add smoke test training

* fix: load_in_*bit not properly read

* fix: load_*bit check

* fix: typo

* refactor: load * bit handling

* feat: add test dpo lora multi-gpu

* fix: turn off sample packing for dpo

* fix: missing warmup_steps

* fix: test to load in 8bit for lora

* skip 8bit lora on h100, add 4bit lora on h100 to multi gpu tests

* chore: reduce max_steps

---------

Co-authored-by: Wing Lian <wing.lian@gmail.com>

.github/CONTRIBUTING.md

@@ -21,12 +21,12 @@ All contributors are expected to adhere to our [Code of Conduct](CODE_OF_CONDUCT
 
 ## Getting Started
 
-Bugs? Please check for open issue else create a new [Issue](https://github.com/OpenAccess-AI-Collective/axolotl/issues/new).
+Bugs? Please check for open issue else create a new [Issue](https://github.com/axolotl-ai-cloud/axolotl/issues/new).
 
 PRs are **greatly welcome**!
 
 1. Fork the repository and clone it to your local machine.
-2. Set up the development environment by following the instructions in the [README.md](https://github.com/OpenAccess-AI-Collective/axolotl/tree/main/README.md) file.
+2. Set up the development environment by following the instructions in the [README.md](https://github.com/axolotl-ai-cloud/axolotl/tree/main/README.md) file.
 3. Explore the codebase, run tests, and verify that everything works as expected.
 
 Please run below to setup env
@@ -42,11 +42,11 @@ pytest tests/
 
 ### Reporting Bugs
 
-If you encounter a bug or issue while using axolotl, please open a new issue on the [GitHub Issues](https://github.com/OpenAccess-AI-Collective/axolotl/issues) page. Provide a clear and concise description of the problem, steps to reproduce it, and any relevant error messages or logs.
+If you encounter a bug or issue while using axolotl, please open a new issue on the [GitHub Issues](https://github.com/axolotl-ai-cloud/axolotl/issues) page. Provide a clear and concise description of the problem, steps to reproduce it, and any relevant error messages or logs.
 
 ### Suggesting Enhancements
 
-We welcome ideas for improvements and new features. To suggest an enhancement, open a new issue on the [GitHub Issues](https://github.com/OpenAccess-AI-Collective/axolotl/issues) page. Describe the enhancement in detail, explain the use case, and outline the benefits it would bring to the project.
+We welcome ideas for improvements and new features. To suggest an enhancement, open a new issue on the [GitHub Issues](https://github.com/axolotl-ai-cloud/axolotl/issues) page. Describe the enhancement in detail, explain the use case, and outline the benefits it would bring to the project.
 
 ### Submitting Pull Requests
 

.github/ISSUE_TEMPLATE/bug-report.yaml

@@ -15,7 +15,7 @@ body:
       label: "Please check that this issue hasn't been reported before."
       description: "The **Label filters** may help make your search more focussed."
       options:
-        - label: "I searched previous [Bug Reports](https://github.com/OpenAccess-AI-Collective/axolotl/labels/bug) didn't find any similar reports."
+        - label: "I searched previous [Bug Reports](https://github.com/axolotl-ai-cloud/axolotl/labels/bug) didn't find any similar reports."
           required: true
 
   - type: textarea

.github/ISSUE_TEMPLATE/config.yml

@@ -1,7 +1,7 @@
 blank_issues_enabled: false
 contact_links:
   - name: Ask a question
-    url: https://github.com/OpenAccess-AI-Collective/axolotl/discussions/categories/q-a
+    url: https://github.com/axolotl-ai-cloud/axolotl/discussions/categories/q-a
     about: Ask questions and discuss with other community members
   - name: Discuss the Project in Discord
     url: https://discord.gg/HhrNrHJPRb

.github/ISSUE_TEMPLATE/docs.yml

@@ -10,7 +10,7 @@ body:
       value: |
         * Ask questions in [Discord](https://discord.gg/HhrNrHJPRb).
         * Before you file an issue read the [Contributing guide](./CONTRIBUTING.md).
-        * Check to make sure someone hasn't already opened a [similar issue](https://github.com/OpenAccess-AI-Collective/axolotl/issues).
+        * Check to make sure someone hasn't already opened a [similar issue](https://github.com/axolotl-ai-cloud/axolotl/issues).
   - type: textarea
     attributes:
       label: What piece of documentation is affected?

.github/ISSUE_TEMPLATE/feature-request.yaml

@@ -8,9 +8,9 @@ body:
       label: "⚠️ Please check that this feature request hasn't been suggested before."
       description: "There are two locations for previous feature requests. Please search in both. Thank you. The **Label filters** may help make your search more focussed."
       options:
-        - label: "I searched previous [Ideas in Discussions](https://github.com/OpenAccess-AI-Collective/axolotl/discussions/categories/ideas) didn't find any similar feature requests."
+        - label: "I searched previous [Ideas in Discussions](https://github.com/axolotl-ai-cloud/axolotl/discussions/categories/ideas) didn't find any similar feature requests."
           required: true
-        - label: "I searched previous [Issues](https://github.com/OpenAccess-AI-Collective/axolotl/labels/enhancement) didn't find any similar feature requests."
+        - label: "I searched previous [Issues](https://github.com/axolotl-ai-cloud/axolotl/labels/enhancement) didn't find any similar feature requests."
           required: true
 
   - type: textarea

.github/workflows/base.yml

@@ -5,32 +5,42 @@ on:
 
 jobs:
   build-base:
-    if: github.repository_owner == 'OpenAccess-AI-Collective'
+    if: github.repository_owner == 'axolotl-ai-cloud'
     # this job needs to be run on self-hosted GPU runners...
     runs-on: axolotl-gpu-runner
     strategy:
       fail-fast: false
       matrix:
         include:
-          - cuda: "118"
-            cuda_version: 11.8.0
+          - cuda: "121"
+            cuda_version: 12.1.1
+            cudnn_version: 8
             python_version: "3.10"
-            pytorch: 2.1.2
+            pytorch: 2.3.1
             torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
           - cuda: "121"
-            cuda_version: 12.1.0
-            python_version: "3.10"
-            pytorch: 2.1.2
-            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
-          - cuda: "121"
-            cuda_version: 12.1.0
+            cuda_version: 12.1.1
+            cudnn_version: 8
             python_version: "3.11"
-            pytorch: 2.1.2
+            pytorch: 2.3.1
             torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
-          - cuda: "121"
-            cuda_version: 12.1.0
+          - cuda: "124"
+            cuda_version: 12.4.1
+            cudnn_version: ""
             python_version: "3.11"
-            pytorch: 2.2.1
+            pytorch: 2.4.1
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+          - cuda: "124"
+            cuda_version: 12.4.1
+            cudnn_version: ""
+            python_version: "3.11"
+            pytorch: 2.4.1
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+          - cuda: "124"
+            cuda_version: 12.4.1
+            cudnn_version: ""
+            python_version: "3.11"
+            pytorch: 2.5.0
             torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
     steps:
       - name: Checkout
@@ -57,6 +67,7 @@ jobs:
           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 }}

.github/workflows/lint.yml

@@ -6,7 +6,8 @@ on:
        - '**.py'
        - 'requirements.txt'
        - '.github/workflows/*.yml'
-       - "*.md"
+       - "*.[q]md"
+       - "examples/**/*.y[a]?ml"
   workflow_dispatch:
 
 jobs:

.github/workflows/main.yml

@@ -8,27 +8,31 @@ on:
 
 jobs:
   build-axolotl:
-    if: ${{ ! contains(github.event.commits[0].message, '[skip docker]]') && github.repository_owner == 'OpenAccess-AI-Collective' }}
+    if: ${{ ! contains(github.event.commits[0].message, '[skip docker]]') && github.repository_owner == 'axolotl-ai-cloud' }}
     strategy:
       fail-fast: false
       matrix:
         include:
-          - cuda: 118
-            cuda_version: 11.8.0
-            python_version: "3.10"
-            pytorch: 2.1.2
-            axolotl_extras:
-            axolotl_args: "--extra-index-url https://download.pytorch.org/whl/cu118"
-            is_latest: true
           - cuda: 121
-            cuda_version: 12.1.0
+            cuda_version: 12.1.1
             python_version: "3.10"
-            pytorch: 2.1.2
-            axolotl_extras:
+            pytorch: 2.3.1
+            axolotl_extras: mamba-ssm
           - cuda: 121
-            cuda_version: 12.1.0
+            cuda_version: 12.1.1
             python_version: "3.11"
-            pytorch: 2.2.1
+            pytorch: 2.3.1
+            axolotl_extras: mamba-ssm
+            is_latest: true
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.4.1
+            axolotl_extras:
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.5.0
             axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:
@@ -60,31 +64,37 @@ jobs:
           push: ${{ github.event_name != 'pull_request' }}
           tags: |
             ${{ steps.metadata.outputs.tags }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
+            ${{ steps.metadata.outputs.tags }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}
             ${{ (matrix.is_latest) && format('{0}-latest', steps.metadata.outputs.tags) || '' }}
           labels: ${{ steps.metadata.outputs.labels }}
 
   build-axolotl-cloud:
     needs: build-axolotl
-    if: ${{ ! contains(github.event.commits[0].message, '[skip docker]]') && github.repository_owner == 'OpenAccess-AI-Collective' }}
+    if: ${{ ! contains(github.event.commits[0].message, '[skip docker]]') && github.repository_owner == 'axolotl-ai-cloud' }}
     # this job needs to be run on self-hosted GPU runners...
     strategy:
       matrix:
         include:
-          - cuda: 118
-            cuda_version: 11.8.0
+          - cuda: 121
+            cuda_version: 12.1.1
             python_version: "3.10"
-            pytorch: 2.1.2
+            pytorch: 2.3.1
+            axolotl_extras:
+          - cuda: 121
+            cuda_version: 12.1.1
+            python_version: "3.11"
+            pytorch: 2.3.1
             axolotl_extras:
             is_latest: true
-          - cuda: 121
-            cuda_version: 12.1.0
-            python_version: "3.10"
-            pytorch: 2.1.2
-            axolotl_extras:
-          - cuda: 121
-            cuda_version: 12.1.0
+          - cuda: 124
+            cuda_version: 12.4.1
             python_version: "3.11"
-            pytorch: 2.2.1
+            pytorch: 2.4.1
+            axolotl_extras:
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.5.0
             axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:
@@ -115,3 +125,45 @@ jobs:
              ${{ steps.metadata.outputs.tags }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
              ${{ (matrix.is_latest) && format('{0}-latest', steps.metadata.outputs.tags) || '' }}
           labels: ${{ steps.metadata.outputs.labels }}
+
+  build-axolotl-cloud-no-tmux:
+    needs: build-axolotl
+    if: ${{ ! contains(github.event.commits[0].message, '[skip docker]]') && github.repository_owner == 'axolotl-ai-cloud' }}
+    # this job needs to be run on self-hosted GPU runners...
+    strategy:
+      matrix:
+        include:
+          - cuda: 121
+            cuda_version: 12.1.1
+            python_version: "3.11"
+            pytorch: 2.3.1
+            axolotl_extras:
+    runs-on: axolotl-gpu-runner
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Docker metadata
+        id: metadata
+        uses: docker/metadata-action@v5
+        with:
+          images: winglian/axolotl-cloud-term
+      - name: Login to Docker Hub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v2
+      - name: Build
+        uses: docker/build-push-action@v5
+        with:
+          context: .
+          build-args: |
+            BASE_TAG=${{ github.ref_name }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
+            CUDA=${{ matrix.cuda }}
+          file: ./docker/Dockerfile-cloud-no-tmux
+          push: ${{ github.event_name != 'pull_request' }}
+          tags: |
+             ${{ steps.metadata.outputs.tags }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
+             ${{ (matrix.is_latest) && format('{0}-latest', steps.metadata.outputs.tags) || '' }}
+          labels: ${{ steps.metadata.outputs.labels }}

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

@@ -0,0 +1,62 @@
+name: docker-multigpu-tests-biweekly
+
+on:
+  pull_request:
+    paths:
+      - 'tests/e2e/multigpu/*.py'
+  workflow_dispatch:
+  schedule:
+    - cron: '0 0 * * 1,4'  # Runs at 00:00 UTC every monday & thursday
+
+jobs:
+  test-axolotl-multigpu:
+    if: ${{ ! contains(github.event.commits[0].message, '[skip docker]]') && github.repository_owner == 'axolotl-ai-cloud' }}
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - cuda: 121
+            cuda_version: 12.1.1
+            python_version: "3.11"
+            pytorch: 2.3.1
+            axolotl_extras:
+            num_gpus: 2
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.4.1
+            axolotl_extras:
+            num_gpus: 2
+            nightly_build: "true"
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.5.0
+            axolotl_extras:
+            num_gpus: 2
+            nightly_build: "true"
+    runs-on: [self-hosted, modal]
+    timeout-minutes: 120
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Install Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.10"
+      - name: Install Modal
+        run: |
+          python -m pip install --upgrade pip
+          pip install modal==0.63.64 jinja2
+      - name: Update env vars
+        run: |
+          echo "BASE_TAG=main-base-py${{ 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 }}" >> $GITHUB_ENV
+          echo "N_GPUS=${{ matrix.num_gpus }}" >> $GITHUB_ENV
+          echo "NIGHTLY_BUILD=${{ matrix.nightly_build }}" >> $GITHUB_ENV
+      - name: Run tests job on Modal
+        run: |
+          modal run cicd.multigpu

.github/workflows/nightlies.yml

@@ -7,27 +7,31 @@ on:
 
 jobs:
   build-axolotl:
-    if: ${{ ! contains(github.event.commits[0].message, '[skip docker]]') && github.repository_owner == 'OpenAccess-AI-Collective' }}
+    if: ${{ ! contains(github.event.commits[0].message, '[skip docker]]') && github.repository_owner == 'axolotl-ai-cloud' }}
     strategy:
       fail-fast: false
       matrix:
         include:
-          - cuda: 118
-            cuda_version: 11.8.0
-            python_version: "3.10"
-            pytorch: 2.1.2
-            axolotl_extras:
-            axolotl_args: "--extra-index-url https://download.pytorch.org/whl/cu118"
-            is_latest: true
           - cuda: 121
-            cuda_version: 12.1.0
+            cuda_version: 12.1.1
             python_version: "3.10"
-            pytorch: 2.1.2
+            pytorch: 2.3.1
             axolotl_extras:
           - cuda: 121
-            cuda_version: 12.1.0
+            cuda_version: 12.1.1
             python_version: "3.11"
-            pytorch: 2.2.1
+            pytorch: 2.3.1
+            axolotl_extras:
+            is_latest: true
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.4.1
+            axolotl_extras:
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.5.0
             axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:
@@ -65,26 +69,31 @@ jobs:
 
   build-axolotl-cloud:
     needs: build-axolotl
-    if: ${{ ! contains(github.event.commits[0].message, '[skip docker]]') && github.repository_owner == 'OpenAccess-AI-Collective' }}
+    if: ${{ ! contains(github.event.commits[0].message, '[skip docker]]') && github.repository_owner == 'axolotl-ai-cloud' }}
     # this job needs to be run on self-hosted GPU runners...
     strategy:
       matrix:
         include:
-          - cuda: 118
-            cuda_version: 11.8.0
+          - cuda: 121
+            cuda_version: 12.1.1
             python_version: "3.10"
-            pytorch: 2.1.2
+            pytorch: 2.3.1
+            axolotl_extras:
+          - cuda: 121
+            cuda_version: 12.1.1
+            python_version: "3.11"
+            pytorch: 2.3.1
             axolotl_extras:
             is_latest: true
-          - cuda: 121
-            cuda_version: 12.1.0
-            python_version: "3.10"
-            pytorch: 2.1.2
-            axolotl_extras:
-          - cuda: 121
-            cuda_version: 12.1.0
+          - cuda: 124
+            cuda_version: 12.4.1
             python_version: "3.11"
-            pytorch: 2.2.1
+            pytorch: 2.4.1
+            axolotl_extras:
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.5.0
             axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:

.github/workflows/pypi.yml

@@ -27,7 +27,7 @@ jobs:
         run: |
           pip3 install wheel packaging
           pip3 install -e .
-          pip3 install -r requirements-tests.txt
+          pip3 install -r requirements-dev.txt -r requirements-tests.txt
 
       - name: Extract tag name
         id: tag

.github/workflows/tests-nightly.yml

@@ -0,0 +1,128 @@
+name: Tests Nightly against upstream main
+on:
+  workflow_dispatch:
+  schedule:
+    - cron: '0 0 * * *'  # Runs at 00:00 UTC every day
+
+jobs:
+  pre-commit:
+    name: pre-commit
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v3
+      - uses: actions/setup-python@v4
+        with:
+          python-version: "3.10"
+          cache: 'pip' # caching pip dependencies
+      - uses: pre-commit/action@v3.0.0
+        env:
+          SKIP: no-commit-to-branch
+
+  pytest:
+    name: PyTest
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python_version: ["3.10", "3.11"]
+        pytorch_version: ["2.3.1", "2.4.1", "2.5.0"]
+    timeout-minutes: 20
+
+    steps:
+      - name: Check out repository code
+        uses: actions/checkout@v3
+
+      - name: Setup Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: ${{ matrix.python_version }}
+          cache: 'pip' # caching pip dependencies
+
+      - name: Install PyTorch
+        run: |
+          pip3 install torch==${{ matrix.pytorch_version }} --index-url https://download.pytorch.org/whl/cpu
+
+      - name: Update requirements.txt
+        run: |
+          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
+
+      - name: Install dependencies
+        run: |
+          pip3 install --upgrade pip
+          pip3 install --upgrade packaging
+          pip3 install -U -e .
+          pip3 install -r requirements-dev.txt -r requirements-tests.txt
+
+      - name: Run tests
+        run: |
+          pytest --ignore=tests/e2e/ tests/
+
+      - name: cleanup pip cache
+        run: |
+          find "$(pip cache dir)/http-v2" -type f -mtime +14 -exec rm {} \;
+
+  docker-e2e-tests:
+    if: github.repository_owner == 'axolotl-ai-cloud'
+    # this job needs to be run on self-hosted GPU runners...
+    runs-on: [self-hosted, modal]
+    timeout-minutes: 60
+    needs: [pre-commit, pytest]
+
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - cuda: 121
+            cuda_version: 12.1.1
+            python_version: "3.10"
+            pytorch: 2.3.1
+            num_gpus: 1
+            axolotl_extras: mamba-ssm
+            nightly_build: "true"
+          - cuda: 121
+            cuda_version: 12.1.1
+            python_version: "3.11"
+            pytorch: 2.3.1
+            num_gpus: 1
+            axolotl_extras: mamba-ssm
+            nightly_build: "true"
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.4.1
+            num_gpus: 1
+            axolotl_extras:
+            nightly_build: "true"
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.5.0
+            num_gpus: 1
+            axolotl_extras:
+            nightly_build: "true"
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Install Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.10"
+      - name: Install Modal
+        run: |
+          python -m pip install --upgrade pip
+          pip install modal==0.63.64 jinja2
+      - name: Update env vars
+        run: |
+          echo "BASE_TAG=main-base-py${{ 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 }}" >> $GITHUB_ENV
+          echo "N_GPUS=${{ matrix.num_gpus }}" >> $GITHUB_ENV
+          echo "NIGHTLY_BUILD=${{ matrix.nightly_build }}" >> $GITHUB_ENV
+      - name: Run tests job on Modal
+        run: |
+          modal run cicd.tests

.github/workflows/tests.yml

@@ -26,6 +26,8 @@ jobs:
           python-version: "3.10"
           cache: 'pip' # caching pip dependencies
       - uses: pre-commit/action@v3.0.0
+        env:
+          SKIP: no-commit-to-branch
 
   pytest:
     name: PyTest
@@ -34,6 +36,7 @@ jobs:
       fail-fast: false
       matrix:
         python_version: ["3.10", "3.11"]
+        pytorch_version: ["2.3.1", "2.4.1", "2.5.0"]
     timeout-minutes: 20
 
     steps:
@@ -46,44 +49,64 @@ jobs:
           python-version: ${{ matrix.python_version }}
           cache: 'pip' # caching pip dependencies
 
-      - name: Install dependencies
+      - name: upgrade pip
         run: |
           pip3 install --upgrade pip
-          pip3 install --upgrade packaging
+          pip3 install --upgrade packaging setuptools wheel
+
+      - name: Install PyTorch
+        run: |
+          pip3 install torch==${{ matrix.pytorch_version }}
+
+      - name: Install dependencies
+        run: |
+          pip3 show torch
           pip3 install -U -e .
-          pip3 install -r requirements-tests.txt
+          pip3 install -r requirements-dev.txt -r requirements-tests.txt
 
       - name: Run tests
         run: |
           pytest --ignore=tests/e2e/ tests/
 
+      - name: cleanup pip cache
+        run: |
+          find "$(pip cache dir)/http-v2" -type f -mtime +14 -exec rm {} \;
+
   docker-e2e-tests:
-    if: github.repository_owner == 'OpenAccess-AI-Collective'
+    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: 60
+    timeout-minutes: 90
     needs: [pre-commit, pytest]
 
     strategy:
       fail-fast: false
       matrix:
         include:
-          - cuda: 118
-            cuda_version: 11.8.0
-            python_version: "3.10"
-            pytorch: 2.1.2
-            axolotl_args: "--extra-index-url https://download.pytorch.org/whl/cu118"
-            num_gpus: 1
           - cuda: 121
-            cuda_version: 12.1.0
+            cuda_version: 12.1.1
             python_version: "3.10"
-            pytorch: 2.1.2
+            pytorch: 2.3.1
             num_gpus: 1
+            axolotl_extras: mamba-ssm
           - cuda: 121
-            cuda_version: 12.1.0
+            cuda_version: 12.1.1
             python_version: "3.11"
-            pytorch: 2.2.1
+            pytorch: 2.3.1
             num_gpus: 1
+            axolotl_extras: mamba-ssm
+          - 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: 2.5.0
+            num_gpus: 1
+            axolotl_extras:
     steps:
       - name: Checkout
         uses: actions/checkout@v4
@@ -94,12 +117,13 @@ jobs:
       - name: Install Modal
         run: |
           python -m pip install --upgrade pip
-          pip install modal jinja2
+          pip install modal==0.63.64 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 }}" >> $GITHUB_ENV
           echo "N_GPUS=${{ matrix.num_gpus }}" >> $GITHUB_ENV
       - name: Run tests job on Modal

.gitignore

@@ -133,6 +133,7 @@ venv/
 ENV/
 env.bak/
 venv.bak/
+venv3.10/
 
 # Spyder project settings
 .spyderproject
@@ -175,3 +176,9 @@ qlora-out/*
 mlruns/*
 
 /.quarto/
+prepared-datasets/
+submit.sh
+*.out*
+
+typings/
+out/

.isort.cfg

@@ -1,3 +1,3 @@
 [settings]
 profile=black
-known_third_party=wandb
+known_third_party=wandb,comet_ml

.mypy.ini

@@ -11,6 +11,9 @@ ignore_errors = True
 [mypy-axolotl.models.mixtral.*]
 ignore_errors = True
 
+[mypy-axolotl.integrations.liger.models.*]
+ignore_errors = True
+
 [mypy-axolotl.models.phi.*]
 ignore_errors = True
 

.pre-commit-config.yaml

@@ -8,6 +8,8 @@ repos:
     -   id: check-yaml
     -   id: end-of-file-fixer
     -   id: trailing-whitespace
+    -   id: no-commit-to-branch
+        args: ['--branch', 'main']
 -   repo: https://github.com/psf/black
     rev: 23.3.0
     hooks:

README.md

@@ -1,5 +1,9 @@
 # Axolotl
 
+![tests](https://github.com/axolotl-ai-cloud/axolotl/actions/workflows/tests.yml/badge.svg)
+![tests-nightly](https://github.com/axolotl-ai-cloud/axolotl/actions/workflows/tests-nightly.yml/badge.svg)
+![multigpu-semi-weekly tests](https://github.com/axolotl-ai-cloud/axolotl/actions/workflows/multi-gpu-e2e.yml/badge.svg)
+
 Axolotl is a tool designed to streamline the fine-tuning of various AI models, offering support for multiple configurations and architectures.
 
 Features:
@@ -7,10 +11,10 @@ Features:
 - 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 xformer, flash attention, rope scaling, and multipacking
+- Integrated with xformer, 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 or mlflow
+- Log results and optionally checkpoints to wandb, mlflow or Comet
 - And more!
 
 <a href="https://www.phorm.ai/query?projectId=e315ba4a-4e14-421f-ab05-38a1f9076f25">
@@ -22,36 +26,50 @@ Features:
 <td>
 
 ## Table of Contents
-- [Introduction](#axolotl)
-- [Supported Features](#axolotl-supports)
-- [Quickstart](#quickstart-)
-- [Environment](#environment)
-  - [Docker](#docker)
-  - [Conda/Pip venv](#condapip-venv)
-  - [Cloud GPU](#cloud-gpu) - Latitude.sh, JarvisLabs, RunPod
-  - [Bare Metal Cloud GPU](#bare-metal-cloud-gpu)
-  - [Windows](#windows)
-  - [Mac](#mac)
-  - [Google Colab](#google-colab)
-  - [Launching on public clouds via SkyPilot](#launching-on-public-clouds-via-skypilot)
-- [Dataset](#dataset)
-- [Config](#config)
-  - [Train](#train)
-  - [Inference](#inference-playground)
-  - [Merge LORA to Base](#merge-lora-to-base)
-  - [Special Tokens](#special-tokens)
-  - [All Config Options](#all-config-options)
-- Advanced Topics
-  - [Multipack](./docs/multipack.qmd)<svg width="24" height="24" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg"><path d="M17 13.5v6H5v-12h6m3-3h6v6m0-6-9 9" class="icon_svg-stroke" stroke="#666" stroke-width="1.5" fill="none" fill-rule="evenodd" stroke-linecap="round" stroke-linejoin="round"></path></svg>
-  - [RLHF & DPO](./docs/rlhf.qmd)<svg width="24" height="24" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg"><path d="M17 13.5v6H5v-12h6m3-3h6v6m0-6-9 9" class="icon_svg-stroke" stroke="#666" stroke-width="1.5" fill="none" fill-rule="evenodd" stroke-linecap="round" stroke-linejoin="round"></path></svg>
-- [Common Errors](#common-errors-)
-  - [Tokenization Mismatch b/w Training & Inference](#tokenization-mismatch-bw-inference--training)
-- [Debugging Axolotl](#debugging-axolotl)
-- [Need Help?](#need-help-)
-- [Badge](#badge-)
-- [Community Showcase](#community-showcase)
-- [Contributing](#contributing-)
-- [Sponsors](#sponsors-)
+- [Axolotl](#axolotl)
+  - [Table of Contents](#table-of-contents)
+  - [Axolotl supports](#axolotl-supports)
+  - [Quickstart ⚡](#quickstart-)
+    - [Usage](#usage)
+  - [Advanced Setup](#advanced-setup)
+    - [Environment](#environment)
+      - [Docker](#docker)
+      - [Conda/Pip venv](#condapip-venv)
+      - [Cloud GPU](#cloud-gpu)
+      - [Bare Metal Cloud GPU](#bare-metal-cloud-gpu)
+        - [LambdaLabs](#lambdalabs)
+        - [GCP](#gcp)
+      - [Windows](#windows)
+      - [Mac](#mac)
+      - [Google Colab](#google-colab)
+      - [Launching on public clouds via SkyPilot](#launching-on-public-clouds-via-skypilot)
+      - [Launching on public clouds via dstack](#launching-on-public-clouds-via-dstack)
+    - [Dataset](#dataset)
+    - [Config](#config)
+      - [All Config Options](#all-config-options)
+    - [Train](#train)
+      - [Preprocess dataset](#preprocess-dataset)
+      - [Multi-GPU](#multi-gpu)
+        - [DeepSpeed](#deepspeed)
+        - [FSDP](#fsdp)
+        - [FSDP + QLoRA](#fsdp--qlora)
+        - [Weights \& Biases Logging](#weights--biases-logging)
+        - [Special Tokens](#special-tokens)
+      - [Liger Kernel](#liger-kernel)
+    - [Inference Playground](#inference-playground)
+    - [Merge LORA to base](#merge-lora-to-base)
+  - [Common Errors 🧰](#common-errors-)
+    - [Tokenization Mismatch b/w Inference \& Training](#tokenization-mismatch-bw-inference--training)
+  - [Debugging Axolotl](#debugging-axolotl)
+  - [Need help? 🙋](#need-help-)
+  - [Badge ❤🏷️](#badge-️)
+  - [Community Showcase](#community-showcase)
+  - [Contributing 🤝](#contributing-)
+  - [Sponsors 🤝❤](#sponsors-)
+      - [💎 Diamond Sponsors - Contact directly](#-diamond-sponsors---contact-directly)
+      - [🥇 Gold Sponsors - $5000/mo](#-gold-sponsors---5000mo)
+      - [🥈 Silver Sponsors - $1000/mo](#-silver-sponsors---1000mo)
+      - [🥉 Bronze Sponsors - $500/mo](#-bronze-sponsors---500mo)
 
 </td>
 <td>
@@ -65,8 +83,8 @@ Features:
     <p>
       Go ahead and Axolotl questions!!
     </p>
-    <img src="https://github.com/OpenAccess-AI-Collective/axolotl/actions/workflows/pre-commit.yml/badge.svg?branch=main" alt="pre-commit">
-    <img alt="PyTest Status" src="https://github.com/OpenAccess-AI-Collective/axolotl/actions/workflows/tests.yml/badge.svg?branch=main">
+    <img src="https://github.com/axolotl-ai-cloud/axolotl/actions/workflows/pre-commit.yml/badge.svg?branch=main" alt="pre-commit">
+    <img alt="PyTest Status" src="https://github.com/axolotl-ai-cloud/axolotl/actions/workflows/tests.yml/badge.svg?branch=main">
   </div>
 </div>
 
@@ -81,6 +99,7 @@ Features:
 | llama       | ✅         | ✅    | ✅     | ✅             | ✅                 | ✅          | ✅            |
 | Mistral     | ✅         | ✅    | ✅     | ✅             | ✅                 | ✅          | ✅            |
 | Mixtral-MoE | ✅         | ✅    | ✅     | ❓             | ❓                 | ❓          | ❓            |
+| Mixtral8X22 | ✅         | ✅    | ✅     | ❓             | ❓                 | ❓          | ❓            |
 | Pythia      | ✅         | ✅    | ✅     | ❌             | ❌                 | ❌          | ❓            |
 | cerebras    | ✅         | ✅    | ✅     | ❌             | ❌                 | ❌          | ❓            |
 | btlm        | ✅         | ✅    | ✅     | ❌             | ❌                 | ❌          | ❓            |
@@ -92,6 +111,7 @@ Features:
 | RWKV        | ✅         | ❓    | ❓     | ❓             | ❓                 | ❓          | ❓            |
 | Qwen        | ✅         | ✅    | ✅     | ❓             | ❓                 | ❓          | ❓            |
 | Gemma       | ✅         | ✅    | ✅     | ❓             | ❓                 | ✅          | ❓            |
+| Jamba       | ✅         | ✅    | ✅     | ❓             | ❓                 | ✅          | ❓            |
 
 ✅: supported
 ❌: not supported
@@ -101,10 +121,10 @@ Features:
 
 Get started with Axolotl in just a few steps! This quickstart guide will walk you through setting up and running a basic fine-tuning task.
 
-**Requirements**: Python >=3.10 and Pytorch >=2.1.1.
+**Requirements**: Nvidia GPU (Ampere architecture or newer for `bf16` and Flash Attention), Python >=3.10 and PyTorch >=2.3.1.
 
 ```bash
-git clone https://github.com/OpenAccess-AI-Collective/axolotl
+git clone https://github.com/axolotl-ai-cloud/axolotl
 cd axolotl
 
 pip3 install packaging ninja
@@ -121,15 +141,15 @@ accelerate launch -m axolotl.cli.train examples/openllama-3b/lora.yml
 
 # inference
 accelerate launch -m axolotl.cli.inference examples/openllama-3b/lora.yml \
-    --lora_model_dir="./lora-out"
+    --lora_model_dir="./outputs/lora-out"
 
 # gradio
 accelerate launch -m axolotl.cli.inference examples/openllama-3b/lora.yml \
-    --lora_model_dir="./lora-out" --gradio
+    --lora_model_dir="./outputs/lora-out" --gradio
 
 # remote yaml files - the yaml config can be hosted on a public URL
 # Note: the yaml config must directly link to the **raw** yaml
-accelerate launch -m axolotl.cli.train https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/examples/openllama-3b/lora.yml
+accelerate launch -m axolotl.cli.train https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/examples/openllama-3b/lora.yml
 ```
 
 ## Advanced Setup
@@ -290,11 +310,47 @@ HF_TOKEN=xx sky launch axolotl.yaml --env HF_TOKEN
 HF_TOKEN=xx BUCKET=<unique-name> sky spot launch axolotl-spot.yaml --env HF_TOKEN --env BUCKET
 ```
 
+#### Launching on public clouds via dstack
+To launch on GPU instance (both on-demand and spot instances) on public clouds (GCP, AWS, Azure, Lambda Labs, TensorDock, Vast.ai, and CUDO), you can use [dstack](https://dstack.ai/).
+
+Write a job description in YAML as below:
+
+```yaml
+# dstack.yaml
+type: task
+
+image: winglian/axolotl-cloud:main-20240429-py3.11-cu121-2.2.2
+
+env:
+  - HUGGING_FACE_HUB_TOKEN
+  - WANDB_API_KEY
+
+commands:
+  - accelerate launch -m axolotl.cli.train config.yaml
+
+ports:
+  - 6006
+
+resources:
+  gpu:
+    memory: 24GB..
+    count: 2
+```
+
+then, simply run the job with `dstack run` command. Append `--spot` option if you want spot instance. `dstack run` command will show you the instance with cheapest price across multi cloud services:
+
+```bash
+pip install dstack
+HUGGING_FACE_HUB_TOKEN=xxx WANDB_API_KEY=xxx dstack run . -f dstack.yaml # --spot
+```
+
+For further and fine-grained use cases, please refer to the official [dstack documents](https://dstack.ai/docs/) and the detailed description of [axolotl example](https://github.com/dstackai/dstack/tree/master/examples/fine-tuning/axolotl) on the official repository.
+
 ### Dataset
 
 Axolotl supports a variety of dataset formats.  It is recommended to use a JSONL.  The schema of the JSONL depends upon the task and the prompt template you wish to use.  Instead of a JSONL, you can also use a HuggingFace dataset with columns for each JSONL field.
 
-See [these docs](https://openaccess-ai-collective.github.io/axolotl/docs/dataset-formats/) for more information on how to use different dataset formats.
+See [the documentation](https://axolotl-ai-cloud.github.io/axolotl/docs/dataset-formats/) for more information on how to use different dataset formats.
 
 ### Config
 
@@ -327,7 +383,7 @@ See [examples](examples) for quick start. It is recommended to duplicate and mod
         - typescript
       type: ... # unimplemented custom format
 
-      # fastchat conversation
+      # fastchat conversation (deprecation soon, use chat_template https://axolotl-ai-cloud.github.io/axolotl/docs/dataset-formats/conversation.html#chat_template)
       # See 'conversation' options: https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py
     - path: ...
       type: sharegpt
@@ -425,7 +481,7 @@ deepspeed: deepspeed_configs/zero1.json
 ```
 
 ```shell
-accelerate launch -m axolotl.cli.train examples/llama-2/config.py --deepspeed deepspeed_configs/zero1.json
+accelerate launch -m axolotl.cli.train examples/llama-2/config.yml --deepspeed deepspeed_configs/zero1.json
 ```
 
 ##### FSDP
@@ -459,6 +515,22 @@ wandb_name:
 wandb_log_model:
 ```
 
+##### Comet Logging
+
+Make sure your `COMET_API_KEY` environment variable is set (recommended) or you login to wandb with `comet login`.
+
+- wandb options
+```yaml
+use_comet:
+comet_api_key:
+comet_workspace:
+comet_project_name:
+comet_experiment_key:
+comet_mode:
+comet_online:
+comet_experiment_config:
+```
+
 ##### Special Tokens
 
 It is important to have special tokens like delimiters, end-of-sequence, beginning-of-sequence in your tokenizer's vocabulary.  This will help you avoid tokenization issues and help your model train better.  You can do this in axolotl like this:
@@ -475,6 +547,25 @@ tokens: # these are delimiters
 
 When you include these tokens in your axolotl config, axolotl adds these tokens to the tokenizer's vocabulary.
 
+##### Liger Kernel
+
+Liger Kernel: Efficient Triton Kernels for LLM Training
+
+https://github.com/linkedin/Liger-Kernel
+
+Liger (LinkedIn GPU Efficient Runtime) Kernel is a collection of Triton kernels designed specifically for LLM training.
+It can effectively increase multi-GPU training throughput by 20% and reduces memory usage by 60%. The Liger Kernel
+composes well and is compatible with both FSDP and Deepspeed.
+
+```yaml
+plugins:
+  - axolotl.integrations.liger.LigerPlugin
+liger_rope: true
+liger_rms_norm: true
+liger_swiglu: true
+liger_fused_linear_cross_entropy: true
+```
+
 ### Inference Playground
 
 Axolotl allows you to load your model in an interactive terminal playground for quick experimentation.
@@ -570,7 +661,7 @@ If you decode a prompt constructed by axolotl, you might see spaces between toke
 3. Make sure the inference string from #2 looks **exactly** like the data you fine tuned on from #1, including spaces and new lines.  If they aren't the same, adjust your inference server accordingly.
 4. As an additional troubleshooting step, you can look at the token ids between 1 and 2 to make sure they are identical.
 
-Having misalignment between your prompts during training and inference can cause models to perform very poorly, so it is worth checking this.  See [this blog post](https://hamel.dev/notes/llm/05_tokenizer_gotchas.html) for a concrete example.
+Having misalignment between your prompts during training and inference can cause models to perform very poorly, so it is worth checking this.  See [this blog post](https://hamel.dev/notes/llm/finetuning/05_tokenizer_gotchas.html) for a concrete example.
 
 ## Debugging Axolotl
 
@@ -587,10 +678,10 @@ Need dedicated support? Please contact us at [✉️wing@openaccessaicollective.
 Building something cool with Axolotl? Consider adding a badge to your model card.
 
 ```markdown
-[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
+[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
 ```
 
-[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)
+[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
 
 ## Community Showcase
 
@@ -608,7 +699,7 @@ PocketDoc Labs
 
 Please read the [contributing guide](./.github/CONTRIBUTING.md)
 
-Bugs? Please check the [open issues](https://github.com/OpenAccess-AI-Collective/axolotl/issues/bug) else create a new Issue.
+Bugs? Please check the [open issues](https://github.com/axolotl-ai-cloud/axolotl/issues/bug) else create a new Issue.
 
 PRs are **greatly welcome**!
 
@@ -626,7 +717,7 @@ pre-commit run --all-files
 
 Thanks to all of our contributors to date. Help drive open source AI progress forward by contributing to Axolotl.
 
-<a href="https://github.com/openaccess-ai-collective/axolotl/graphs/contributors">
+<a href="https://github.com/axolotl-ai-cloud/axolotl/graphs/contributors">
   <img src="https://contrib.rocks/image?repo=openaccess-ai-collective/axolotl" alt="contributor chart by https://contrib.rocks"/>
 </a>
 

_quarto.yml

@@ -14,7 +14,7 @@ website:
     - icon: twitter
       href: https://twitter.com/axolotl_ai
     - icon: github
-      href: https://github.com/OpenAccess-AI-Collective/axolotl/
+      href: https://github.com/axolotl-ai-cloud/axolotl/
     - icon: discord
       href: https://discord.gg/7m9sfhzaf3
 
@@ -36,6 +36,8 @@ website:
             - docs/nccl.qmd
             - docs/mac.qmd
             - docs/multi-node.qmd
+            - docs/unsloth.qmd
+            - docs/amd_hpc.qmd
         - section: "Dataset Formats"
           contents: docs/dataset-formats/*
         - section: "Reference"

cicd/Dockerfile.jinja

@@ -8,13 +8,14 @@ ENV BNB_CUDA_VERSION="{{ CUDA }}"
 ENV PYTORCH_VERSION="{{ PYTORCH_VERSION }}"
 ENV GITHUB_REF="{{ GITHUB_REF }}"
 ENV GITHUB_SHA="{{ GITHUB_SHA }}"
+ENV NIGHTLY_BUILD="{{ NIGHTLY_BUILD }}"
 
 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/OpenAccess-AI-Collective/axolotl.git
+RUN git clone --depth=1 https://github.com/axolotl-ai-cloud/axolotl.git
 
 WORKDIR /workspace/axolotl
 
@@ -22,15 +23,21 @@ RUN git fetch origin +$GITHUB_REF && \
     git checkout FETCH_HEAD
 
 # If AXOLOTL_EXTRAS is set, append it in brackets
-RUN pip install causal_conv1d
+RUN if [ "$NIGHTLY_BUILD" = "true" ] ; then \
+        sed -i 's#^transformers.*#transformers @ git+https://github.com/huggingface/transformers.git@main#' requirements.txt; \
+        sed -i 's#^peft.*#peft @ git+https://github.com/huggingface/peft.git@main#' requirements.txt; \
+        sed -i 's#^accelerate.*#accelerate @ git+https://github.com/huggingface/accelerate.git@main#' requirements.txt; \
+        sed -i 's#^trl.*#trl @ git+https://github.com/huggingface/trl.git@main#' requirements.txt; \
+    fi
+
 RUN if [ "$AXOLOTL_EXTRAS" != "" ] ; then \
-        pip install -e .[deepspeed,flash-attn,mamba-ssm,galore,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
+        pip install -e .[deepspeed,flash-attn,optimizers,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
     else \
-        pip install -e .[deepspeed,flash-attn,mamba-ssm,galore] $AXOLOTL_ARGS; \
+        pip install -e .[deepspeed,flash-attn,optimizers] $AXOLOTL_ARGS; \
     fi
 
 # So we can test the Docker image
-RUN pip install pytest
+RUN 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/*" && \

cicd/cicd.sh

@@ -1,5 +1,6 @@
 #!/bin/bash
+set -e
 
-pytest --ignore=tests/e2e/ /workspace/axolotl/tests/
-pytest /workspace/axolotl/tests/e2e/patched/
-pytest --ignore=tests/e2e/patched/ /workspace/axolotl/tests/e2e/
+pytest -n4 --ignore=tests/e2e/ /workspace/axolotl/tests/
+pytest -n1 --dist loadfile -v /workspace/axolotl/tests/e2e/patched/ /workspace/axolotl/tests/e2e/integrations/
+pytest --ignore=tests/e2e/patched/ --ignore=tests/e2e/multigpu/ --ignore=tests/e2e/integrations/ /workspace/axolotl/tests/e2e/

cicd/multigpu.py

@@ -0,0 +1,77 @@
+"""
+ modal application to run axolotl gpu tests in Modal
+ """
+# pylint: disable=duplicate-code
+
+import os
+import pathlib
+import tempfile
+
+import jinja2
+import modal
+from jinja2 import select_autoescape
+from modal import Image, Stub
+
+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", "2.3.1"),
+    "BASE_TAG": os.environ.get("BASE_TAG", "main-base-py3.11-cu121-2.3.1"),
+    "CUDA": os.environ.get("CUDA", "121"),
+    "GITHUB_REF": os.environ.get("GITHUB_REF", "refs/heads/main"),
+    "GITHUB_SHA": os.environ.get("GITHUB_SHA", ""),
+}
+
+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",
+        force_build=True,
+        gpu="A10G",
+    )
+    .env(df_args)
+    .pip_install("fastapi==0.110.0", "pydantic==2.6.3")
+)
+
+stub = Stub("Axolotl CI/CD", secrets=[])
+
+
+N_GPUS = int(os.environ.get("N_GPUS", 2))
+GPU_CONFIG = modal.gpu.H100(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
+
+
+@stub.function(
+    image=cicd_image,
+    gpu=GPU_CONFIG,
+    timeout=60 * 60,
+    cpu=8.0,
+    memory=131072 * N_GPUS,
+)
+def cicd_pytest():
+    run_cmd("./cicd/multigpu.sh", "/workspace/axolotl")
+
+
+@stub.local_entrypoint()
+def main():
+    cicd_pytest.remote()

cicd/multigpu.sh

@@ -0,0 +1,5 @@
+#!/bin/bash
+set -e
+
+# only run one test at a time so as not to OOM the GPU
+pytest -n1 /workspace/axolotl/tests/e2e/multigpu/

cicd/tests.py

@@ -1,6 +1,8 @@
 """
  modal application to run axolotl gpu tests in Modal
  """
+# pylint: disable=duplicate-code
+
 import os
 import pathlib
 import tempfile
@@ -21,11 +23,12 @@ 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", "2.0.1"),
-    "BASE_TAG": os.environ.get("BASE_TAG", "main-base-py3.10-cu118-2.0.1"),
-    "CUDA": os.environ.get("CUDA", "118"),
+    "PYTORCH_VERSION": os.environ.get("PYTORCH_VERSION", "2.3.1"),
+    "BASE_TAG": os.environ.get("BASE_TAG", "main-base-py3.11-cu121-2.3.1"),
+    "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", ""),
 }
 
 dockerfile_contents = df_template.render(**df_args)
@@ -62,7 +65,7 @@ def run_cmd(cmd: str, run_folder: str):
 @stub.function(
     image=cicd_image,
     gpu=GPU_CONFIG,
-    timeout=45 * 60,
+    timeout=60 * 60,
     cpu=8.0,
     memory=131072,
 )

deepspeed_configs/zero3_bf16.json

@@ -14,15 +14,6 @@
   "bf16": {
     "enabled": true
   },
-  "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",

deepspeed_configs/zero3_bf16_cpuoffload_all.json

@@ -1,4 +1,6 @@
 {
+  "zero_force_ds_cpu_optimizer": false,
+  "zero_allow_untested_optimizer": true,
   "zero_optimization": {
     "stage": 3,
     "offload_optimizer": {
@@ -22,15 +24,6 @@
   "bf16": {
     "enabled": true
   },
-  "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",

deepspeed_configs/zero3_bf16_cpuoffload_params.json

@@ -1,4 +1,6 @@
 {
+  "zero_force_ds_cpu_optimizer": false,
+  "zero_allow_untested_optimizer": true,
   "zero_optimization": {
     "stage": 3,
     "offload_param": {
@@ -18,15 +20,6 @@
   "bf16": {
     "enabled": true
   },
-  "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",

devtools/dev_chat_template.yml

@@ -7,8 +7,8 @@ load_in_8bit: true
 load_in_4bit: false
 
 datasets:
-  - path: philschmid/guanaco-sharegpt-style
-    type: sharegpt
+  - path: fozziethebeat/alpaca_messages_2k_test
+    type: chat_template
     shards: 10
 val_set_size: 0
 output_dir: temp_debug/axolotl_outputs/model

docker/Dockerfile

@@ -11,20 +11,19 @@ ARG PYTORCH_VERSION="2.1.2"
 ENV PYTORCH_VERSION=$PYTORCH_VERSION
 
 RUN apt-get update && \
-    apt-get install -y --allow-change-held-packages vim curl nano libnccl2 libnccl-dev
+    apt-get install -y --allow-change-held-packages vim curl nano libnccl2 libnccl-dev rsync s3fs
 
 WORKDIR /workspace
 
-RUN git clone --depth=1 https://github.com/OpenAccess-AI-Collective/axolotl.git
+RUN git clone --depth=1 https://github.com/axolotl-ai-cloud/axolotl.git
 
 WORKDIR /workspace/axolotl
 
 # If AXOLOTL_EXTRAS is set, append it in brackets
-RUN pip install causal_conv1d
 RUN if [ "$AXOLOTL_EXTRAS" != "" ] ; then \
-        pip install -e .[deepspeed,flash-attn,mamba-ssm,galore,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
+        pip install -e .[deepspeed,flash-attn,optimizers,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
     else \
-        pip install -e .[deepspeed,flash-attn,mamba-ssm,galore] $AXOLOTL_ARGS; \
+        pip install -e .[deepspeed,flash-attn,optimizers] $AXOLOTL_ARGS; \
     fi
 
 # So we can test the Docker image

docker/Dockerfile-base

@@ -3,7 +3,7 @@ ARG CUDNN_VERSION="8"
 ARG UBUNTU_VERSION="22.04"
 ARG MAX_JOBS=4
 
-FROM nvidia/cuda:$CUDA_VERSION-cudnn$CUDNN_VERSION-devel-ubuntu$UBUNTU_VERSION as base-builder
+FROM nvidia/cuda:$CUDA_VERSION-cudnn$CUDNN_VERSION-devel-ubuntu$UBUNTU_VERSION AS base-builder
 
 ENV PATH="/root/miniconda3/bin:${PATH}"
 

docker/Dockerfile-cloud

@@ -3,7 +3,6 @@ FROM winglian/axolotl:$BASE_TAG
 
 ENV HF_DATASETS_CACHE="/workspace/data/huggingface-cache/datasets"
 ENV HUGGINGFACE_HUB_CACHE="/workspace/data/huggingface-cache/hub"
-ENV TRANSFORMERS_CACHE="/workspace/data/huggingface-cache/hub"
 ENV HF_HOME="/workspace/data/huggingface-cache/hub"
 ENV HF_HUB_ENABLE_HF_TRANSFER="1"
 

docker/Dockerfile-cloud-no-tmux

@@ -0,0 +1,26 @@
+ARG BASE_TAG=main
+FROM winglian/axolotl:$BASE_TAG
+
+ENV HF_DATASETS_CACHE="/workspace/data/huggingface-cache/datasets"
+ENV HUGGINGFACE_HUB_CACHE="/workspace/data/huggingface-cache/hub"
+ENV HF_HOME="/workspace/data/huggingface-cache/hub"
+ENV HF_HUB_ENABLE_HF_TRANSFER="1"
+
+EXPOSE 8888
+EXPOSE 22
+
+COPY scripts/cloud-entrypoint-term.sh /root/cloud-entrypoint.sh
+COPY scripts/motd /etc/motd
+
+RUN pip install jupyterlab notebook ipywidgets && \
+    jupyter lab clean
+RUN apt install --yes --no-install-recommends openssh-server tmux sudo && \
+    pip3 install -U --no-cache-dir grpcio ray[default]==2.9.3 && \
+    mkdir -p ~/.ssh && \
+    chmod 700 ~/.ssh && \
+    printf "[ ! -z \"\$TERM\" -a -r /etc/motd ] && cat /etc/motd\n" >> ~/.bashrc && \
+    chmod +x /workspace/axolotl/scripts/cloud-entrypoint.sh && \
+    chmod +x /root/cloud-entrypoint.sh
+
+ENTRYPOINT ["/root/cloud-entrypoint.sh"]
+CMD ["sleep", "infinity"]

docker/Dockerfile-tests

@@ -16,7 +16,7 @@ RUN apt-get update && \
 
 WORKDIR /workspace
 
-RUN git clone --depth=1 https://github.com/OpenAccess-AI-Collective/axolotl.git
+RUN git clone --depth=1 https://github.com/axolotl-ai-cloud/axolotl.git
 
 WORKDIR /workspace/axolotl
 

docs/amd_hpc.qmd

@@ -0,0 +1,108 @@
+---
+title: Training with AMD GPUs on HPC Systems
+description: A comprehensive guide for using Axolotl on distributed systems with AMD GPUs
+---
+
+This guide provides step-by-step instructions for installing and configuring Axolotl on a High-Performance Computing (HPC) environment equipped with AMD GPUs.
+
+## Setup
+
+### 1. Install Python
+
+We recommend using Miniforge, a minimal conda-based Python distribution:
+
+```bash
+curl -L -O "https://github.com/conda-forge/miniforge/releases/latest/download/Miniforge3-$(uname)-$(uname -m).sh"
+bash Miniforge3-$(uname)-$(uname -m).sh
+```
+
+### 2. Configure Python Environment
+Add Python to your PATH and ensure it's available at login:
+
+```bash
+echo 'export PATH=~/miniforge3/bin:$PATH' >> ~/.bashrc
+echo 'if [ -f ~/.bashrc ]; then . ~/.bashrc; fi' >> ~/.bash_profile
+```
+
+### 3. Load AMD GPU Software
+
+Load the ROCm module:
+
+```bash
+module load rocm/5.7.1
+```
+
+Note: The specific module name and version may vary depending on your HPC system. Consult your system documentation for the correct module name.
+
+### 4. Install PyTorch
+
+Install PyTorch with ROCm support:
+
+```bash
+pip install -U torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm5.7 --force-reinstall
+```
+
+### 5. Install Flash Attention
+
+Clone and install the Flash Attention repository:
+
+```bash
+git clone --recursive https://github.com/ROCmSoftwarePlatform/flash-attention.git
+export GPU_ARCHS="gfx90a"
+cd flash-attention
+export PYTHON_SITE_PACKAGES=$(python -c 'import site; print(site.getsitepackages()[0])')
+patch "${PYTHON_SITE_PACKAGES}/torch/utils/hipify/hipify_python.py" hipify_patch.patch
+pip install .
+```
+
+### 6. Install Axolotl
+
+Clone and install Axolotl:
+
+```bash
+git clone https://github.com/axolotl-ai-cloud/axolotl
+cd axolotl
+pip install packaging ninja
+pip install -e .
+```
+
+### 7. Apply xformers Workaround
+
+xformers appears to be incompatible with ROCm. Apply the following workarounds:
+ - Edit $HOME/packages/axolotl/src/axolotl/monkeypatch/llama_attn_hijack_flash.py modifying the code to always return `False` for SwiGLU availability from xformers.
+ - Edit $HOME/miniforge3/lib/python3.10/site-packages/xformers/ops/swiglu_op.py replacing the "SwiGLU" function with a pass statement.
+
+### 8. Prepare Job Submission Script
+
+Create a script for job submission using your HPC's particular software (e.g. Slurm, PBS). Include necessary environment setup and the command to run Axolotl training. If the compute node(s) do(es) not have internet access, it is recommended to include
+
+```bash
+export TRANSFORMERS_OFFLINE=1
+export HF_DATASETS_OFFLINE=1
+```
+
+### 9. Download Base Model
+
+Download a base model using the Hugging Face CLI:
+
+```bash
+huggingface-cli download meta-llama/Meta-Llama-3.1-8B --local-dir ~/hfdata/llama3.1-8B
+```
+
+### 10. Create Axolotl Configuration
+
+Create an Axolotl configuration file (YAML format) tailored to your specific training requirements and dataset. Use FSDP for multi-node training.
+
+Note: Deepspeed did not work at the time of testing. However, if anyone managed to get it working, please let us know.
+
+### 11. Preprocess Data
+
+Run preprocessing on the login node:
+
+```bash
+CUDA_VISIBLE_DEVICES="" python -m axolotl.cli.preprocess /path/to/your/config.yaml
+```
+
+### 12. Train
+
+You are now ready to submit your previously prepared job script. 🚂

docs/config.qmd

@@ -83,13 +83,14 @@ lora_on_cpu: true
 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, sharegpt, gpteacher, oasst, reflection]
+    # The type of prompt to use for training. [alpaca, sharegpt, 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] number of shards to split data into
     name: # Optional[str] name of dataset configuration to load
     train_on_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.
 
     # Optional[str] fastchat conversation type, only used with type: sharegpt
     conversation: # Options (see Conversation 'name'): https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py
@@ -123,6 +124,48 @@ datasets:
       # 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 template for chat template. This will be only used if `chat_template` is set to `jinja` or empty (in which case chat_template is automatically set to `jinja`).
+    chat_template_jinja:
+    # The key in the data example that contains the messages. Default is "messages".
+    field_messages: messages
+    # The key in the message turn that contains the role. Default is "role".
+    message_field_role: role
+    # The key in the message turn that contains the content. Default is "content".
+    message_field_content: content
+    # Optional[Dict[str, List]]. Roles mapping for the messages.
+    roles:
+      user: ["human", "user"]
+      assistant: ["gpt", "assistant", "ai"]
+      system: ["system"]
+
+    ## NOTE: Leaving the below empty will default to using the simple legacy tokenization strategy where only last message is trained on.
+
+    # Optional[List[str]]. Roles to train on. The tokens from these roles will be considered for the loss.
+    roles_to_train: ["gpt", "assistant"]
+    # Optional[str]. Which EOS tokens to train on in the conversation. Possible values are:
+    # - all: train on all EOS tokens
+    # - turn: train on the EOS token at the end of each trainable turn
+    # - last: train on the last EOS token in the conversation
+    train_on_eos: last
+    # 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).
+    # See example at `docs/dataset-formats/conversation.qmd`
+    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
@@ -138,12 +181,19 @@ test_datasets:
     data_files:
       - /workspace/data/eval.jsonl
 
-# use RL training: 'dpo', 'ipo', 'kto_pair'
+# use RL training: 'dpo', 'ipo', 'kto'
 rl:
 
-# Saves the desired chat template to the tokenizer_config.json for easier inferencing
-# Currently supports chatml and inst (mistral/mixtral)
-chat_template: chatml
+# 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
 # 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
@@ -186,6 +236,11 @@ eval_sample_packing:
 # 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
 
 # Passed through to transformers when loading the model when launched without accelerate
 # Use `sequential` when training w/ model parallelism to limit memory
@@ -227,6 +282,12 @@ lora_modules_to_save:
 
 lora_fan_in_fan_out: false
 
+# 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
@@ -254,8 +315,21 @@ wandb_log_model: # "checkpoint" to log model to wandb Artifacts every `save_step
 # 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.
+
 # Where to save the full-finetuned model to
 output_dir: ./completed-model
 
@@ -268,6 +342,7 @@ torch_compile_backend:  # Optional[str]
 # 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
@@ -278,7 +353,7 @@ lr_quadratic_warmup:
 logging_steps:
 eval_steps: # Leave empty to eval at each epoch, integers for every N steps. decimal for fraction of total steps
 evals_per_epoch: # number of times per epoch to run evals, mutually exclusive with eval_steps
-save_strategy: # Set to `no` to skip checkpoint saves
+save_strategy: # Set to `"no"` to skip checkpoint saves
 save_steps: # Leave empty to save at each epoch
 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
@@ -289,7 +364,7 @@ max_steps:
 
 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
-eval_causal_lm_metrics: # HF evaluate metrics used during evaluation. Default is ["sacrebleu", "comet", "ter", chrf]
+eval_causal_lm_metrics: # HF evaluate metrics used during evaluation. Default is ["sacrebleu", "comet", "ter", "chrf", "perplexity"]
 
 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)
@@ -412,6 +487,7 @@ special_tokens:
   # bos_token: "<s>"
   # eos_token: "</s>"
   # unk_token: "<unk>"
+  # pad_token: "[PAD]"
 
 # Add extra tokens.
 tokens:

docs/dataset-formats/conversation.qmd

@@ -6,6 +6,8 @@ order: 3
 
 ## sharegpt
 
+UPDATE: ShareGPT is being deprecated in the next release. Please see `chat_template` section below.
+
 conversations where `from` is `human`/`gpt`. (optional: first row with role `system` to override default system prompt)
 
 ```{.json filename="data.jsonl"}
@@ -54,6 +56,14 @@ conversations where `from` is `prompter` `assistant` instead of default sharegpt
 {"conversations": [{"from": "...", "value": "..."}]}
 ```
 
+## sharegpt.load_ultrachat
+
+conversations where the turns field is 'messages', human is 'user' and gpt is 'assistant'.
+
+```{.json filename="data.jsonl"}
+{"messages": [{"user": "...", "assistant": "..."}]}
+```
+
 ## sharegpt_jokes
 
 creates a chat where bot is asked to tell a joke, then explain why the joke is funny
@@ -61,3 +71,138 @@ creates a chat where bot is asked to tell a joke, then explain why the joke is f
 ```{.json filename="data.jsonl"}
 {"conversations": [{"title": "...", "text": "...", "explanation": "..."}]}
 ```
+
+
+## chat_template
+
+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": "..."}]}
+```
+
+See `config.qmd` for full configs and supported templates.
+
+### Migrating from sharegpt
+
+Most configs can be adapted as follows:
+
+```yaml
+# old
+chat_template: chatml
+datasets:
+  - path: ...
+    type: sharegpt
+    conversation: chatml
+
+# new (if using tokenizer's chat_template)
+datasets:
+  - path: ...
+    type: chat_template
+
+    field_messages: conversations
+    message_field_role: from
+    message_field_content: value
+
+# new (if setting a new chat_template like chatml, gemma, etc)
+chat_template: chatml
+datasets:
+  - path: ...
+    type: chat_template
+
+    field_messages: conversations
+    message_field_role: from
+    message_field_content: value
+```
+
+We recommend checking the below examples for other usecases.
+
+### Examples
+
+1. Using the default chat template in the tokenizer_config.json on OpenAI messages format, training on only last message.
+
+```yaml
+datasets:
+  - path: ...
+    type: 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.
+
+```yaml
+chat_template: gemma # this overwrites the tokenizer's chat_template
+datasets:
+  - path: ...
+    type: chat_template
+    roles_to_train: ["assistant"]
+```
+
+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.
+
+```yaml
+chat_template: tokenizer_default_fallback_chatml # this overwrites the tokenizer's chat_template
+datasets:
+  - path: ...
+    type: chat_template
+    roles_to_train: ["assistant"]
+```
+
+4. 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
+chat_template_jinja: "{{ bos_token }}{% for message in messages %}{% if (message['role'] == 'system') %}{{'<|system|>' + '\n' + message['content'] + '<|end|>' + '\n'}}{% elif (message['role'] == 'user') %}{{'<|user|>' + '\n' + message['content'] + '<|end|>' + '\n' + '<|assistant|>' + '\n'}}{% elif message['role'] == 'assistant' %}{{message['content'] + '<|end|>' + '\n'}}{% endif %}{% endfor %}"
+
+datasets:
+  - path: ...
+    type: chat_template
+    roles_to_train: ["assistant"]
+```
+
+5. (Advanced) Using fine-grained control over tokens and turns to train in a conversation
+
+For a data sample that looks like:
+
+```{.json filename="data.jsonl"}
+{
+  "conversations": [
+    {"from": "system", "value": "You are an AI assistant.", "train": false},
+    {"from": "human", "value": "Hello", "train": false},
+    {"from": "assistant", "value": "Hello", "train": true},
+    {"from": "human", "value": "How are you?", "train": true},
+    {
+      "from": "assistant",
+      "value": "I'm doing very well, thank you!",
+      "train_detail": [
+        {"begin_offset": 0, "end_offset": 8, "train": false},
+        {"begin_offset": 9, "end_offset": 18, "train": true},
+        {"begin_offset": 19, "end_offset": 30, "train": false},
+      ],
+    },
+    {
+        "from": "human",
+        "value": "I'm doing very well, thank you!",
+        "train": true,
+    },
+    {"from": "assistant", "value": "Hi there!", "train": true}
+  ]
+}
+```
+
+The configuration would look like:
+
+```yaml
+datasets:
+  - path: ...
+    type: chat_template
+    chat_template: tokenizer_default
+    field_messages: conversations
+    message_field_role: from
+    message_field_content: value
+    roles_to_train: []
+    train_on_eos: turn
+    message_field_training: train
+    message_field_training_detail: train_detail
+```
+
+Tip: It is not necessary to use both `message_field_training` and `message_field_training_detail` at a time.

docs/dataset-formats/tokenized.qmd

@@ -4,9 +4,25 @@ description: How to use a custom pre-tokenized dataset.
 order: 5
 ---
 
-- Do not pass a `type:` in your axolotl config.
+- Pass an empty `type:` in your axolotl config.
 - Columns in Dataset must be exactly `input_ids`, `attention_mask`, `labels`
+- To indicate that a token should be ignored during training, set its corresponding label to `-100`.
+- You must add BOS and EOS, and make sure that you are training on EOS by not setting its label to -100.
+- For pretraining, do not truncate/pad documents to the context window length.
+- For instruction training, documents must be truncated/padded as desired.
+
+Sample config:
 
 ```{.yaml filename="config.yml"}
-- path: ...
+datasets:
+  - path: /path/to/your/file.jsonl
+    ds_type: json
+    type:
+```
+
+Sample jsonl:
+
+```jsonl
+{"input_ids":[271,299,99],"attention_mask":[1,1,1],"labels":[271,-100,99]}
+{"input_ids":[87,227,8383,12],"attention_mask":[1,1,1,1],"labels":[87,227,8383,12]}
 ```

docs/dataset_preprocessing.qmd

@@ -0,0 +1,35 @@
+---
+title: Dataset Preprocessing
+description: How datasets are processed
+---
+
+Dataset pre-processing is the step where Axolotl takes each dataset you've configured alongside
+the (dataset format)[../dataset-formats/] and prompt strategies to:
+ - parse the dataset based on the *dataset format*
+ - transform the dataset to how you would interact with the model based on the *prompt strategy*
+ - tokenize the dataset based on the configured model & tokenizer
+ - shuffle and merge multiple datasets together if using more than one
+
+The processing of the datasets can happen one of two ways:
+
+1. Before kicking off training by calling `python -m axolotl.cli.preprocess /path/to/your.yaml --debug`
+2. When training is started
+
+What are the benefits of pre-processing? When training interactively or for sweeps
+(e.g. you are restarting the trainer often), processing the datasets can oftentimes be frustratingly
+slow. Pre-processing will cache the tokenized/formatted datasets according to a hash of dependent
+training parameters so that it will intelligently pull from its cache when possible.
+
+The path of the cache is controlled by `dataset_prepared_path:` and is often left blank in example
+YAMLs as this leads to a more robust solution that prevents unexpectedly reusing cached data.
+
+If `dataset_prepared_path:` is left empty, when training, the processed dataset will be cached in a
+default path of `./last_run_prepared/`, but will ignore anything already cached there. By explicitly
+setting `dataset_prepared_path: ./last_run_prepared`, the trainer will use whatever pre-processed
+data is in the cache.
+
+What are the edge cases? Let's say you are writing a custom prompt strategy or using a user-defined
+prompt template. Because the trainer cannot readily detect these changes, we cannot change the
+calculated hash value for the pre-processed dataset. If you have `dataset_prepared_path: ...` set
+and change your prompt templating logic, it may not pick up the changes you made and you will be
+training over the old prompt.

docs/debugging.qmd

@@ -51,12 +51,12 @@ While debugging it's helpful to simplify your test scenario as much as possible.
 
 ### Background
 
-The below example shows how to configure VSCode to debug data preprocessing of the `sharegpt` format.  This is the format used when you have the following in your axolotl config:
+The below example shows how to configure VSCode to debug data preprocessing of the `chat_template` format.  This is the format used when you have the following in your axolotl config:
 
 ```yaml
 datasets:
-  - path: <path to your sharegpt formatted dataset> # example on HF Hub: philschmid/guanaco-sharegpt-style
-    type: sharegpt
+  - path: <path to your chat_template formatted dataset> # example on HF Hub: fozziethebeat/alpaca_messages_2k_test
+    type: chat_template
 ```
 
 >[!Important]
@@ -83,7 +83,7 @@ If you developing on a remote host, you can easily use VSCode to debug remotely.
 
 The easiest way to get started is to modify the [.vscode/launch.json](../.vscode/launch.json) file in this project.  This is just an example configuration, so you may need to modify or copy it to suit your needs.
 
-For example, to mimic the command `cd devtools && CUDA_VISIBLE_DEVICES=0 accelerate launch -m axolotl.cli.train dev_sharegpt.yml`, you would use the below configuration[^1].  Note that we add additional flags that override the axolotl config and incorporate the tips above (see the comments). We also set the working directory to `devtools` and set the `env` variable `HF_HOME` to a temporary folder that is later partially deleted.  This is because we want to delete the HF dataset cache before each run in order to ensure that the data preprocessing code is run from scratch.
+For example, to mimic the command `cd devtools && CUDA_VISIBLE_DEVICES=0 accelerate launch -m axolotl.cli.train dev_chat_template.yml`, you would use the below configuration[^1].  Note that we add additional flags that override the axolotl config and incorporate the tips above (see the comments). We also set the working directory to `devtools` and set the `env` variable `HF_HOME` to a temporary folder that is later partially deleted.  This is because we want to delete the HF dataset cache before each run in order to ensure that the data preprocessing code is run from scratch.
 
 ```jsonc
 // .vscode/launch.json
@@ -91,12 +91,12 @@ For example, to mimic the command `cd devtools && CUDA_VISIBLE_DEVICES=0 acceler
     "version": "0.2.0",
     "configurations": [
         {
-            "name": "Debug axolotl prompt - sharegpt",
+            "name": "Debug axolotl prompt - chat_template",
             "type": "python",
             "module": "accelerate.commands.launch",
             "request": "launch",
             "args": [
-                "-m", "axolotl.cli.train", "dev_sharegpt.yml",
+                "-m", "axolotl.cli.train", "dev_chat_template.yml",
                 // The flags below simplify debugging by overriding the axolotl config
                 // with the debugging tips above.  Modify as needed.
                 "--dataset_processes=1",      // limits data preprocessing to one process
@@ -192,7 +192,7 @@ Using [official Axolotl Docker images](https://hub.docker.com/r/winglian/axolotl
 On the host that is running axolotl (ex: if you are using a remote host), clone the axolotl repo and change your current directory to the root:
 
 ```bash
-git clone https://github.com/OpenAccess-AI-Collective/axolotl
+git clone https://github.com/axolotl-ai-cloud/axolotl
 cd axolotl
 ```
 
@@ -240,6 +240,6 @@ style="border-radius: 10px; display: block; margin: auto;" width="560" height="3
 </div>
 <br>
 
-[^1]: The config actually mimics the command `CUDA_VISIBLE_DEVICES=0 python -m accelerate.commands.launch -m axolotl.cli.train devtools/sharegpt.yml`, but this is the same thing.
+[^1]: The config actually mimics the command `CUDA_VISIBLE_DEVICES=0 python -m accelerate.commands.launch -m axolotl.cli.train devtools/chat_template.yml`, but this is the same thing.
 
 [^2]: Many of the below flags are recommended best practices by Nvidia when using nvidia-container-toolkit.  You can read more about these flags [here](https://docs.nvidia.com/deeplearning/frameworks/user-guide/index.html).

docs/fsdp_qlora.qmd

@@ -20,7 +20,7 @@ To enable `QLoRA` with `FSDP`, you need to perform the following steps:
 > See the [example config](#example-config) file in addition to reading these instructions.
 
 1. Set `adapter: qlora` in your axolotl config file.
-2. Enable FSDP in your axolotl config, as [described here](https://github.com/OpenAccess-AI-Collective/axolotl?tab=readme-ov-file#fsdp).
+2. Enable FSDP in your axolotl config, as [described here](https://github.com/axolotl-ai-cloud/axolotl?tab=readme-ov-file#fsdp).
 3. Use one of the supported model types: `llama`, `mistral` or `mixtral`.
 
 ## Example Config
@@ -29,7 +29,7 @@ To enable `QLoRA` with `FSDP`, you need to perform the following steps:
 
 ## References
 
-- [PR #1378](https://github.com/OpenAccess-AI-Collective/axolotl/pull/1378) enabling QLoRA in FSDP in Axolotl.
+- [PR #1378](https://github.com/axolotl-ai-cloud/axolotl/pull/1378) enabling QLoRA in FSDP in Axolotl.
 - [Blog Post](https://www.answer.ai/posts/2024-03-06-fsdp-qlora.html) from the [Answer.AI](https://www.answer.ai/) team describing the work that enabled QLoRA in FSDP.
 - Related HuggingFace PRs Enabling FDSP + QLoRA:
     - Accelerate [PR#2544](https://github.com/huggingface/accelerate/pull/2544 )

docs/input_output.qmd

@@ -25,7 +25,7 @@ description: "Template-free prompt construction with the `input_output` format"
 ### Masking Inputs
 
 One of the most popular features of
-[axolotl](https://github.com/OpenAccess-AI-Collective/axolotl) is
+[axolotl](https://github.com/axolotl-ai-cloud/axolotl) is
 setting the following configuration value:
 
 
@@ -33,7 +33,7 @@ setting the following configuration value:
 train_on_inputs: false
 ```
 
-If you declare a [dataset formats](https://github.com/OpenAccess-AI-Collective/axolotl?tab=readme-ov-file#dataset)
+If you declare a [dataset formats](https://github.com/axolotl-ai-cloud/axolotl?tab=readme-ov-file#dataset)
 such as `alpaca` or `chatml`, axolotl knows what is an input
 (i.e. human) vs. an output (i.e. the assistant) and masks the input
 labels so that your model can focus on predicting the outputs only.
@@ -205,7 +205,7 @@ ds = load_from_disk(f'last_run_prepared/{directory[0]}/')
     hi there!.  goodbye  farewell</s>
 ```
 
-We can check that the right tokens are ingored by comparing the labels
+We can check that the right tokens are ignored by comparing the labels
 to each token:
 
 ```python

docs/multimodal.qmd

@@ -0,0 +1,28 @@
+# MultiModal / Vision Language Models (BETA)
+
+### Supported Models
+
+- Mllama, i.e. llama with vision models
+
+### Usage
+
+Currently multimodal support is limited and doesn't have full feature parity. To finetune a multimodal Llama w/ LoRA,
+you'll need to use the following in YAML in combination with the rest of the required hyperparams.
+
+```yaml
+base_model: alpindale/Llama-3.2-11B-Vision-Instruct
+processor_type: AutoProcessor
+skip_prepare_dataset: true
+
+chat_template: llama3_2_vision
+datasets:
+  - path: HuggingFaceH4/llava-instruct-mix-vsft
+    type: chat_template
+    split: train[:1%]
+    field_messages: messages
+remove_unused_columns: false
+sample_packing: false
+
+# only finetune the Language model, leave the vision model and vision tower frozen
+lora_target_modules: 'language_model.model.layers.[\d]+.(mlp|cross_attn|self_attn).(up|down|gate|q|k|v|o)_proj'
+```

docs/rlhf.qmd

@@ -49,7 +49,7 @@ remove_unused_columns: false
 chat_template: chatml
 datasets:
   - path: argilla/ultrafeedback-binarized-preferences-cleaned
-    type: orpo.chat_template
+    type: chat_template.argilla
 ```
 
 #### Using local dataset files

docs/torchao.qmd

@@ -0,0 +1,19 @@
+---
+title: "PyTorch ao"
+description: "Custom data types and layouts for training and inference"
+---
+
+### Installation
+
+Stable Release from the PyTorch index
+
+```bash
+pip install torchao --extra-index-url https://download.pytorch.org/whl/cu121 # full options are cpu/cu118/cu121/cu124
+```
+
+
+Nightly release
+
+```bash
+pip install --pre torchao-nightly --index-url https://download.pytorch.org/whl/nightly/cu121 # full options are cpu/cu118/cu121/cu124
+```

docs/unsloth.qmd

@@ -0,0 +1,49 @@
+---
+title: "Unsloth"
+description: "Hyper-optimized QLoRA finetuning for single GPUs"
+---
+
+### Overview
+
+Unsloth provides hand-written optimized kernels for LLM finetuning that slightly improve speed and VRAM over
+standard industry baselines.
+
+
+### Installation
+
+The following will install unsloth from source and downgrade xformers as unsloth is incompatible with the most up
+to date libraries.
+
+```bash
+pip install --no-deps "unsloth @ git+https://github.com/unslothai/unsloth.git"
+pip install --no-deps --force-reinstall xformers==0.0.26.post1
+```
+
+### Using unsloth w Axolotl
+
+Axolotl exposes a few configuration options to try out unsloth and get most of the performance gains.
+
+Our unsloth integration is currently limited to the following model architectures:
+ - llama
+
+These options are specific to LoRA finetuning and cannot be used for multi-GPU finetuning
+```yaml
+unsloth_lora_mlp: true
+unsloth_lora_qkv: true
+unsloth_lora_o: true
+```
+
+These options are composable and can be used with multi-gpu finetuning
+```yaml
+unsloth_cross_entropy_loss: true
+unsloth_rms_norm: true
+unsloth_rope: true
+```
+
+### Limitations
+
+- Single GPU only; e.g. no multi-gpu support
+- No deepspeed or FSDP support (requires multi-gpu)
+- LoRA + QLoRA support only. No full fine tunes or fp8 support.
+- Limited model architecture support. Llama, Phi, Gemma, Mistral only
+- No MoE support.

examples/cerebras/btlm-ft.yml

@@ -38,7 +38,7 @@ wandb_watch:
 wandb_name:
 wandb_log_model:
 
-output_dir: btlm-out
+output_dir: ./outputs/btlm-out
 gradient_accumulation_steps: 1
 micro_batch_size: 1
 num_epochs: 1

examples/cerebras/qlora.yml

@@ -25,7 +25,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 batch_size: 4
 micro_batch_size: 4
 num_epochs: 2

examples/code-llama/13b/lora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/code-llama/13b/qlora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 adapter: qlora
 lora_model_dir:

examples/code-llama/34b/lora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/code-llama/34b/qlora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 adapter: qlora
 lora_model_dir:

examples/code-llama/7b/lora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/code-llama/7b/qlora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 adapter: qlora
 lora_model_dir:

examples/colab-notebooks/colab-axolotl-example.ipynb

@@ -1,216 +1,222 @@
 {
-  "cells": [
-    {
-      "cell_type": "markdown",
-      "metadata": {
-        "id": "AKjdG7tbTb-n"
-      },
-      "source": [
-        "# Example notebook for running Axolotl on google colab"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {
-        "id": "RcbNpOgWRcii"
-      },
-      "outputs": [],
-      "source": [
-        "import torch\n",
-        "# Check so there is a gpu available, a T4(free tier) is enough to run this notebook\n",
-        "assert (torch.cuda.is_available()==True)"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {
-        "id": "h3nLav8oTRA5"
-      },
-      "source": [
-        "## Install Axolotl and dependencies"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {
-        "colab": {
-          "base_uri": "https://localhost:8080/"
-        },
-        "id": "3c3yGAwnOIdi",
-        "outputId": "e3777b5a-40ef-424f-e181-62dfecd1dd01"
-      },
-      "outputs": [],
-      "source": [
-        "!pip install torch==\"2.1.2\"\n",
-        "!pip install -e git+https://github.com/OpenAccess-AI-Collective/axolotl#egg=axolotl\n",
-        "!pip install flash-attn==\"2.5.0\"\n",
-        "!pip install deepspeed==\"0.13.1\""
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {
-        "id": "BW2MFr7HTjub"
-      },
-      "source": [
-        "## Create an yaml config file"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {
-        "id": "9pkF2dSoQEUN"
-      },
-      "outputs": [],
-      "source": [
-        "import yaml\n",
-        "\n",
-        "# Your YAML string\n",
-        "yaml_string = \"\"\"\n",
-        "base_model: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T\n",
-        "model_type: LlamaForCausalLM\n",
-        "tokenizer_type: LlamaTokenizer\n",
-        "is_llama_derived_model: true\n",
-        "\n",
-        "load_in_8bit: false\n",
-        "load_in_4bit: true\n",
-        "strict: false\n",
-        "\n",
-        "datasets:\n",
-        "  - path: mhenrichsen/alpaca_2k_test\n",
-        "    type: alpaca\n",
-        "dataset_prepared_path:\n",
-        "val_set_size: 0.05\n",
-        "output_dir: ./qlora-out\n",
-        "\n",
-        "adapter: qlora\n",
-        "lora_model_dir:\n",
-        "\n",
-        "sequence_len: 1096\n",
-        "sample_packing: true\n",
-        "pad_to_sequence_len: true\n",
-        "\n",
-        "lora_r: 32\n",
-        "lora_alpha: 16\n",
-        "lora_dropout: 0.05\n",
-        "lora_target_modules:\n",
-        "lora_target_linear: true\n",
-        "lora_fan_in_fan_out:\n",
-        "\n",
-        "wandb_project:\n",
-        "wandb_entity:\n",
-        "wandb_watch:\n",
-        "wandb_name:\n",
-        "wandb_log_model:\n",
-        "\n",
-        "mlflow_experiment_name: colab-example\n",
-        "\n",
-        "gradient_accumulation_steps: 1\n",
-        "micro_batch_size: 1\n",
-        "num_epochs: 4\n",
-        "max_steps: 20\n",
-        "optimizer: paged_adamw_32bit\n",
-        "lr_scheduler: cosine\n",
-        "learning_rate: 0.0002\n",
-        "\n",
-        "train_on_inputs: false\n",
-        "group_by_length: false\n",
-        "bf16: false\n",
-        "fp16: true\n",
-        "tf32: false\n",
-        "\n",
-        "gradient_checkpointing: true\n",
-        "early_stopping_patience:\n",
-        "resume_from_checkpoint:\n",
-        "local_rank:\n",
-        "logging_steps: 1\n",
-        "xformers_attention:\n",
-        "flash_attention: false\n",
-        "\n",
-        "warmup_steps: 10\n",
-        "evals_per_epoch:\n",
-        "saves_per_epoch:\n",
-        "debug:\n",
-        "deepspeed:\n",
-        "weight_decay: 0.0\n",
-        "fsdp:\n",
-        "fsdp_config:\n",
-        "special_tokens:\n",
-        "\n",
-        "\"\"\"\n",
-        "\n",
-        "# Convert the YAML string to a Python dictionary\n",
-        "yaml_dict = yaml.safe_load(yaml_string)\n",
-        "\n",
-        "# Specify your file path\n",
-        "file_path = 'test_axolotl.yaml'\n",
-        "\n",
-        "# Write the YAML file\n",
-        "with open(file_path, 'w') as file:\n",
-        "    yaml.dump(yaml_dict, file)\n"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {
-        "id": "bidoj8YLTusD"
-      },
-      "source": [
-        "## Launch the training"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {
-        "colab": {
-          "base_uri": "https://localhost:8080/"
-        },
-        "id": "ydTI2Jk2RStU",
-        "outputId": "d6d0df17-4b53-439c-c802-22c0456d301b"
-      },
-      "outputs": [],
-      "source": [
-        "# Buy using the ! the comand will be executed as a bash command\n",
-        "!accelerate launch -m axolotl.cli.train /content/test_axolotl.yaml"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "## Play with inference"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "# Buy using the ! the comand will be executed as a bash command\n",
-        "!accelerate launch -m axolotl.cli.inference /content/test_axolotl.yaml \\\n",
-        "    --qlora_model_dir=\"./qlora-out\" --gradio"
-      ]
-    }
-  ],
-  "metadata": {
-    "accelerator": "GPU",
-    "colab": {
-      "gpuType": "T4",
-      "provenance": []
-    },
-    "kernelspec": {
-      "display_name": "Python 3",
-      "name": "python3"
-    },
-    "language_info": {
-      "name": "python"
-    }
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "id": "AKjdG7tbTb-n"
+   },
+   "source": [
+    "# Example notebook for running Axolotl on google colab"
+   ]
   },
-  "nbformat": 4,
-  "nbformat_minor": 0
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "id": "RcbNpOgWRcii"
+   },
+   "outputs": [],
+   "source": [
+    "import torch\n",
+    "# Check so there is a gpu available, a T4(free tier) is enough to run this notebook\n",
+    "assert (torch.cuda.is_available()==True)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "id": "h3nLav8oTRA5"
+   },
+   "source": [
+    "## Install Axolotl and dependencies"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "colab": {
+     "base_uri": "https://localhost:8080/"
+    },
+    "id": "3c3yGAwnOIdi",
+    "outputId": "e3777b5a-40ef-424f-e181-62dfecd1dd01"
+   },
+   "outputs": [],
+   "source": [
+    "!pip install -e git+https://github.com/axolotl-ai-cloud/axolotl#egg=axolotl\n",
+    "!pip install flash-attn==\"2.5.0\"\n",
+    "!pip install deepspeed==\"0.13.1\"!pip install mlflow==\"2.13.0\""
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "id": "BW2MFr7HTjub"
+   },
+   "source": [
+    "## Create an yaml config file"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "id": "9pkF2dSoQEUN"
+   },
+   "outputs": [],
+   "source": [
+    "import yaml\n",
+    "\n",
+    "# Your YAML string\n",
+    "yaml_string = \"\"\"\n",
+    "base_model: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T\n",
+    "model_type: LlamaForCausalLM\n",
+    "tokenizer_type: LlamaTokenizer\n",
+    "\n",
+    "load_in_8bit: false\n",
+    "load_in_4bit: true\n",
+    "strict: false\n",
+    "\n",
+    "datasets:\n",
+    "  - path: mhenrichsen/alpaca_2k_test\n",
+    "    type: alpaca\n",
+    "dataset_prepared_path:\n",
+    "val_set_size: 0.05\n",
+    "output_dir: ./outputs/qlora-out\n",
+    "\n",
+    "adapter: qlora\n",
+    "lora_model_dir:\n",
+    "\n",
+    "sequence_len: 4096\n",
+    "sample_packing: true\n",
+    "eval_sample_packing: false\n",
+    "pad_to_sequence_len: true\n",
+    "\n",
+    "lora_r: 32\n",
+    "lora_alpha: 16\n",
+    "lora_dropout: 0.05\n",
+    "lora_target_modules:\n",
+    "lora_target_linear: true\n",
+    "lora_fan_in_fan_out:\n",
+    "\n",
+    "wandb_project:\n",
+    "wandb_entity:\n",
+    "wandb_watch:\n",
+    "wandb_name:\n",
+    "wandb_log_model:\n",
+    "\n",
+    "gradient_accumulation_steps: 4\n",
+    "micro_batch_size: 2\n",
+    "num_epochs: 4\n",
+    "optimizer: paged_adamw_32bit\n",
+    "lr_scheduler: cosine\n",
+    "learning_rate: 0.0002\n",
+    "\n",
+    "train_on_inputs: false\n",
+    "group_by_length: false\n",
+    "bf16: auto\n",
+    "fp16:\n",
+    "tf32: false\n",
+    "\n",
+    "gradient_checkpointing: true\n",
+    "early_stopping_patience:\n",
+    "resume_from_checkpoint:\n",
+    "local_rank:\n",
+    "logging_steps: 1\n",
+    "xformers_attention:\n",
+    "flash_attention: true\n",
+    "\n",
+    "warmup_steps: 10\n",
+    "evals_per_epoch: 4\n",
+    "saves_per_epoch: 1\n",
+    "debug:\n",
+    "deepspeed:\n",
+    "weight_decay: 0.0\n",
+    "fsdp:\n",
+    "fsdp_config:\n",
+    "special_tokens:\n",
+    "\n",
+    "\"\"\"\n",
+    "\n",
+    "# Convert the YAML string to a Python dictionary\n",
+    "yaml_dict = yaml.safe_load(yaml_string)\n",
+    "\n",
+    "# Specify your file path\n",
+    "file_path = 'test_axolotl.yaml'\n",
+    "\n",
+    "# Write the YAML file\n",
+    "with open(file_path, 'w') as file:\n",
+    "    yaml.dump(yaml_dict, file)\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "id": "bidoj8YLTusD"
+   },
+   "source": [
+    "## Launch the training"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "colab": {
+     "base_uri": "https://localhost:8080/"
+    },
+    "id": "ydTI2Jk2RStU",
+    "outputId": "d6d0df17-4b53-439c-c802-22c0456d301b"
+   },
+   "outputs": [],
+   "source": [
+    "# By using the ! the comand will be executed as a bash command\n",
+    "!accelerate launch -m axolotl.cli.train /content/test_axolotl.yaml"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Play with inference"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# By using the ! the comand will be executed as a bash command\n",
+    "!accelerate launch -m axolotl.cli.inference /content/test_axolotl.yaml \\\n",
+    "    --qlora_model_dir=\"./qlora-out\" --gradio"
+   ]
+  }
+ ],
+ "metadata": {
+  "accelerator": "GPU",
+  "colab": {
+   "gpuType": "T4",
+   "provenance": []
+  },
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.12.1"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
 }

examples/dbrx/16bit-lora.yaml

@@ -0,0 +1,81 @@
+base_model: LnL-AI/dbrx-base-converted-v2
+trust_remote_code: true
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: tatsu-lab/alpaca
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.0
+output_dir: ./outputs/out
+
+sequence_len: 512
+sample_packing: false
+pad_to_sequence_len: false
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+adapter: lora
+lora_model_dir:
+lora_r: 8
+lora_alpha: 16
+lora_dropout: 0.05
+# w1, w2, & v1 will hang the trainer
+lora_target_modules:
+  - q_proj # attn
+  - k_proj # attn
+  - v_proj # attn
+  - out_proj # attn
+  - layer # router
+#  - w1
+#  - w2
+#  - v1
+
+gradient_accumulation_steps: 1
+micro_batch_size: 1
+num_epochs: 1
+optimizer: paged_adamw_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: false  # don't use with fsdp_activation_checkpointing
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 10
+evals_per_epoch:
+saves_per_epoch: 1
+debug:
+weight_decay: 0.0
+fsdp:
+  - full_shard
+  - auto_wrap
+fsdp_config:
+  fsdp_limit_all_gathers: true
+  fsdp_sync_module_states: true
+  fsdp_offload_params: false
+  fsdp_use_orig_params: false
+  fsdp_cpu_ram_efficient_loading: true
+  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
+  fsdp_transformer_layer_cls_to_wrap: DbrxBlock
+  fsdp_state_dict_type: FULL_STATE_DICT
+  fsdp_activation_checkpointing: true

examples/dbrx/8bit-lora.yaml

@@ -0,0 +1,81 @@
+base_model: LnL-AI/dbrx-base-converted-v2
+trust_remote_code: true
+
+load_in_8bit: true
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: tatsu-lab/alpaca
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.0
+output_dir: ./outputs/out
+
+sequence_len: 512
+sample_packing: false
+pad_to_sequence_len: false
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+adapter: lora
+lora_model_dir:
+lora_r: 8
+lora_alpha: 16
+lora_dropout: 0.05
+# w1, w2, & v1 will hang the trainer
+lora_target_modules:
+  - q_proj # attn
+  - k_proj # attn
+  - v_proj # attn
+  - out_proj # attn
+  - layer # router
+#  - w1
+#  - w2
+#  - v1
+
+gradient_accumulation_steps: 1
+micro_batch_size: 1
+num_epochs: 1
+optimizer: paged_adamw_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: false  # don't use with fsdp_activation_checkpointing
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 10
+evals_per_epoch:
+saves_per_epoch: 1
+debug:
+weight_decay: 0.0
+fsdp:
+  - full_shard
+  - auto_wrap
+fsdp_config:
+  fsdp_limit_all_gathers: true
+  fsdp_sync_module_states: true
+  fsdp_offload_params: false
+  fsdp_use_orig_params: false
+  fsdp_cpu_ram_efficient_loading: true
+  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
+  fsdp_transformer_layer_cls_to_wrap: DbrxBlock
+  fsdp_state_dict_type: FULL_STATE_DICT
+  fsdp_activation_checkpointing: true

examples/dbrx/README.md

@@ -0,0 +1,26 @@
+# DBRX MoE
+
+Currently, for LoRA, only the `q_proj`, `k_proj`, `v_proj` `out_proj` and `layer` Linear layers are trainable.
+
+We are using the "converted" base models based on [this issue](https://huggingface.co/databricks/dbrx-instruct/discussions/10)
+where the Experts are fused as an `nn.Parameter` rather than a `nn.Linear` layer. However, the implementation
+is still a bit buggy and attempting to train a LoRA adapter over those `w1`, `w2` and `v1` layers
+results in the trainer hanging.
+
+
+### FSDP
+We've tested using the [`LnL-AI/dbrx-base-converted-v2`](https://huggingface.co/LnL-AI/dbrx-base-converted-v2) model as the base model for FSDP.
+
+The high memory usage seen w/ FSDP is due to FSDP not supporting 8bit optimizers.
+
+- 16-bit LoRA w/ FSDP
+  - ✅ w/o CPU Offload - 8x80GB uses ~80GiB/gpu
+  - ❌ w/ CPU Offload - `paged_adamw_8bit` optimizer errors from being on cpu
+- ✅ 8-bit LoRA w/ FSDP
+- ❌ 4-bit QLoRA w/ FSDP - errors w/: `Error an illegal memory access was encountered at line 90 in file /src/csrc/ops.cu`
+- ✅ bf16 full finetune w/ FSDP, freezing all but first 8 layers (8x80GB uses ~78GiB/gpu)
+
+
+### Deepspeed
+
+WIP

examples/dbrx/fft-ds-zero3.yaml

@@ -0,0 +1,56 @@
+base_model: LnL-AI/dbrx-base-converted-v2
+trust_remote_code: true
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: tatsu-lab/alpaca
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.0
+output_dir: ./outputs/out
+
+sequence_len: 512
+sample_packing: false
+pad_to_sequence_len: false
+
+unfrozen_parameters:
+  - transformer.blocks.[0-7].
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 1
+micro_batch_size: 1
+num_epochs: 1
+optimizer: paged_adamw_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 10
+evals_per_epoch:
+saves_per_epoch: 1
+debug:
+weight_decay: 0.0
+deepspeed: deepspeed_configs/zero3_bf16.json

examples/deepseek-v2/fft-fsdp-16b.yaml

@@ -0,0 +1,67 @@
+base_model: deepseek-ai/DeepSeek-V2-Lite
+trust_remote_code: true
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: tatsu-lab/alpaca
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.0
+output_dir: ./outputs/out
+
+sequence_len: 2048
+sample_packing: true
+pad_to_sequence_len: true
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 8
+micro_batch_size: 1
+num_epochs: 1
+optimizer: adamw_torch
+lr_scheduler: cosine
+learning_rate: 2e-5
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+early_stopping_patience:
+resume_from_checkpoint:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 100
+evals_per_epoch: 2
+eval_table_size:
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+special_tokens:
+fsdp:
+  - full_shard
+  - auto_wrap
+fsdp_config:
+  fsdp_limit_all_gathers: true
+  fsdp_sync_module_states: true
+  fsdp_offload_params: true
+  fsdp_use_orig_params: false
+  fsdp_cpu_ram_efficient_loading: true
+  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
+  fsdp_transformer_layer_cls_to_wrap: DeepseekV2DecoderLayer
+  fsdp_state_dict_type: FULL_STATE_DICT
+  fsdp_sharding_strategy: FULL_SHARD

examples/deepseek-v2/qlora-fsdp-2_5.yaml

@@ -0,0 +1,86 @@
+base_model: axolotl-quants/DeepSeek-V2.5-bnb-nf4-bf16
+trust_remote_code: true
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+
+plugins:
+  - axolotl.integrations.liger.LigerPlugin
+liger_rms_norm: true
+liger_swiglu: true
+liger_fused_linear_cross_entropy: true
+
+chat_template: deepseek_v2
+datasets:
+  - path: mlabonne/FineTome-100k
+    type: chat_template
+    split: train[:20%]
+    field_messages: conversations
+    message_field_role: from
+    message_field_content: value
+
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.0
+output_dir: ./outputs/out
+
+sequence_len: 4096
+sample_packing: true
+pad_to_sequence_len: true
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+adapter: qlora
+lora_r: 256
+lora_alpha: 256
+lora_target_linear: true
+peft_use_rslora: true
+
+gradient_accumulation_steps: 1
+micro_batch_size: 8
+num_epochs: 1
+optimizer: adamw_torch
+lr_scheduler: cosine
+learning_rate: 2e-5
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+early_stopping_patience:
+resume_from_checkpoint:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 100
+evals_per_epoch: 2
+eval_table_size:
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+special_tokens:
+fsdp:
+  - full_shard
+  - auto_wrap
+fsdp_config:
+  fsdp_limit_all_gathers: true
+  fsdp_sync_module_states: true
+  fsdp_offload_params: true
+  fsdp_use_orig_params: false
+  fsdp_cpu_ram_efficient_loading: true
+  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
+  fsdp_transformer_layer_cls_to_wrap: DeepseekV2DecoderLayer
+  fsdp_state_dict_type: FULL_STATE_DICT
+  fsdp_sharding_strategy: FULL_SHARD

examples/falcon/config-7b-lora.yml

@@ -28,7 +28,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./falcon-7b
+output_dir: ./outputs/falcon-7b
 batch_size: 2
 micro_batch_size: 1
 num_epochs: 4

examples/falcon/config-7b-qlora.yml

@@ -42,7 +42,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 # QLoRA paper Table 9
 # - 16 for 7b & 13b

examples/falcon/config-7b.yml

@@ -28,7 +28,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./falcon-7b
+output_dir: ./outputs/falcon-7b
 batch_size: 2
 micro_batch_size: 1
 num_epochs: 4

examples/gemma/qlora.yml

@@ -12,7 +12,7 @@ datasets:
   - path: mhenrichsen/alpaca_2k_test
     type: alpaca
 val_set_size: 0.1
-output_dir: ./out
+output_dir: ./outputs/out
 
 adapter: qlora
 lora_r: 32

examples/gemma2/qlora.yml

@@ -0,0 +1,71 @@
+base_model: google/gemma-2-9b
+model_type: AutoModelForCausalLM
+tokenizer_type: AutoTokenizer
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+# huggingface repo
+chat_template: gemma
+datasets:
+  - path: cgato/SlimOrcaDedupCleaned
+    type: chat_template
+    drop_system_message: true
+    field_messages: conversations
+    message_field_role: from
+    message_field_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_linear: true
+
+sequence_len: 2048
+sample_packing: true
+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
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: true
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_ratio: 0.1
+evals_per_epoch:
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:

examples/gemma2/reward-model.yaml

@@ -0,0 +1,63 @@
+base_model: google/gemma-2-2b
+model_type: AutoModelForSequenceClassification
+tokenizer_type: AutoTokenizer
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+reward_model: true
+chat_template: gemma
+datasets:
+  - path: argilla/distilabel-intel-orca-dpo-pairs
+    type: bradley_terry.chat_template
+val_set_size: 0.0
+output_dir: ./outputs/out
+remove_unused_columns: false
+
+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: 2
+num_epochs: 4
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: true
+fp16:
+tf32: true
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_ratio: 0.1
+evals_per_epoch:
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:

examples/gptj/qlora.yml

@@ -23,7 +23,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 gradient_accumulation_steps: 2
 micro_batch_size: 2
 num_epochs: 2

examples/jamba/README.md

@@ -6,5 +6,5 @@
 - ✅ qlora w/ deepspeed Zero-3 needs at least 2x GPUs and 67GiB VRAM (wtf?)
 - ✅ qlora single-gpu, ~51GiB VRAM
 - ✅ multipack
-- ❓ FSDP
+- ✅ FSDP
 - ❓ 8-bit LoRA

examples/jamba/qlora.yaml

@@ -10,7 +10,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.0
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 4096
 sample_packing: false

examples/jamba/qlora_deepspeed.yaml

@@ -10,7 +10,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.0
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 4096
 sample_packing: false

examples/jamba/qlora_fsdp_large.yaml

@@ -0,0 +1,65 @@
+base_model: ai21labs/AI21-Jamba-1.5-Large
+tokenizer_type: AutoTokenizer
+
+load_in_4bit: true
+strict: false
+use_tensorboard: true
+chat_template: jamba
+datasets:
+  - path: cgato/SlimOrcaDedupCleaned
+    type: chat_template
+    drop_system_message: true
+    field_messages: conversations
+    message_field_role: from
+    message_field_content: value
+
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.0
+output_dir: jamba-large-fsdp-qlora-ft
+save_safetensors: true
+adapter: qlora
+sequence_len: 2048
+sample_packing: true
+pad_to_sequence_len: true
+
+lora_r: 16
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_modules: [down_proj,gate_proj,in_proj,k_proj,o_proj,out_proj,q_proj,up_proj,v_proj,x_proj]
+lora_target_linear: false
+
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 2
+optimizer: adamw_torch
+lr_scheduler: cosine
+learning_rate: 0.00001
+
+train_on_inputs: false
+group_by_length: false
+bf16: true
+tf32: true
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: true
+logging_steps: 1
+flash_attention: true
+
+warmup_steps: 10
+evals_per_epoch: 1
+saves_per_epoch: 1
+weight_decay: 0.0
+fsdp:
+  - full_shard
+  - auto_wrap
+fsdp_config:
+  fsdp_limit_all_gathers: true
+  fsdp_sync_module_states: true
+  fsdp_offload_params: false
+  fsdp_use_orig_params: false
+  fsdp_cpu_ram_efficient_loading: true
+  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
+  fsdp_transformer_layer_cls_to_wrap: JambaAttentionDecoderLayer,JambaMambaDecoderLayer
+  fsdp_state_dict_type: FULL_STATE_DICT
+  fsdp_sharding_strategy: FULL_SHARD

examples/jeopardy-bot/config.yml

@@ -21,7 +21,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./jeopardy-bot-7b
+output_dir: ./outputs/jeopardy-bot-7b
 gradient_accumulation_steps: 1
 micro_batch_size: 1
 num_epochs: 4

examples/llama-2/fft_optimized.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.05
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 4096
 sample_packing: true

examples/llama-2/gptq-lora.yml

@@ -33,7 +33,7 @@ wandb_project:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./model-out
+output_dir: ./outputs/model-out
 gradient_accumulation_steps: 1
 micro_batch_size: 1
 num_epochs: 4

examples/llama-2/lisa.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.05
-output_dir: ./lisa-out
+output_dir: ./outputs/lisa-out
 
 sequence_len: 4096
 sample_packing: true

examples/llama-2/loftq.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/llama-2/lora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/llama-2/qlora-fsdp.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.05
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 adapter: qlora
 lora_model_dir:
@@ -65,12 +65,14 @@ deepspeed:
 weight_decay: 0.0
 fsdp:
   - full_shard
+  - auto_wrap
 fsdp_config:
   fsdp_limit_all_gathers: true
   fsdp_sync_module_states: true
   fsdp_offload_params: true
   fsdp_use_orig_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: SHARDED_STATE_DICT
+  fsdp_state_dict_type: FULL_STATE_DICT
 special_tokens:

examples/llama-2/qlora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 adapter: qlora
 lora_model_dir:

examples/llama-2/relora.yml

@@ -12,7 +12,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./relora-out
+output_dir: ./outputs/relora-out
 
 adapter: qlora
 lora_model_dir:

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

@@ -0,0 +1,63 @@
+base_model: alpindale/Llama-3.2-11B-Vision-Instruct
+processor_type: AutoProcessor
+strict: false
+
+# 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: llama3_2_vision
+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: 'language_model.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
+
+train_on_inputs: false
+group_by_length: false
+bf16: true
+fp16:
+tf32: true
+
+gradient_checkpointing: true
+local_rank:
+logging_steps: 1
+flash_attention: true
+eager_attention:
+
+warmup_ratio: 0.1
+evals_per_epoch: 1
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:

examples/llama-3/README.md

@@ -0,0 +1,13 @@
+# Llama-3
+
+https://llama.meta.com/llama3/
+
+[8B Base Model](https://huggingface.co/meta-llama/Meta-Llama-3-8B)
+ - [Full Fine Tune](./fft-8b.yaml)
+   - Single GPU @ 48GB VRAM
+ - [LoRA](./lora-8b.yml)
+   - Single GPU @ 11GB VRAM
+
+[70B Base Model](https://huggingface.co/meta-llama/Meta-Llama-3-70B)
+ - [QLORA+FSDP](./qlora-fsdp-70b.yaml)
+   - Dual GPU @ 21GB VRAM

examples/llama-3/fft-8b-liger-fsdp.yaml

@@ -0,0 +1,80 @@
+base_model: NousResearch/Meta-Llama-3.1-8B
+
+plugins:
+  - axolotl.integrations.liger.LigerPlugin
+liger_rope: true
+liger_rms_norm: true
+liger_swiglu: true
+liger_fused_linear_cross_entropy: true
+
+strict: false
+
+chat_template: llama3
+datasets:
+  - path: mlabonne/FineTome-100k
+    type: chat_template
+    split: train[:20%]
+    field_messages: conversations
+    message_field_role: from
+    message_field_content: value
+
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.02
+output_dir: ./outputs/out
+
+sequence_len: 4096
+sample_packing: true
+pad_to_sequence_len: true
+
+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
+lr_scheduler: cosine
+learning_rate: 2e-5
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+early_stopping_patience:
+resume_from_checkpoint:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 100
+evals_per_epoch: 2
+eval_table_size:
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+  - full_shard
+  - auto_wrap
+fsdp_config:
+  fsdp_limit_all_gathers: true
+  fsdp_sync_module_states: true
+  fsdp_offload_params: true
+  fsdp_use_orig_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_backward_prefetch: BACKWARD_PRE
+special_tokens:
+  pad_token: <|finetune_right_pad_id|>
+  eos_token: <|eot_id|>

examples/llama-3/fft-8b.yaml

@@ -0,0 +1,56 @@
+base_model: NousResearch/Meta-Llama-3.1-8B
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: tatsu-lab/alpaca
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.05
+output_dir: ./outputs/out
+
+sequence_len: 8192
+sample_packing: true
+pad_to_sequence_len: true
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 8
+micro_batch_size: 1
+num_epochs: 1
+optimizer: paged_adamw_8bit
+lr_scheduler: cosine
+learning_rate: 2e-5
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+early_stopping_patience:
+resume_from_checkpoint:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 100
+evals_per_epoch: 2
+eval_table_size:
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:
+  pad_token: <|end_of_text|>

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

@@ -0,0 +1,80 @@
+base_model: meta-llama/Meta-Llama-3-8B-Instruct
+model_type: LlamaForCausalLM
+tokenizer_type: AutoTokenizer
+
+load_in_8bit: true
+load_in_4bit: false
+strict: false
+
+chat_template: llama3
+rl: dpo
+datasets:
+  - path: fozziethebeat/alpaca_messages_2k_dpo_test
+    type: chat_template.default
+    field_messages: conversation
+    field_chosen: chosen
+    field_rejected: rejected
+    message_field_role: role
+    message_field_content: content
+    roles:
+      system:
+        - system
+      user:
+        - user
+      assistant:
+        - assistant
+
+dataset_prepared_path:
+val_set_size: 0.05
+output_dir: ./outputs/lora-out
+
+sequence_len: 4096
+sample_packing: false
+pad_to_sequence_len: true
+
+adapter: lora
+lora_model_dir:
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 2
+num_epochs: 4
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+s2_attention:
+
+warmup_steps: 10
+evals_per_epoch: 4
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:

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

@@ -0,0 +1,77 @@
+base_model: NousResearch/Meta-Llama-3-8B-Instruct
+model_type: LlamaForCausalLM
+tokenizer_type: AutoTokenizer
+
+load_in_8bit: true
+load_in_4bit: false
+strict: false
+
+chat_template: llama3
+datasets:
+  - path: fozziethebeat/alpaca_messages_2k_test
+    type: chat_template
+    field_messages: messages
+    message_field_role: role
+    message_field_content: content
+    roles:
+      user:
+        - user
+      assistant:
+        - assistant
+
+dataset_prepared_path:
+val_set_size: 0.05
+output_dir: ./outputs/lora-out
+
+sequence_len: 4096
+sample_packing: false
+pad_to_sequence_len: true
+
+adapter: lora
+lora_model_dir:
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 2
+num_epochs: 4
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+s2_attention:
+
+warmup_steps: 10
+evals_per_epoch: 4
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:
+   pad_token: <|end_of_text|>

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

@@ -0,0 +1,71 @@
+base_model: NousResearch/Meta-Llama-3-8B
+model_type: LlamaForCausalLM
+tokenizer_type: AutoTokenizer
+
+load_in_8bit: true
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: mhenrichsen/alpaca_2k_test
+    type: alpaca
+dataset_prepared_path:
+val_set_size: 0.05
+output_dir: ./outputs/lora-out
+
+sequence_len: 4096
+sample_packing: true
+eval_sample_packing: false
+pad_to_sequence_len: true
+
+adapter: lora
+lora_model_dir:
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+lora_modules_to_save:
+  - embed_tokens
+  - lm_head
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 2
+num_epochs: 4
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+s2_attention:
+
+warmup_steps: 10
+evals_per_epoch: 4
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:
+   pad_token: <|end_of_text|>

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

@@ -0,0 +1,77 @@
+base_model: meta-llama/Llama-3.2-1B
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+datasets:
+  - path: teknium/GPT4-LLM-Cleaned
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.1
+output_dir: ./outputs/qlora-out
+
+adapter: qlora
+lora_model_dir:
+
+sequence_len: 2048
+sample_packing: true
+eval_sample_packing: true
+pad_to_sequence_len: true
+
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+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
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+loss_watchdog_threshold: 5.0
+loss_watchdog_patience: 3
+
+warmup_steps: 10
+evals_per_epoch: 4
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:
+  pad_token: "<|end_of_text|>"

examples/llama-3/qlora-fsdp-405b.yaml

@@ -0,0 +1,63 @@
+base_model: hugging-quants/Meta-Llama-3.1-405B-BNB-NF4-BF16
+tokenizer_type: AutoTokenizer
+
+load_in_4bit: true
+strict: false
+
+datasets:
+  - path: tatsu-lab/alpaca
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.0
+output_dir: ./outputs/out/qlora-llama3_1-405b
+save_safetensors: true
+
+adapter: qlora
+
+sequence_len: 2048
+sample_packing: true
+pad_to_sequence_len: true
+
+lora_r: 16
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_modules:
+lora_target_linear: true
+
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 2
+optimizer: adamw_torch
+lr_scheduler: cosine
+learning_rate: 0.00001
+
+train_on_inputs: false
+group_by_length: false
+bf16: true
+tf32: true
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: true
+logging_steps: 1
+flash_attention: true
+
+warmup_steps: 10
+evals_per_epoch: 4
+saves_per_epoch: 1
+weight_decay: 0.0
+fsdp:
+  - full_shard
+  - auto_wrap
+fsdp_config:
+  fsdp_limit_all_gathers: true
+  fsdp_sync_module_states: true
+  fsdp_offload_params: true
+  fsdp_use_orig_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
+special_tokens:
+  pad_token: <|finetune_right_pad_id|>

examples/llama-3/qlora-fsdp-70b.yaml

@@ -0,0 +1,80 @@
+base_model: casperhansen/llama-3-70b-fp16
+model_type: LlamaForCausalLM
+tokenizer_type: AutoTokenizer  # PreTrainedTokenizerFast
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+datasets:
+  - path: tatsu-lab/alpaca
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.05
+output_dir: ./outputs/out/qlora-llama3-70b
+
+adapter: qlora
+lora_model_dir:
+
+sequence_len: 512
+sample_packing: false
+pad_to_sequence_len: true
+
+lora_r: 8
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_modules:
+lora_target_linear: true
+lora_fan_in_fan_out:
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 4
+optimizer: adamw_torch
+lr_scheduler: cosine
+learning_rate: 0.00001
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 10
+evals_per_epoch: 4
+eval_table_size:
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+  - full_shard
+  - auto_wrap
+fsdp_config:
+  fsdp_limit_all_gathers: true
+  fsdp_sync_module_states: true
+  fsdp_offload_params: true
+  fsdp_use_orig_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
+special_tokens:
+  pad_token: <|end_of_text|>

examples/llama-3/qlora.yml

@@ -0,0 +1,67 @@
+base_model: NousResearch/Meta-Llama-3-8B
+model_type: AutoModelForCausalLM
+tokenizer_type: AutoTokenizer
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+datasets:
+  - path: aaditya/alpaca_subset_1
+    type: alpaca
+dataset_prepared_path:
+val_set_size: 0
+output_dir: ./outputs/qlora-out
+
+adapter: qlora
+lora_model_dir:
+
+sequence_len: 4096
+sample_packing: true
+pad_to_sequence_len: true
+
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_modules:
+lora_target_linear: true
+lora_fan_in_fan_out:
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 2
+num_epochs: 4
+optimizer: paged_adamw_32bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 10
+evals_per_epoch: 4
+eval_table_size:
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:
+  pad_token: "<|end_of_text|>"

examples/mamba/config.yml

@@ -12,7 +12,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.0
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 2048
 sample_packing: false

examples/mistral/bigstral-ds-zero3.yaml

@@ -0,0 +1,63 @@
+base_model: mistral-community/Mixtral-8x22B-v0.1
+model_type: AutoModelForCausalLM
+tokenizer_type: LlamaTokenizer
+trust_remote_code: true
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+unfrozen_parameters:
+  - ^lm_head.weight$
+  - ^model.embed_tokens.weight$
+  - model.layers.4[4-9]+.block_sparse_moe.gate
+  - model.layers.4[4-9]+.block_sparse_moe.experts
+  - model.layers.5[0-5]+.block_sparse_moe.gate
+  - model.layers.5[0-5]+.block_sparse_moe.experts
+
+model_config:
+  output_router_logits: true
+
+datasets:
+  - path: tatsu-lab/alpaca
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.05
+output_dir: ./outputs/out
+
+sequence_len: 2048
+sample_packing: true
+pad_to_sequence_len: true
+
+gradient_accumulation_steps: 1
+micro_batch_size: 1
+num_epochs: 3
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0001
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+save_total_limit: 1
+save_steps:
+debug:
+deepspeed: deepspeed_configs/zero3_bf16_cpuoffload_params.json
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:
+  eos_token: "<|im_end|>"
+tokens:
+  - "<|im_start|>"

examples/mistral/config.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 8192
 sample_packing: true

examples/mistral/lora-mps.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 eval_sample_packing: false
 
 adapter: lora

examples/mistral/lora.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.1
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 adapter: lora
 lora_model_dir:

examples/mistral/mistral-qlora-fsdp.yml

@@ -0,0 +1,82 @@
+base_model: mistralai/Mixtral-8x7B-v0.1
+model_type: AutoModelForCausalLM
+tokenizer_type: LlamaTokenizer
+trust_remote_code: true
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+datasets:
+  - path: tatsu-lab/alpaca
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.02
+output_dir: ./outputs/qlora-out
+
+model_config:
+  output_router_logits: true
+
+adapter: qlora
+lora_model_dir:
+
+sequence_len: 1024
+sample_packing: false
+pad_to_sequence_len: false
+
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 2
+num_epochs: 1
+optimizer: paged_adamw_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+loss_watchdog_threshold: 5.0
+loss_watchdog_patience: 3
+
+warmup_steps: 10
+evals_per_epoch: 4
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+weight_decay: 0.0
+fsdp:
+  - full_shard
+  - auto_wrap
+fsdp_config:
+  fsdp_limit_all_gathers: true
+  fsdp_sync_module_states: true
+  fsdp_offload_params: false
+  fsdp_use_orig_params: false
+  fsdp_cpu_ram_efficient_loading: false
+  fsdp_transformer_layer_cls_to_wrap: MistralDecoderLayer
+  fsdp_state_dict_type: FULL_STATE_DICT
+  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
+special_tokens:

examples/mistral/mistral-qlora-orpo.yml

@@ -0,0 +1,82 @@
+base_model: mistralai/Mistral-7B-v0.1
+model_type: MistralForCausalLM
+tokenizer_type: LlamaTokenizer
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+rl: orpo
+orpo_alpha: 0.1
+remove_unused_columns: false
+
+chat_template: chatml
+datasets:
+  - path: argilla/ultrafeedback-binarized-preferences-cleaned
+    type: chat_template.argilla
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.1
+output_dir: ./outputs/mistral-qlora-orpo-out
+
+adapter: qlora
+lora_model_dir:
+
+sequence_len: 4096
+sample_packing: false
+pad_to_sequence_len: true
+
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+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
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+loss_watchdog_threshold: 5.0
+loss_watchdog_patience: 3
+
+warmup_steps: 10
+evals_per_epoch: 4
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:

examples/mistral/mixtral-8x22b-qlora-fsdp.yml

@@ -0,0 +1,81 @@
+base_model: mistral-community/Mixtral-8x22B-v0.1
+model_type: AutoModelForCausalLM
+tokenizer_type: LlamaTokenizer
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+datasets:
+  - path: tatsu-lab/alpaca
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.02
+output_dir: ./outputs/qlora-out
+
+model_config:
+  output_router_logits: true
+
+adapter: qlora
+lora_model_dir:
+
+sequence_len: 1024
+sample_packing: false
+pad_to_sequence_len: false
+
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+
+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
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: true
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+loss_watchdog_threshold: 5.0
+loss_watchdog_patience: 3
+
+warmup_steps: 10
+evals_per_epoch: 4
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+weight_decay: 0.0
+fsdp:
+  - full_shard
+  - auto_wrap
+fsdp_config:
+  fsdp_limit_all_gathers: true
+  fsdp_sync_module_states: true
+  fsdp_offload_params: true
+  fsdp_use_orig_params: false
+  fsdp_cpu_ram_efficient_loading: true
+  fsdp_transformer_layer_cls_to_wrap: MixtralSparseMoeBlock
+  fsdp_state_dict_type: FULL_STATE_DICT
+  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
+special_tokens:

examples/mistral/mixtral-qlora-fsdp.yml

@@ -12,7 +12,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.02
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 model_config:
   output_router_logits: true
@@ -39,7 +39,7 @@ wandb_log_model:
 gradient_accumulation_steps: 4
 micro_batch_size: 2
 num_epochs: 1
-optimizer: paged_adamw_8bit
+optimizer: adamw_torch
 lr_scheduler: cosine
 learning_rate: 0.0002
 
@@ -47,7 +47,7 @@ train_on_inputs: false
 group_by_length: false
 bf16: auto
 fp16:
-tf32: false
+tf32: true
 
 gradient_checkpointing: true
 early_stopping_patience:
@@ -69,6 +69,17 @@ debug:
 weight_decay: 0.0
 fsdp:
   - full_shard
+  - auto_wrap
 fsdp_config:
+  fsdp_limit_all_gathers: true
+  fsdp_sync_module_states: true
+  fsdp_offload_params: true
+  fsdp_use_orig_params: false
+  fsdp_cpu_ram_efficient_loading: true
   fsdp_transformer_layer_cls_to_wrap: MixtralSparseMoeBlock
+  fsdp_state_dict_type: FULL_STATE_DICT
+  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
+  fsdp_sharding_strategy: FULL_SHARD
+  fsdp_forward_prefetch: false
+  fsdp_backward_prefetch: BACKWARD_PRE
 special_tokens: