kwargs["do_bench_eval"] = self.cfg.do_bench_eval
+            if self.cfg.bench_dataset:
+                training_arguments_kwargs["bench_dataset"] = self.cfg.bench_dataset
+        if self.cfg.do_causal_lm_eval:
+            training_arguments_kwargs["do_causal_lm_eval"] = self.cfg.do_causal_lm_eval
+        if self.cfg.metric_for_best_model:
+            training_arguments_kwargs["metric_for_best_model"] = (
+                self.cfg.metric_for_best_model
+            )
+        if self.cfg.greater_is_better:
+            training_arguments_kwargs["greater_is_better"] = self.cfg.greater_is_better
+
+        # DDP Config
+        if self.cfg.ddp_timeout:
+            training_arguments_kwargs["ddp_timeout"] = self.cfg.ddp_timeout
+        # see https://pytorch.org/docs/stable/generated/torch.nn.parallel.DistributedDataParallel.html
+        if self.cfg.ddp_bucket_cap_mb:
+            training_arguments_kwargs["ddp_bucket_cap_mb"] = self.cfg.ddp_bucket_cap_mb
+        if self.cfg.ddp_broadcast_buffers is not None:
+            training_arguments_kwargs["ddp_broadcast_buffers"] = (
+                self.cfg.ddp_broadcast_buffers
+            )
+
+        # these are all the "standard" kwargs that are def used
+        training_arguments_kwargs["max_seq_length"] = self.cfg.sequence_len
+
+        if self.cfg.auto_find_batch_size is not None:
+            training_arguments_kwargs["auto_find_batch_size"] = (
+                self.cfg.auto_find_batch_size
+            )
+
+        training_arguments_kwargs["eval_accumulation_steps"] = (
+            self.cfg.gradient_accumulation_steps
+        )
+
+        training_arguments_kwargs["load_best_model_at_end"] = (
+            (
+                self.cfg.load_best_model_at_end is not False
+                or self.cfg.early_stopping_patience
+            )
+            and (
+                (not self.cfg.test_datasets and self.cfg.val_set_size > 0)
+                or (self.cfg.test_datasets and self.cfg.val_set_size == 0)
+            )
+            and self.cfg.save_steps
+            and self.cfg.eval_steps
+            and self.cfg.save_steps % self.cfg.eval_steps == 0
+        ) or False
+
+        # handle ddp
+        ddp_find_unused_parameters = None
+        if self.cfg.ddp:
+            ddp_find_unused_parameters = bool(self.cfg.ddp_find_unused_parameters)
+        training_arguments_kwargs["ddp_find_unused_parameters"] = (
+            ddp_find_unused_parameters
+        )
+
+        training_arguments_kwargs["group_by_length"] = self.cfg.group_by_length
+        training_arguments_kwargs["curriculum_sampling"] = self.cfg.curriculum_sampling
+
+        training_arguments_kwargs["sample_packing"] = bool(self.cfg.sample_packing)
+        training_arguments_kwargs["multipack_real_batches"] = (
+            self.cfg.multipack_real_batches
+            if self.cfg.multipack_real_batches is not None
+            else not self.cfg.flash_attention
+        )
+        training_arguments_kwargs["eval_sample_packing"] = bool(
+            self.cfg.eval_sample_packing
+        )
+        if self.cfg.sample_packing_bin_size is not None:
+            training_arguments_kwargs["sample_packing_bin_size"] = (
+                self.cfg.sample_packing_bin_size
+            )
+        if self.cfg.sample_packing_group_size is not None:
+            training_arguments_kwargs["sample_packing_group_size"] = (
+                self.cfg.sample_packing_group_size
+            )
+        if self.cfg.sample_packing_eff_est:
+            training_arguments_kwargs["sample_packing_efficiency"] = (
+                self.cfg.sample_packing_eff_est
+            )
+
+        if self.cfg.relora_steps:
+            training_arguments_kwargs["relora_steps"] = self.cfg.relora_steps
+            training_arguments_kwargs["relora_warmup_steps"] = (
+                self.cfg.relora_warmup_steps
+            )
+            if self.cfg.relora_anneal_steps:
+                training_arguments_kwargs["relora_anneal_steps"] = (
+                    self.cfg.relora_anneal_steps
+                )
+            if self.cfg.relora_prune_ratio:
+                training_arguments_kwargs["relora_prune_ratio"] = (
+                    self.cfg.relora_prune_ratio
+                )
+
+        if self.cfg.lisa_step_interval and self.cfg.lisa_n_layers:
+            training_arguments_kwargs["lisa_n_layers"] = self.cfg.lisa_n_layers
+            training_arguments_kwargs["lisa_step_interval"] = (
+                self.cfg.lisa_step_interval
+            )
+            training_arguments_kwargs["lisa_layers_attribute"] = (
+                self.cfg.lisa_layers_attribute
+            )
+
+        training_arguments_kwargs = self.hook_pre_create_training_args(
+            training_arguments_kwargs
+        )
+        training_arguments_kwargs["model_type"] = self.cfg.model_config_type
+        training_arguments_kwargs["pretraining"] = bool(self.cfg.pretraining_dataset)
+        if self.cfg.chat_template:
+            training_arguments_kwargs["chat_template"] = get_chat_template_from_config(
+                cfg=self.cfg,
+                tokenizer=self.tokenizer,
+            )
+
+        if self.cfg.neftune_noise_alpha is not None:
+            training_arguments_kwargs["neftune_noise_alpha"] = (
+                self.cfg.neftune_noise_alpha
+            )
+
+        if self.cfg.accelerator_config:
+            training_arguments_kwargs["accelerator_config"] = (
+                self.cfg.accelerator_config
+            )
+
+        if self.cfg.image_size:
+            training_arguments_kwargs["image_size"] = self.cfg.image_size
+        if self.cfg.image_resize_algorithm:
+            training_arguments_kwargs["image_resize_algorithm"] = (
+                self.cfg.image_resize_algorithm
+            )
+
+        if self.cfg.plugins:
+            plugin_manager = PluginManager.get_instance()
+            plugin_training_args = plugin_manager.get_training_args(self.cfg)
+            if plugin_training_args:
+                training_arguments_kwargs.update(plugin_training_args)
+
+        if self.cfg.reward_model:
+            training_args_cls = AxolotlRewardConfig
+        elif self.cfg.process_reward_model:
+            training_args_cls = AxolotlPRMConfig
+        else:
+            training_args_cls = AxolotlTrainingArguments
+        training_args = training_args_cls(  # pylint: disable=unexpected-keyword-arg
+            **training_arguments_kwargs,
+        )
+        training_args = self.hook_post_create_training_args(training_args)
+
+        # unset run_name so wandb sets up experiment names
+        if self.cfg.use_wandb and training_args.run_name == training_args.output_dir:
+            training_args.run_name = (  # pylint: disable=attribute-defined-outside-init
+                None
+            )
+
+        data_collator_kwargs = {
+            "padding": True,  # True/"longest" is the default
+        }
+        multiple = 64
+        if self.cfg.pad_to_sequence_len:
+            data_collator_kwargs["pad_to_multiple_of"] = multiple * math.ceil(
+                self.cfg.sequence_len / multiple
+            )
+        else:
+            # A100 is best at 64, while others at 8. Let's use the larger so we don't have to check
+            # https://docs.nvidia.com/deeplearning/performance/dl-performance-matrix-multiplication/index.html
+            data_collator_kwargs["pad_to_multiple_of"] = multiple
+
+        trainer_cls = self._get_trainer_cls()
+
+        trainer_kwargs, trainer_cls = self.hook_pre_create_trainer(
+            trainer_kwargs, trainer_cls
+        )
+        if eval_data_collator := self.build_collator(
+            training_args, is_eval=True, **data_collator_kwargs
+        ):
+            if not (self.cfg.reward_model or self.cfg.process_reward_model):
+                trainer_kwargs["eval_data_collator"] = eval_data_collator
+        if not (self.cfg.reward_model or self.cfg.process_reward_model):
+            trainer_kwargs["bench_data_collator"] = transformers.DataCollatorForSeq2Seq(
+                self.tokenizer,
+                return_tensors="pt",
+                **data_collator_kwargs,
+            )
+        sig = inspect.signature(trainer_cls)
+        if "processing_class" in sig.parameters:
+            trainer_kwargs["processing_class"] = self.tokenizer
+        elif "tokenizer" in sig.parameters:
+            trainer_kwargs["tokenizer"] = self.tokenizer
+        if (
+            trainer_cls not in [AxolotlRewardTrainer, AxolotlPRMTrainer]
+            and self.cfg.datasets is not None
+        ):
+            trainer_kwargs["dataset_tags"] = [
+                d["path"] for d in self.cfg.datasets if not Path(d["path"]).is_dir()
+            ]
+        trainer = trainer_cls(
+            model=self.model,
+            train_dataset=self.train_dataset,
+            eval_dataset=self.eval_dataset,
+            args=training_args,
+            data_collator=self.build_collator(training_args, **data_collator_kwargs),
+            callbacks=self.get_callbacks(),
+            **trainer_kwargs,
+        )
+        trainer = self.hook_post_create_trainer(trainer)
+        for callback in self.get_post_trainer_create_callbacks(trainer):
+            trainer.add_callback(callback)
+
+        if self.cfg.deepspeed and self.cfg.sample_packing:
+            trainer.accelerator.state.deepspeed_plugin.deepspeed_config[
+                "train_micro_batch_size_per_gpu"
+            ] = self.cfg.micro_batch_size
+
+        return trainer
+
+    def build_collator(
+        self,
+        training_args,  # type: "AxolotlTrainingArguments"  # type: ignore
+        is_eval=False,
+        **kwargs,
+    ):
+        if training_args.pretraining:
+            if (
+                self.cfg.pretraining_sample_concatenation is False
+                or self.cfg.micro_batch_size > 1
+            ):
+                return DataCollatorForSeq2Seq(self.tokenizer, **kwargs)
+            if not (self.cfg.sample_packing and self.cfg.pretrain_multipack_attn):
+                return None
+
+        if self.cfg.model_config_type == "mamba":
+            return MambaDataCollator(tokenizer=self.tokenizer)
+
+        use_batch_sampler_collator = False
+        if is_eval is False and training_args.sample_packing:
+            use_batch_sampler_collator = True
+        if is_eval and training_args.eval_sample_packing:
+            use_batch_sampler_collator = True
+
+        collator: Type[
+            Union[
+                V2BatchSamplerDataCollatorForSeq2Seq,
+                BatchSamplerDataCollatorForSeq2Seq,
+                DataCollatorForSeq2Seq,
+                DataCollatorWithFlattening,
+                RewardDataCollatorWithPadding,
+            ]
+        ]
+        collator_args = [self.tokenizer]
+
+        collator_cls_and_kwargs = None
+        if self.cfg.plugins:
+            plugin_manager = PluginManager.get_instance()
+            collator_cls_and_kwargs = plugin_manager.get_collator_cls_and_kwargs(
+                self.cfg, is_eval=is_eval
+            )
+
+        if collator_cls_and_kwargs:
+            collator = collator_cls_and_kwargs[0]
+            if kwargs and isinstance(kwargs, dict):
+                kwargs.update(collator_cls_and_kwargs[1])
+        elif self.cfg.reward_model:
+            collator = RewardDataCollatorWithPadding
+        elif use_batch_sampler_collator:
+            # Use V2BatchSamplerDataCollatorForSeq2Seq for flex attention,
+            # supported multipack models, or non-flash-attention llama
+            if (
+                self.cfg.flex_attention
+                or self.cfg.model_config_type in SUPPORTED_MULTIPACK_MODEL_TYPES
+                or (
+                    self.cfg.model_config_type in ["llama"]
+                    and self.cfg.flash_attention is not True
+                )
+            ):
+                collator = V2BatchSamplerDataCollatorForSeq2Seq
+            else:
+                collator = BatchSamplerDataCollatorForSeq2Seq
+        else:
+            if self.cfg.processor_type and self.processor:
+                collator = MultiModalChatDataCollator
+                kwargs["processing_strategy"] = get_processing_strategy(
+                    self.processor,
+                    training_args.chat_template,
+                    self.cfg.chat_template,
+                    image_size=training_args.image_size,
+                    image_resize_algorithm=training_args.image_resize_algorithm,
+                )
+            elif self.cfg.batch_flattening:
+                collator = DataCollatorWithFlattening
+                collator_args.pop(0)
+                kwargs.pop("pad_to_multiple_of", None)
+                kwargs.pop("padding", None)
+            else:
+                collator = DataCollatorForSeq2Seq
+
+        kwargs["return_tensors"] = "pt"
+
+        return collator(
+            *collator_args,
+            **kwargs,
+        )

src/axolotl/core/builders/rl.py

@@ -0,0 +1,238 @@
+"""Builder for RLHF trainers"""
+
+import inspect
+from pathlib import Path
+
+from axolotl.core.builders.base import TrainerBuilderBase
+from axolotl.core.trainers import (
+    AxolotlCPOTrainer,
+    AxolotlKTOTrainer,
+    AxolotlORPOTrainer,
+)
+from axolotl.core.trainers.dpo import DPOStrategy
+from axolotl.core.trainers.dpo.args import AxolotlDPOConfig
+from axolotl.core.trainers.grpo import GRPOStrategy
+from axolotl.integrations.base import PluginManager
+from axolotl.loaders.utils import ensure_dtype
+from axolotl.utils.callbacks.qat import QATCallback
+from axolotl.utils.logging import get_logger
+from axolotl.utils.schemas.enums import RLType
+
+LOG = get_logger(__name__)
+
+
+class HFRLTrainerBuilder(TrainerBuilderBase):
+    """Trainer factory class for TRL-based RLHF trainers (e.g. DPO)"""
+
+    def get_callbacks(self):
+        callbacks = super().get_callbacks()
+
+        if self.cfg.qat:
+            callbacks.append(QATCallback(self.cfg.qat))
+
+        return callbacks
+
+    def get_post_trainer_create_callbacks(self, trainer):
+        callbacks = super().get_post_trainer_create_callbacks(trainer=trainer)
+        return callbacks
+
+    def _get_trainer_cls(self, trainer_kwargs: dict):
+        """
+        Returns trainer_cls and trainer_cls_args
+        """
+        if self.cfg.plugins:
+            plugin_manager = PluginManager.get_instance()
+            trainer_cls = plugin_manager.get_trainer_cls(self.cfg)
+            trainer_cls_args = []  # type: ignore
+
+            if trainer_cls is not None:
+                return trainer_cls, trainer_cls_args
+
+        trainer_cls = None
+        trainer_cls_args = [self.model]
+
+        if self.cfg.rl is RLType.GRPO:
+            trainer_cls = GRPOStrategy.get_trainer_class(
+                sequence_parallel=self.cfg.sequence_parallel_degree > 1
+            )
+            trainer_cls_args.extend(GRPOStrategy.set_trainer_args(self.cfg))
+
+            trainer_kwargs.update(GRPOStrategy.set_trainer_kwargs(self.cfg))
+
+        elif self.cfg.rl in [RLType.DPO, RLType.IPO]:
+            trainer_cls = DPOStrategy.get_trainer_class()
+            trainer_cls_args.append(self.model_ref)
+
+        elif self.cfg.rl is RLType.ORPO:
+            trainer_cls = AxolotlORPOTrainer
+        elif self.cfg.rl is RLType.KTO:
+            trainer_cls = AxolotlKTOTrainer
+        elif self.cfg.rl is RLType.SIMPO:
+            trainer_cls = AxolotlCPOTrainer
+        else:
+            raise ValueError(f"Unsupported RL: {self.cfg.rl}")
+
+        return trainer_cls, trainer_cls_args
+
+    def _build_training_arguments(self, total_num_steps):
+        """
+        Returns training_args and trainer_kwargs
+        """
+        from axolotl.core.training_args import (
+            AxolotlCPOConfig,
+            AxolotlKTOConfig,
+            AxolotlORPOConfig,
+        )
+
+        training_args_kwargs, trainer_kwargs = self._set_base_training_args(
+            total_num_steps=total_num_steps
+        )
+
+        if self.cfg.remove_unused_columns is not None:
+            training_args_kwargs["remove_unused_columns"] = (
+                self.cfg.remove_unused_columns
+            )
+        else:
+            training_args_kwargs["remove_unused_columns"] = False
+
+        if self.cfg.trl and self.cfg.trl.beta is not None:
+            training_args_kwargs["beta"] = self.cfg.trl.beta
+        elif self.cfg.rl_beta is not None:
+            training_args_kwargs["beta"] = self.cfg.rl_beta
+        elif self.cfg.orpo_alpha is not None:
+            # trl does some odd mapping of alpha to beta to reuse the beta parameter ???
+            training_args_kwargs["beta"] = self.cfg.orpo_alpha
+
+        if self.cfg.rpo_alpha is not None:
+            training_args_kwargs["rpo_alpha"] = self.cfg.rpo_alpha
+
+        if self.cfg.use_wandb:
+            training_args_kwargs["run_name"] = self.cfg.wandb_name
+
+        training_args_cls = None
+        blocklist_args_kwargs = []
+        if self.cfg.rl is RLType.SIMPO:
+            training_args_cls = AxolotlCPOConfig
+            training_args_kwargs["loss_type"] = "simpo"
+            training_args_kwargs["simpo_gamma"] = self.cfg.simpo_gamma
+            if self.cfg.cpo_alpha is not None:
+                training_args_kwargs["cpo_alpha"] = self.cfg.cpo_alpha
+
+        elif self.cfg.rl is RLType.ORPO:
+            training_args_cls = AxolotlORPOConfig
+            if self.cfg.max_prompt_len:
+                training_args_kwargs["max_prompt_length"] = self.cfg.max_prompt_len
+
+        elif self.cfg.rl is RLType.KTO:
+            training_args_cls = AxolotlKTOConfig
+
+            training_args_kwargs["desirable_weight"] = (
+                self.cfg.kto_desirable_weight or 1.0
+            )
+            training_args_kwargs["undesirable_weight"] = (
+                self.cfg.kto_undesirable_weight or 1.0
+            )
+
+            if self.cfg.max_prompt_len:
+                training_args_kwargs["max_prompt_length"] = self.cfg.max_prompt_len
+
+        elif self.cfg.rl is RLType.GRPO:
+            training_args_cls = GRPOStrategy.get_training_args_class()
+            training_args_kwargs.update(GRPOStrategy.set_training_args_kwargs(self.cfg))
+            blocklist_args_kwargs = GRPOStrategy.get_blocklist_args_kwargs()
+
+        elif self.cfg.rl in [RLType.DPO, RLType.IPO]:
+            training_args_cls = AxolotlDPOConfig
+            training_args_kwargs.update(DPOStrategy.set_training_args_kwargs(self.cfg))
+        else:
+            raise ValueError(f"Unsupported RL: {self.cfg.rl}")
+
+        for blocklist_key in blocklist_args_kwargs:
+            if blocklist_key in training_args_kwargs:
+                del training_args_kwargs[blocklist_key]
+
+        if self.cfg.plugins:
+            plugin_manager = PluginManager.get_instance()
+            plugin_training_args = plugin_manager.get_training_args(self.cfg)
+            if plugin_training_args:
+                training_args_kwargs.update(plugin_training_args)
+
+        training_args = training_args_cls(  # pylint: disable=unexpected-keyword-arg
+            logging_first_step=True,
+            **training_args_kwargs,
+        )
+
+        # unset run_name so wandb sets up experiment names
+        if self.cfg.use_wandb and training_args.run_name == training_args.output_dir:
+            training_args.run_name = (  # pylint: disable=attribute-defined-outside-init
+                None
+            )
+
+        return training_args, trainer_kwargs
+
+    def build(self, total_num_steps):
+        training_args, trainer_kwargs = self._build_training_arguments(total_num_steps)
+
+        if self.eval_dataset:
+            trainer_kwargs["eval_dataset"] = self.eval_dataset
+        if self.cfg.adapter and self.peft_config and self.cfg.rl is not RLType.GRPO:
+            trainer_kwargs["peft_config"] = self.peft_config
+        if self.cfg.precompute_ref_log_probs is not None:
+            trainer_kwargs["precompute_ref_log_probs"] = (
+                self.cfg.precompute_ref_log_probs
+            )
+
+        trainer_cls, trainer_cls_args = self._get_trainer_cls(trainer_kwargs)
+
+        sig = inspect.signature(trainer_cls)
+        if "tokenizer" in sig.parameters:
+            trainer_kwargs["tokenizer"] = self.tokenizer
+        else:
+            trainer_kwargs["processing_class"] = self.tokenizer
+
+        if self.cfg.datasets is not None and (
+            trainer_cls is DPOStrategy.get_trainer_class()
+        ):
+            trainer_kwargs["dataset_tags"] = [
+                d["path"] for d in self.cfg.datasets if not Path(d["path"]).is_dir()
+            ]
+
+        trainer_kwargs, trainer_cls = self.hook_pre_create_trainer(
+            trainer_kwargs, trainer_cls
+        )
+
+        trainer = trainer_cls(
+            *trainer_cls_args,
+            args=training_args,
+            train_dataset=self.train_dataset,
+            callbacks=self.get_callbacks(),
+            **trainer_kwargs,
+        )
+        if self.cfg.fsdp:
+            ensure_dtype(trainer.model, dtype=self.cfg.torch_dtype)
+            if self.cfg.rl in [RLType.DPO, RLType.IPO] and trainer.ref_model:
+                ensure_dtype(trainer.ref_model, dtype=self.cfg.torch_dtype)
+
+        trainer = self.hook_post_create_trainer(trainer)
+        for callback in self.get_post_trainer_create_callbacks(trainer):
+            trainer.add_callback(callback)
+
+        return trainer
+
+
+class HFPPOTrainerBuilder(TrainerBuilderBase):
+    """
+    HF Factory class for PPO Trainer
+    """
+
+    def get_callbacks(self):
+        callbacks = super().get_callbacks()
+        return callbacks
+
+    def get_post_trainer_create_callbacks(self, trainer):
+        callbacks = super().get_post_trainer_create_callbacks(trainer=trainer)
+        return callbacks
+
+    def build(self, total_num_steps):
+        # TODO: build PPOConfig
+        raise NotImplementedError("PPO trainer builder is not implemented yet.")

src/axolotl/core/chat/messages.py

@@ -156,7 +156,6 @@ class Messages(BaseModel):
                         len(input_ids) : len(input_ids) + len(pending_input_ids)
                     ]
                     if new_pending_inputs != pending_input_ids:
-                        # logging.warning("tokenization mismatch from concatenation.")
                         pending_input_ids = new_pending_inputs
                     input_ids.extend(pending_input_ids)
                     if pending_weight: