add transformers module to cleanup

* allow flexibility in transformers version for FSDP

* more flexibility with dev versions of 4.47.0.dev0

* add patch for fsdp

* fix typo

* correct fn name

* stray character

* fix patch

* reset Trainer too

* also reset Trainer.training_step

* allow tests/patched to run more than one process on e2e runner

* skip tests/patched in e2e for now since it's run in regular pytest

.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

@@ -1,57 +1,74 @@
 name: ci-cd-base
 
 on:
+  push:
+    branches:
+      - "main"
+    paths:
+      - 'Dockerfile-base'
+      - '.github/workflows/base.yml'
+  pull_request:
+    paths:
+      - 'Dockerfile-base'
+      - '.github/workflows/base.yml'
   workflow_dispatch:
 
 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
+            cuda_version: 12.1.1
+            cudnn_version: 8
+            python_version: "3.11"
+            pytorch: 2.3.1
+            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: "3.10"
-            pytorch: 2.1.2
+            pytorch: 2.4.1
             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: 2.1.2
+            pytorch: 2.4.1
             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: 2.2.2
-            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
-            python_version: "3.11"
-            pytorch: 2.3.0
+            pytorch: 2.5.1
             torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
     steps:
       - name: Checkout
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
       - name: Docker metadata
         id: metadata
-        uses: docker/metadata-action@v3
+        uses: docker/metadata-action@v5
         with:
-          images: winglian/axolotl-base
+          images: |
+            winglian/axolotl-base
+            axolotlai/axolotl-base
       - name: Login to Docker Hub
         uses: docker/login-action@v2
         with:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_TOKEN }}
       - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v2
+        uses: docker/setup-buildx-action@v3
       - name: Build
         uses: docker/build-push-action@v4
         with:
@@ -62,6 +79,7 @@ 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/docs.yml

@@ -17,7 +17,7 @@ jobs:
         - name: Set up Quarto
           uses: quarto-dev/quarto-actions/setup@v2
         - name: Setup Python
-          uses: actions/setup-python@v3
+          uses: actions/setup-python@v5
           with:
             python-version: '3.10'
         - name: install dependencies

.github/workflows/lint.yml

@@ -6,7 +6,7 @@ on:
        - '**.py'
        - 'requirements.txt'
        - '.github/workflows/*.yml'
-       - "*.md"
+       - "*.[q]md"
        - "examples/**/*.y[a]?ml"
   workflow_dispatch:
 
@@ -15,9 +15,9 @@ jobs:
     name: pre-commit
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v3
-      - uses: actions/setup-python@v4
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
         with:
           python-version: "3.10"
           cache: 'pip' # caching pip dependencies
-      - uses: pre-commit/action@v3.0.0
+      - uses: pre-commit/action@v3.0.1

.github/workflows/main.yml

@@ -4,36 +4,37 @@ on:
   push:
     branches:
       - "main"
+    tags:
+      - "v*"
   workflow_dispatch:
 
 jobs:
   build-axolotl:
-    if: ${{ ! contains(github.event.commits[0].message, '[skip docker]]') && github.repository_owner == 'OpenAccess-AI-Collective' }}
+    if: ${{ ! contains(github.event.commits[0].message, '[skip docker]') && github.repository_owner == 'axolotl-ai-cloud' }}
     strategy:
       fail-fast: false
       matrix:
         include:
-          - cuda: 118
-            cuda_version: 11.8.0
+          - cuda: 121
+            cuda_version: 12.1.1
             python_version: "3.10"
-            pytorch: 2.1.2
-            axolotl_extras:
-            axolotl_args: "--extra-index-url https://download.pytorch.org/whl/cu118"
+            pytorch: 2.3.1
+            axolotl_extras: mamba-ssm
+          - cuda: 121
+            cuda_version: 12.1.1
+            python_version: "3.11"
+            pytorch: 2.3.1
+            axolotl_extras: mamba-ssm
             is_latest: true
-          - cuda: 121
-            cuda_version: 12.1.0
-            python_version: "3.10"
-            pytorch: 2.1.2
-            axolotl_extras:
-          - cuda: 121
-            cuda_version: 12.1.0
+          - cuda: 124
+            cuda_version: 12.4.1
             python_version: "3.11"
-            pytorch: 2.2.2
+            pytorch: 2.4.1
             axolotl_extras:
-          - cuda: 121
-            cuda_version: 12.1.0
+          - cuda: 124
+            cuda_version: 12.4.1
             python_version: "3.11"
-            pytorch: 2.3.0
+            pytorch: 2.5.1
             axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:
@@ -43,7 +44,12 @@ jobs:
         id: metadata
         uses: docker/metadata-action@v5
         with:
-          images: winglian/axolotl
+          images: |
+            winglian/axolotl
+            axolotlai/axolotl
+          tags: |
+            type=ref,event=branch
+            type=pep440,pattern={{version}}
       - name: Set up Docker Buildx
         uses: docker/setup-buildx-action@v3
       - name: Login to Docker Hub
@@ -57,7 +63,7 @@ jobs:
         with:
           context: .
           build-args: |
-            BASE_TAG=${{ github.ref_name }}-base-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}
+            BASE_TAG=${{ github.ref_type == 'tag' && 'main' || github.ref_name }}-base-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}
             CUDA=${{ matrix.cuda }}
             PYTORCH_VERSION=${{ matrix.pytorch }}
             AXOLOTL_ARGS=${{ matrix.axolotl_args }}
@@ -65,36 +71,37 @@ 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: 2.2.2
+            pytorch: 2.4.1
             axolotl_extras:
-          - cuda: 121
-            cuda_version: 12.1.0
+          - cuda: 124
+            cuda_version: 12.4.1
             python_version: "3.11"
-            pytorch: 2.3.0
+            pytorch: 2.5.1
             axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:
@@ -104,20 +111,25 @@ jobs:
         id: metadata
         uses: docker/metadata-action@v5
         with:
-          images: winglian/axolotl-cloud
+          images: |
+            winglian/axolotl-cloud
+            axolotlai/axolotl-cloud
+          tags: |
+            type=ref,event=branch
+            type=pep440,pattern={{version}}
       - name: Login to Docker Hub
         uses: docker/login-action@v3
         with:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_TOKEN }}
       - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v2
+        uses: docker/setup-buildx-action@v3
       - name: Build
         uses: docker/build-push-action@v5
         with:
           context: .
           build-args: |
-            BASE_TAG=${{ github.ref_name }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
+            BASE_TAG=${{ github.ref_type == 'tag' && 'main' || 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
           push: ${{ github.event_name != 'pull_request' }}
@@ -128,15 +140,15 @@ jobs:
 
   build-axolotl-cloud-no-tmux:
     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: 121
-            cuda_version: 12.1.0
+            cuda_version: 12.1.1
             python_version: "3.11"
-            pytorch: 2.3.0
+            pytorch: 2.3.1
             axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:
@@ -146,20 +158,25 @@ jobs:
         id: metadata
         uses: docker/metadata-action@v5
         with:
-          images: winglian/axolotl-cloud-term
+          images: |
+            winglian/axolotl-cloud-term
+            axolotlai/axolotl-cloud-term
+          tags: |
+            type=ref,event=branch
+            type=pep440,pattern={{version}}
       - name: Login to Docker Hub
         uses: docker/login-action@v3
         with:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_TOKEN }}
       - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v2
+        uses: docker/setup-buildx-action@v3
       - name: Build
         uses: docker/build-push-action@v5
         with:
           context: .
           build-args: |
-            BASE_TAG=${{ github.ref_name }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
+            BASE_TAG=${{ github.ref_type == 'tag' && 'main' || 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' }}

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

@@ -0,0 +1,67 @@
+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
+
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
+jobs:
+  test-axolotl-multigpu:
+    if: ${{ ! contains(github.event.commits[0].message, '[skip e2e]') && github.repository_owner == 'axolotl-ai-cloud' }}
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - cuda: 121
+            cuda_version: 12.1.1
+            python_version: "3.11"
+            pytorch: 2.3.1
+            axolotl_extras:
+            num_gpus: 2
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.4.1
+            axolotl_extras:
+            num_gpus: 2
+            nightly_build: "true"
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.5.1
+            axolotl_extras:
+            num_gpus: 2
+            nightly_build: "true"
+    runs-on: [self-hosted, modal]
+    timeout-minutes: 120
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Install Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.10"
+      - name: Install Modal
+        run: |
+          python -m pip install --upgrade pip
+          pip install modal==0.63.64 jinja2
+      - name: Update env vars
+        run: |
+          echo "BASE_TAG=main-base-py${{ 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

@@ -7,32 +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
+          - 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:
-            axolotl_args: "--extra-index-url https://download.pytorch.org/whl/cu118"
             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: 2.2.2
+            pytorch: 2.4.1
             axolotl_extras:
-          - cuda: 121
-            cuda_version: 12.1.0
+          - cuda: 124
+            cuda_version: 12.4.1
             python_version: "3.11"
-            pytorch: 2.3.0
+            pytorch: 2.5.1
             axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:
@@ -42,7 +41,9 @@ jobs:
         id: metadata
         uses: docker/metadata-action@v5
         with:
-          images: winglian/axolotl
+          images: |
+            winglian/axolotl
+            axolotlai/axolotl
           tags: |
             type=raw,value={{ branch }}-{{ date 'YYYYMMDD' }}
       - name: Set up Docker Buildx
@@ -70,31 +71,31 @@ 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: 2.2.2
+            pytorch: 2.4.1
             axolotl_extras:
-          - cuda: 121
-            cuda_version: 12.1.0
+          - cuda: 124
+            cuda_version: 12.4.1
             python_version: "3.11"
-            pytorch: 2.3.0
+            pytorch: 2.5.1
             axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:
@@ -104,7 +105,9 @@ jobs:
         id: metadata
         uses: docker/metadata-action@v5
         with:
-          images: winglian/axolotl-cloud
+          images: |
+            winglian/axolotl-cloud
+            axolotlai/axolotl-cloud
           tags: |
             type=raw,value={{ branch }}-{{ date 'YYYYMMDD' }}
       - name: Login to Docker Hub
@@ -113,7 +116,7 @@ jobs:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_TOKEN }}
       - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v2
+        uses: docker/setup-buildx-action@v3
       - name: Build
         uses: docker/build-push-action@v5
         with:

.github/workflows/pypi.yml

@@ -3,12 +3,24 @@ name: publish pypi
 on:
   push:
     tags:
-      - '*'
+      - 'v*'
+  workflow_dispatch:
 
 jobs:
+  setup_release:
+    name: Create Release
+    runs-on: ubuntu-latest
+    permissions:
+      contents: write
+    steps:
+      - name: Create release
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: gh release create "$GITHUB_REF_NAME" # GITHUB_REF_NAME is the tag name in `on.push.tags` workflows
   pypi-publish:
     name: Upload release to PyPI
     runs-on: ubuntu-latest
+    needs: [setup_release]
     environment:
       name: pypi
       url: https://pypi.org/p/axolotl
@@ -16,10 +28,10 @@ jobs:
       id-token: write  # IMPORTANT: this permission is mandatory for trusted publishing
     steps:
       - name: Check out repository code
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
 
       - name: Setup Python
-        uses: actions/setup-python@v4
+        uses: actions/setup-python@v5
         with:
           python-version: "3.10"
 
@@ -27,7 +39,7 @@ 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
@@ -37,9 +49,9 @@ jobs:
         run: |
           sed -i -E 's/version="([0-9.]+)",/version="${{ steps.tag.outputs.TAG_NAME }}",/g' setup.py
 
-      - name: Build a binary wheel
+      - name: Build a source dist
         run: |
-          python setup.py sdist bdist_wheel
+          python setup.py sdist
 
       - name: Publish package distributions to PyPI
         uses: pypa/gh-action-pypi-publish@release/v1

.github/workflows/tests-nightly.yml

@@ -0,0 +1,135 @@
+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@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.10"
+          cache: 'pip' # caching pip dependencies
+      - uses: pre-commit/action@v3.0.1
+        env:
+          SKIP: no-commit-to-branch
+
+  pytest:
+    name: PyTest
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      max-parallel: 2
+      matrix:
+        python_version: ["3.10", "3.11"]
+        pytorch_version: ["2.3.1", "2.4.1", "2.5.1"]
+        exclude:
+          - python_version: "3.10"
+            pytorch_version: "2.4.1"
+          - python_version: "3.10"
+            pytorch_version: "2.5.1"
+    timeout-minutes: 20
+
+    steps:
+      - name: Check out repository code
+        uses: actions/checkout@v4
+
+      - name: Setup Python
+        uses: actions/setup-python@v5
+        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
+          sed -i 's#^datasets.*#datasets @ git+https://github.com/huggingface/datasets.git@main#' requirements.txt
+
+      - name: Install dependencies
+        run: |
+          pip3 install --upgrade pip
+          pip3 install --upgrade packaging
+          pip3 install -U -e .
+          python scripts/unsloth_install.py | sh
+          python scripts/cutcrossentropy_install.py | sh
+          pip3 install -r requirements-dev.txt -r requirements-tests.txt
+
+      - name: Ensure axolotl CLI was installed
+        run: |
+          axolotl --help
+
+      - name: Run tests
+        run: |
+          pytest -n8 --dist loadfile --ignore=tests/e2e/ --ignore=tests/patched/ tests/
+          pytest tests/patched/
+
+      - name: cleanup pip cache
+        run: |
+          find "$(pip cache dir)/http-v2" -type f -mtime +14 -exec rm {} \;
+
+  docker-e2e-tests:
+    if: github.repository_owner == 'axolotl-ai-cloud'
+    # this job needs to be run on self-hosted GPU runners...
+    runs-on: [self-hosted, modal]
+    timeout-minutes: 60
+    needs: [pre-commit, pytest]
+
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - cuda: 121
+            cuda_version: 12.1.1
+            python_version: "3.10"
+            pytorch: 2.3.1
+            num_gpus: 1
+            axolotl_extras: mamba-ssm
+            nightly_build: "true"
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.4.1
+            num_gpus: 1
+            axolotl_extras:
+            nightly_build: "true"
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.5.1
+            num_gpus: 1
+            axolotl_extras:
+            nightly_build: "true"
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Install Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.10"
+      - name: Install Modal
+        run: |
+          python -m pip install --upgrade pip
+          pip install modal==0.63.64 jinja2
+      - name: Update env vars
+        run: |
+          echo "BASE_TAG=main-base-py${{ 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

@@ -8,87 +8,163 @@ on:
       - '**.py'
       - 'requirements.txt'
       - '.github/workflows/*.yml'
+      - 'requirements-tests.txt'
+      - 'cicd/cicd.sh'
+      - 'cicd/Dockerfile.jinja'
   pull_request:
       paths:
        - '**.py'
        - 'requirements.txt'
        - '.github/workflows/*.yml'
+       - 'requirements-tests.txt'
+       - 'cicd/cicd.sh'
+       - 'cicd/Dockerfile.jinja'
   workflow_dispatch:
 
+# Cancel jobs on the same ref if a new one is triggered
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}
+
 jobs:
   pre-commit:
     name: pre-commit
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v3
-      - uses: actions/setup-python@v4
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
         with:
           python-version: "3.10"
           cache: 'pip' # caching pip dependencies
-      - uses: pre-commit/action@v3.0.0
+      - uses: pre-commit/action@v3.0.1
+        env:
+          SKIP: no-commit-to-branch
 
   pytest:
     name: PyTest
     runs-on: ubuntu-latest
     strategy:
       fail-fast: false
+      max-parallel: 2
       matrix:
         python_version: ["3.10", "3.11"]
+        pytorch_version: ["2.3.1", "2.4.1", "2.5.1"]
+        exclude:
+          - python_version: "3.10"
+            pytorch_version: "2.4.1"
+          - python_version: "3.10"
+            pytorch_version: "2.5.1"
     timeout-minutes: 20
 
     steps:
       - name: Check out repository code
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4
 
       - name: Setup Python
-        uses: actions/setup-python@v4
+        uses: actions/setup-python@v5
         with:
           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
+          python scripts/unsloth_install.py | sh
+          python scripts/cutcrossentropy_install.py | sh
+          pip3 install -r requirements-dev.txt -r requirements-tests.txt
+
+      - name: Ensure axolotl CLI was installed
+        run: |
+          axolotl --help
 
       - name: Run tests
         run: |
-          pytest --ignore=tests/e2e/ tests/
+          pytest -n8 --dist loadfile --ignore=tests/e2e/ --ignore=tests/patched/ tests/
+          pytest tests/patched/
 
-  docker-e2e-tests:
-    if: github.repository_owner == 'OpenAccess-AI-Collective'
+      - name: cleanup pip cache
+        run: |
+          find "$(pip cache dir)/http-v2" -type f -mtime +14 -exec rm {} \;
+
+  pytest-sdist:
+    name: PyTest from Source Dist
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      max-parallel: 1
+      matrix:
+        python_version: ["3.11"]
+        pytorch_version: ["2.4.1", "2.5.1"]
+    timeout-minutes: 20
+
+    steps:
+      - name: Check out repository code
+        uses: actions/checkout@v4
+
+      - name: Setup Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python_version }}
+          cache: 'pip' # caching pip dependencies
+
+      - name: upgrade pip
+        run: |
+          pip3 install --upgrade pip
+          pip3 install --upgrade packaging setuptools wheel
+
+      - name: Install PyTorch
+        run: |
+          pip3 install torch==${{ matrix.pytorch_version }}
+
+      - name: Install dependencies
+        run: |
+          pip3 show torch
+          python3 setup.py sdist
+          pip3 install dist/axolotl*.tar.gz
+          python scripts/unsloth_install.py | sh
+          python scripts/cutcrossentropy_install.py | sh
+          pip3 install -r requirements-dev.txt -r requirements-tests.txt
+
+      - name: Ensure axolotl CLI was installed
+        run: |
+          axolotl --help
+
+      - name: Run tests
+        run: |
+          pytest -n8 --dist loadfile --ignore=tests/e2e/ --ignore=tests/patched/ tests/
+          pytest tests/patched/
+
+      - name: cleanup pip cache
+        run: |
+          find "$(pip cache dir)/http-v2" -type f -mtime +14 -exec rm {} \;
+
+  docker-e2e-tests-1st:
+    if: ${{ ! contains(github.event.commits[0].message, '[skip e2e]') && github.repository_owner == 'axolotl-ai-cloud' }}
     # this job needs to be run on self-hosted GPU runners...
     runs-on: [self-hosted, modal]
-    timeout-minutes: 60
-    needs: [pre-commit, pytest]
+    timeout-minutes: 90
+    needs: [pre-commit, pytest, pytest-sdist]
 
     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
-            python_version: "3.10"
-            pytorch: 2.1.2
-            num_gpus: 1
-          - cuda: 121
-            cuda_version: 12.1.0
+          - cuda: 124
+            cuda_version: 12.4.1
             python_version: "3.11"
-            pytorch: 2.2.2
-            num_gpus: 1
-          - cuda: 121
-            cuda_version: 12.1.0
-            python_version: "3.11"
-            pytorch: 2.3.0
+            pytorch: 2.4.1
             num_gpus: 1
+            axolotl_extras:
     steps:
       - name: Checkout
         uses: actions/checkout@v4
@@ -99,12 +175,59 @@ 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 }}" >> $GITHUB_ENV
+      - name: Run tests job on Modal
+        run: |
+          modal run cicd.tests
+
+  docker-e2e-tests:
+    if: github.repository_owner == 'axolotl-ai-cloud'
+    # this job needs to be run on self-hosted GPU runners...
+    runs-on: [self-hosted, modal]
+    timeout-minutes: 90
+    needs: [pre-commit, pytest, docker-e2e-tests-1st]
+
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - cuda: 121
+            cuda_version: 12.1.1
+            python_version: "3.10"
+            pytorch: 2.3.1
+            num_gpus: 1
+            axolotl_extras: mamba-ssm
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.5.1
+            num_gpus: 1
+            axolotl_extras:
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Install Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.10"
+      - name: Install Modal
+        run: |
+          python -m pip install --upgrade pip
+          pip install modal==0.63.64 jinja2
+      - name: Update env vars
+        run: |
+          echo "BASE_TAG=main-base-py${{ 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
       - name: Run tests job on Modal

.gitignore

@@ -182,3 +182,6 @@ submit.sh
 
 typings/
 out/
+
+# vim
+*.swp

.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:

MANIFEST.in

@@ -0,0 +1,4 @@
+include requirements.txt
+include README.md
+include LICENSE
+recursive-include axolotl *.py

README.md

@@ -1,4 +1,21 @@
-# Axolotl
+<p align="center">
+    <picture>
+        <source media="(prefers-color-scheme: dark)" srcset="image/axolotl_logo_digital_white.svg">
+        <source media="(prefers-color-scheme: light)" srcset="image/axolotl_logo_digital_black.svg">
+        <img alt="Axolotl" src="image/axolotl_logo_digital_black.svg" width="400" height="104" style="max-width: 100%;">
+    </picture>
+</p>
+
+<p align="center">
+    <img src="https://img.shields.io/github/license/axolotl-ai-cloud/axolotl.svg?color=blue" alt="GitHub License">
+    <img src="https://github.com/axolotl-ai-cloud/axolotl/actions/workflows/tests.yml/badge.svg" alt="tests">
+    <a href="https://github.com/axolotl-ai-cloud/axolotl/releases"><img src="https://img.shields.io/github/release/axolotl-ai-cloud/axolotl.svg" alt="Releases"></a>
+    <img src="https://img.shields.io/github/stars/axolotl-ai-cloud/axolotl" alt="GitHub Repo stars">
+</p>
+<p align="center">
+    <img src="https://github.com/axolotl-ai-cloud/axolotl/actions/workflows/tests-nightly.yml/badge.svg" alt="tests-nightly">
+    <img src="https://github.com/axolotl-ai-cloud/axolotl/actions/workflows/multi-gpu-e2e.yml/badge.svg" alt="multigpu-semi-weekly tests">
+</p>
 
 Axolotl is a tool designed to streamline the fine-tuning of various AI models, offering support for multiple configurations and architectures.
 
@@ -7,10 +24,10 @@ 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">
@@ -22,44 +39,51 @@ 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)
-  - [Launching on public clouds via dstack](#launching-on-public-clouds-via-dstack)
-- [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>
-  - [Dataset Pre-Processing](./docs/dataset_preprocessing.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)
+  - [Quickstart ⚡](#quickstart-)
+    - [Usage](#usage)
+  - [Badge ❤🏷️](#badge-️)
+  - [Contributing 🤝](#contributing-)
+  - [Sponsors 🤝❤](#sponsors-)
+  - [Axolotl supports](#axolotl-supports)
+  - [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-)
 
 </td>
 <td>
 
 <div align="center">
-  <img src="image/axolotl.png" alt="axolotl" width="160">
+  <img src="image/axolotl_symbol_digital_white.svg" alt="axolotl" width="160">
   <div>
     <p>
       <b>Axolotl provides a unified repository for fine-tuning <br />a variety of AI models with ease</b>
@@ -67,8 +91,8 @@ 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>
 
@@ -76,6 +100,127 @@ Features:
 </tr>
 </table>
 
+## Quickstart ⚡
+
+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**: *Nvidia* GPU (Ampere architecture or newer for `bf16` and Flash Attention) or *AMD* GPU, Python >=3.10 and PyTorch >=2.3.1.
+
+```bash
+git clone https://github.com/axolotl-ai-cloud/axolotl
+cd axolotl
+
+pip3 install packaging ninja
+pip3 install -e '.[flash-attn,deepspeed]'
+```
+
+### Usage
+```bash
+# preprocess datasets - optional but recommended
+CUDA_VISIBLE_DEVICES="0" python -m axolotl.cli.preprocess examples/openllama-3b/lora.yml
+
+# finetune lora
+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="./outputs/lora-out"
+
+# gradio
+accelerate launch -m axolotl.cli.inference examples/openllama-3b/lora.yml \
+    --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/axolotl-ai-cloud/axolotl/main/examples/openllama-3b/lora.yml
+```
+
+### Axolotl CLI
+
+If you've installed this package using `pip` from source, we now support a new, more
+streamlined CLI using [click](https://click.palletsprojects.com/en/stable/). Rewriting
+the above commands:
+
+```bash
+# preprocess datasets - optional but recommended
+CUDA_VISIBLE_DEVICES="0" axolotl preprocess examples/openllama-3b/lora.yml
+
+# finetune lora
+axolotl train examples/openllama-3b/lora.yml
+
+# inference
+axolotl inference examples/openllama-3b/lora.yml \
+    --lora-model-dir="./outputs/lora-out"
+
+# gradio
+axolotl inference examples/openllama-3b/lora.yml \
+    --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
+axolotl train https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/examples/openllama-3b/lora.yml
+```
+
+We've also added a new command for fetching `examples` and `deepspeed_configs` to your
+local machine. This will come in handy when installing `axolotl` from PyPI.
+
+```bash
+# Fetch example YAML files (stores in "examples/" folder)
+axolotl fetch examples
+
+# Fetch deepspeed config files (stores in "deepspeed_configs/" folder)
+axolotl fetch deepspeed_configs
+
+# Optionally, specify a destination folder
+axolotl fetch examples --dest path/to/folder
+```
+
+## Badge ❤🏷️
+
+Building something cool with Axolotl? Consider adding a badge to your model card.
+
+```markdown
+[<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/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)
+
+## Sponsors 🤝❤
+
+If you love axolotl, consider sponsoring the project by reaching out directly to [wing@axolotl.ai](mailto:wing@axolotl.ai).
+
+---
+
+- [Modal](https://modal.com/) Modal lets you run data/AI jobs in the cloud, by just writing a few lines of Python. Customers use Modal to deploy Gen AI models at large scale, fine-tune LLM models, run protein folding simulations, and much more.
+
+---
+
+## Contributing 🤝
+
+Please read the [contributing guide](./.github/CONTRIBUTING.md)
+
+Bugs? Please check the [open issues](https://github.com/axolotl-ai-cloud/axolotl/issues/bug) else create a new Issue.
+
+PRs are **greatly welcome**!
+
+Please run the quickstart instructions followed by the below to setup env:
+```bash
+pip3 install -r requirements-dev.txt -r requirements-tests.txt
+pre-commit install
+
+# test
+pytest tests/
+
+# optional: run against all files
+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/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>
+
 ## Axolotl supports
 
 |             | fp16/fp32 | lora | qlora | gptq | gptq w/flash attn | flash attn | xformers attn |
@@ -95,46 +240,12 @@ Features:
 | RWKV        | ✅         | ❓    | ❓     | ❓             | ❓                 | ❓          | ❓            |
 | Qwen        | ✅         | ✅    | ✅     | ❓             | ❓                 | ❓          | ❓            |
 | Gemma       | ✅         | ✅    | ✅     | ❓             | ❓                 | ✅          | ❓            |
+| Jamba       | ✅         | ✅    | ✅     | ❓             | ❓                 | ✅          | ❓            |
 
 ✅: supported
 ❌: not supported
 ❓: untested
 
-## Quickstart ⚡
-
-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.
-
-```bash
-git clone https://github.com/OpenAccess-AI-Collective/axolotl
-cd axolotl
-
-pip3 install packaging ninja
-pip3 install -e '.[flash-attn,deepspeed]'
-```
-
-### Usage
-```bash
-# preprocess datasets - optional but recommended
-CUDA_VISIBLE_DEVICES="" python -m axolotl.cli.preprocess examples/openllama-3b/lora.yml
-
-# finetune lora
-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="./outputs/lora-out"
-
-# gradio
-accelerate launch -m axolotl.cli.inference examples/openllama-3b/lora.yml \
-    --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
-```
-
 ## Advanced Setup
 
 ### Environment
@@ -142,7 +253,7 @@ accelerate launch -m axolotl.cli.train https://raw.githubusercontent.com/OpenAcc
 #### Docker
 
   ```bash
-  docker run --gpus '"all"' --rm -it winglian/axolotl:main-latest
+  docker run --gpus '"all"' --rm -it axolotlai/axolotl:main-latest
   ```
 
   Or run on the current files for development:
@@ -161,7 +272,7 @@ accelerate launch -m axolotl.cli.train https://raw.githubusercontent.com/OpenAcc
   A more powerful Docker command to run would be this:
 
   ```bash
-docker run --privileged --gpus '"all"' --shm-size 10g --rm -it --name axolotl --ipc=host --ulimit memlock=-1 --ulimit stack=67108864 --mount type=bind,src="${PWD}",target=/workspace/axolotl -v ${HOME}/.cache/huggingface:/root/.cache/huggingface winglian/axolotl:main-latest
+docker run --privileged --gpus '"all"' --shm-size 10g --rm -it --name axolotl --ipc=host --ulimit memlock=-1 --ulimit stack=67108864 --mount type=bind,src="${PWD}",target=/workspace/axolotl -v ${HOME}/.cache/huggingface:/root/.cache/huggingface axolotlai/axolotl:main-latest
   ```
 
   It additionally:
@@ -193,7 +304,7 @@ docker run --privileged --gpus '"all"' --shm-size 10g --rm -it --name axolotl --
 
 #### Cloud GPU
 
-For cloud GPU providers that support docker images, use [`winglian/axolotl-cloud:main-latest`](https://hub.docker.com/r/winglian/axolotl-cloud/tags)
+For cloud GPU providers that support docker images, use [`axolotlai/axolotl-cloud:main-latest`](https://hub.docker.com/r/axolotlai/axolotl-cloud/tags)
 
 - on Latitude.sh use this [direct link](https://latitude.sh/blueprint/989e0e79-3bf6-41ea-a46b-1f246e309d5c)
 - on JarvisLabs.ai use this [direct link](https://jarvislabs.ai/templates/axolotl)
@@ -302,7 +413,7 @@ Write a job description in YAML as below:
 # dstack.yaml
 type: task
 
-image: winglian/axolotl-cloud:main-20240429-py3.11-cu121-2.2.2
+image: axolotlai/axolotl-cloud:main-latest
 
 env:
   - HUGGING_FACE_HUB_TOKEN
@@ -333,7 +444,7 @@ For further and fine-grained use cases, please refer to the official [dstack doc
 
 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
 
@@ -366,11 +477,10 @@ See [examples](examples) for quick start. It is recommended to duplicate and mod
         - typescript
       type: ... # unimplemented custom format
 
-      # fastchat conversation
-      # See 'conversation' options: https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py
+      # chat_template https://axolotl-ai-cloud.github.io/axolotl/docs/dataset-formats/conversation.html#chat_template
     - path: ...
-      type: sharegpt
-      conversation: chatml # default: vicuna_v1.1
+      type: chat_template
+      chat_template: chatml # defaults to tokenizer's chat_template
 
       # local
     - path: data.jsonl # or json
@@ -498,6 +608,22 @@ 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:
@@ -514,6 +640,26 @@ 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_glu_activation: true
+liger_layer_norm: true
+liger_fused_linear_cross_entropy: true
+```
+
 ### Inference Playground
 
 Axolotl allows you to load your model in an interactive terminal playground for quick experimentation.
@@ -617,86 +763,6 @@ See [this debugging guide](docs/debugging.qmd) for tips on debugging Axolotl, al
 
 ## Need help? 🙋
 
-Join our [Discord server](https://discord.gg/HhrNrHJPRb) where we our community members can help you.
+Join our [Discord server](https://discord.gg/HhrNrHJPRb) where our community members can help you.
 
-Need dedicated support? Please contact us at [✉️wing@openaccessaicollective.org](mailto:wing@openaccessaicollective.org) for dedicated support options.
-
-## Badge ❤🏷️
-
-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/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)
-
-## Community Showcase
-
-Check out some of the projects and models that have been built using Axolotl! Have a model you'd like to add to our Community Showcase? Open a PR with your model.
-
-Open Access AI Collective
-- [Minotaur 13b](https://huggingface.co/openaccess-ai-collective/minotaur-13b-fixed)
-- [Manticore 13b](https://huggingface.co/openaccess-ai-collective/manticore-13b)
-- [Hippogriff 30b](https://huggingface.co/openaccess-ai-collective/hippogriff-30b-chat)
-
-PocketDoc Labs
-- [Dan's PersonalityEngine 13b LoRA](https://huggingface.co/PocketDoc/Dans-PersonalityEngine-13b-LoRA)
-
-## Contributing 🤝
-
-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.
-
-PRs are **greatly welcome**!
-
-Please run the quickstart instructions followed by the below to setup env:
-```bash
-pip3 install -r requirements-dev.txt -r requirements-tests.txt
-pre-commit install
-
-# test
-pytest tests/
-
-# optional: run against all files
-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">
-  <img src="https://contrib.rocks/image?repo=openaccess-ai-collective/axolotl" alt="contributor chart by https://contrib.rocks"/>
-</a>
-
-## Sponsors 🤝❤
-
-OpenAccess AI Collective is run by volunteer contributors such as [winglian](https://github.com/winglian),
-[NanoCode012](https://github.com/NanoCode012), [tmm1](https://github.com/tmm1),
-[mhenrichsen](https://github.com/mhenrichsen), [casper-hansen](https://github.com/casper-hansen),
-[hamelsmu](https://github.com/hamelsmu) and many more who help us accelerate forward by fixing bugs, answering
-community questions and implementing new features. Axolotl needs donations from sponsors for the compute needed to
-run our unit & integration tests, troubleshooting community issues, and providing bounties. If you love axolotl,
-consider sponsoring the project via [GitHub Sponsors](https://github.com/sponsors/OpenAccess-AI-Collective),
-[Ko-fi](https://ko-fi.com/axolotl_ai) or reach out directly to
-[wing@openaccessaicollective.org](mailto:wing@openaccessaicollective.org).
-
----
-
-#### 💎 Diamond Sponsors - [Contact directly](mailto:wing@openaccessaicollective.org)
-
----
-
-#### 🥇 Gold Sponsors - $5000/mo
-
----
-
-#### 🥈 Silver Sponsors - $1000/mo
-
----
-
-#### 🥉 Bronze Sponsors - $500/mo
-
- - [JarvisLabs.ai](https://jarvislabs.ai)
-
----
+Need dedicated support? Please contact us at [✉️wing@axolotl.ai](ailto:wing@axolotl.ai) for dedicated support options.

_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

@@ -1,20 +1,20 @@
-FROM winglian/axolotl-base:{{ BASE_TAG }}
+FROM axolotlai/axolotl-base:{{ BASE_TAG }}
 
 ENV TORCH_CUDA_ARCH_LIST="7.0 7.5 8.0 8.6+PTX"
 ENV AXOLOTL_EXTRAS="{{ AXOLOTL_EXTRAS }}"
 ENV AXOLOTL_ARGS="{{ AXOLOTL_ARGS }}"
 ENV CUDA="{{ CUDA }}"
-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 +22,25 @@ 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 [ "$AXOLOTL_EXTRAS" != "" ] ; then \
-        pip install -e .[deepspeed,flash-attn,mamba-ssm,galore,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
-    else \
-        pip install -e .[deepspeed,flash-attn,mamba-ssm,galore] $AXOLOTL_ARGS; \
+RUN if [ "$NIGHTLY_BUILD" = "true" ] ; then \
+        sed -i 's#^transformers.*#transformers @ git+https://github.com/huggingface/transformers.git@main#' requirements.txt; \
+        sed -i 's#^peft.*#peft @ git+https://github.com/huggingface/peft.git@main#' requirements.txt; \
+        sed -i 's#^accelerate.*#accelerate @ git+https://github.com/huggingface/accelerate.git@main#' requirements.txt; \
+        sed -i 's#^trl.*#trl @ git+https://github.com/huggingface/trl.git@main#' requirements.txt; \
+        sed -i 's#^datasets.*#datasets @ git+https://github.com/huggingface/datasets.git@main#' requirements.txt; \
     fi
 
+RUN if [ "$AXOLOTL_EXTRAS" != "" ] ; then \
+        pip install -e .[deepspeed,flash-attn,optimizers,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
+    else \
+        pip install -e .[deepspeed,flash-attn,optimizers] $AXOLOTL_ARGS; \
+    fi
+
+RUN python scripts/unsloth_install.py | sh
+RUN python scripts/cutcrossentropy_install.py | sh
+
 # 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

@@ -1,6 +1,7 @@
 #!/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 -v --durations=10 -n8 --ignore=tests/e2e/ --ignore=tests/patched/ /workspace/axolotl/tests/
+pytest -v --durations=10 -n8 --dist loadfile /workspace/axolotl/tests/patched/
+pytest -v --durations=10 -n1 --dist loadfile /workspace/axolotl/tests/e2e/patched/ /workspace/axolotl/tests/e2e/integrations/
+pytest -v --durations=10 --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 App, Image
+
+cicd_path = pathlib.Path(__file__).parent.resolve()
+
+template_loader = jinja2.FileSystemLoader(searchpath=cicd_path)
+template_env = jinja2.Environment(
+    loader=template_loader, autoescape=select_autoescape()
+)
+df_template = template_env.get_template("Dockerfile.jinja")
+
+df_args = {
+    "AXOLOTL_EXTRAS": os.environ.get("AXOLOTL_EXTRAS", ""),
+    "AXOLOTL_ARGS": os.environ.get("AXOLOTL_ARGS", ""),
+    "PYTORCH_VERSION": os.environ.get("PYTORCH_VERSION", "2.3.1"),
+    "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")
+)
+
+app = App("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
+
+
+@app.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")
+
+
+@app.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 -v -n2 /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
@@ -8,7 +10,7 @@ import tempfile
 import jinja2
 import modal
 from jinja2 import select_autoescape
-from modal import Image, Stub
+from modal import App, Image
 
 cicd_path = pathlib.Path(__file__).parent.resolve()
 
@@ -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)
@@ -37,6 +40,7 @@ with open(pathlib.Path(temp_dir) / "Dockerfile", "w", encoding="utf-8") as f:
 cicd_image = (
     Image.from_dockerfile(
         pathlib.Path(temp_dir) / "Dockerfile",
+        context_mount=None,
         force_build=True,
         gpu="A10G",
     )
@@ -44,7 +48,7 @@ cicd_image = (
     .pip_install("fastapi==0.110.0", "pydantic==2.6.3")
 )
 
-stub = Stub("Axolotl CI/CD", secrets=[])
+app = App("Axolotl CI/CD", secrets=[])
 
 
 N_GPUS = int(os.environ.get("N_GPUS", 1))
@@ -59,10 +63,10 @@ def run_cmd(cmd: str, run_folder: str):
         exit(exit_code)  # pylint: disable=consider-using-sys-exit
 
 
-@stub.function(
+@app.function(
     image=cicd_image,
     gpu=GPU_CONFIG,
-    timeout=45 * 60,
+    timeout=60 * 60,
     cpu=8.0,
     memory=131072,
 )
@@ -70,6 +74,6 @@ def cicd_pytest():
     run_cmd("./cicd/cicd.sh", "/workspace/axolotl")
 
 
-@stub.local_entrypoint()
+@app.local_entrypoint()
 def main():
     cicd_pytest.remote()

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

@@ -24,15 +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

@@ -20,15 +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

@@ -1,4 +1,4 @@
-# Example config for debugging the sharegpt prompt format
+# Example config for debugging the chat_template prompt format
 base_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0
 model_type: LlamaForCausalLM
 tokenizer_type: LlamaTokenizer
@@ -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

@@ -1,11 +1,10 @@
 ARG BASE_TAG=main-base
-FROM winglian/axolotl-base:$BASE_TAG
+FROM axolotlai/axolotl-base:$BASE_TAG
 
 ARG TORCH_CUDA_ARCH_LIST="7.0 7.5 8.0 8.6+PTX"
 ARG AXOLOTL_EXTRAS=""
 ARG AXOLOTL_ARGS=""
 ARG CUDA="118"
-ENV BNB_CUDA_VERSION=$CUDA
 ARG PYTORCH_VERSION="2.1.2"
 
 ENV PYTORCH_VERSION=$PYTORCH_VERSION
@@ -15,18 +14,20 @@ 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
 
 # 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
 
+RUN python scripts/unsloth_install.py | sh
+RUN python scripts/cutcrossentropy_install.py | sh
+
 # So we can test the Docker image
 RUN pip install pytest
 

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}"
 
@@ -16,7 +16,7 @@ ENV PYTHON_VERSION=$PYTHON_VERSION
 ENV TORCH_CUDA_ARCH_LIST=$TORCH_CUDA_ARCH_LIST
 
 RUN apt-get update \
-    && apt-get install -y wget git build-essential ninja-build git-lfs libaio-dev && rm -rf /var/lib/apt/lists/* \
+    && apt-get install -y wget git build-essential ninja-build git-lfs libaio-dev pkg-config && rm -rf /var/lib/apt/lists/* \
     && wget \
     https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh \
     && mkdir /root/.conda \
@@ -29,7 +29,9 @@ ENV PATH="/root/miniconda3/envs/py${PYTHON_VERSION}/bin:${PATH}"
 WORKDIR /workspace
 
 RUN python3 -m pip install --upgrade pip && pip3 install packaging && \
-    python3 -m pip install --no-cache-dir -U torch==${PYTORCH_VERSION}+cu${CUDA} --extra-index-url https://download.pytorch.org/whl/cu$CUDA
+    python3 -m pip install --no-cache-dir -U torch==${PYTORCH_VERSION}+cu${CUDA} --extra-index-url https://download.pytorch.org/whl/cu$CUDA && \
+    python3 -m pip install --no-cache-dir "causal_conv1d @ git+https://github.com/Dao-AILab/causal-conv1d.git@main" && \
+    python3 -m pip install --no-cache-dir "mamba_ssm @ git+https://github.com/state-spaces/mamba.git@main"
 
 RUN git lfs install --skip-repo && \
     pip3 install awscli && \

docker/Dockerfile-cloud

@@ -1,9 +1,8 @@
 ARG BASE_TAG=main
-FROM winglian/axolotl:$BASE_TAG
+FROM axolotlai/axolotl:$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_HUB_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

@@ -1,9 +1,8 @@
 ARG BASE_TAG=main
-FROM winglian/axolotl:$BASE_TAG
+FROM axolotlai/axolotl:$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_HUB_CACHE="/workspace/data/huggingface-cache/hub"
 ENV HF_HOME="/workspace/data/huggingface-cache/hub"
 ENV HF_HUB_ENABLE_HF_TRANSFER="1"
 

docker/Dockerfile-tests

@@ -1,11 +1,10 @@
 ARG BASE_TAG=main-base
-FROM winglian/axolotl-base:$BASE_TAG
+FROM axolotlai/axolotl-base:$BASE_TAG
 
 ARG TORCH_CUDA_ARCH_LIST="7.0 7.5 8.0 8.6+PTX"
 ARG AXOLOTL_EXTRAS=""
 ARG AXOLOTL_ARGS=""
 ARG CUDA="118"
-ENV BNB_CUDA_VERSION=$CUDA
 ARG PYTORCH_VERSION="2.1.2"
 ARG GITHUB_REF="main"
 
@@ -16,7 +15,7 @@ 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

@@ -83,22 +83,15 @@ 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, 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
-
-    # 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
-    field_human: # Optional[str]. Human key to use for conversation.
-    field_model: # Optional[str]. Assistant key to use for conversation.
-    # Add additional keys from your dataset as input or output roles
-    roles:
-      input: # Optional[List[str]]. These will be masked based on train_on_input
-      output: # Optional[List[str]].
+    revision: # Optional[str] The specific revision of the dataset to use when loading from the Hugging Face Hub. This can be a commit hash, tag, or branch name. If not specified, the latest version will be used. This parameter is ignored for local datasets.
+    trust_remote_code: # Optional[bool] Trust remote code for untrusted source
 
   # Custom user instruction prompt
   - path: repo
@@ -123,10 +116,55 @@ 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
 
+Deduplicates datasets and test_datasets with identical entries.
+dataset_exact_deduplication: true
+
 # A list of one or more datasets to eval the model with.
 # You can use either test_datasets, or val_set_size, but not both.
 test_datasets:
@@ -138,12 +176,21 @@ test_datasets:
     data_files:
       - /workspace/data/eval.jsonl
 
-# use RL training: 'dpo', 'ipo', 'kto_pair'
+# use RL training: 'dpo', 'ipo', 'kto'
 rl:
+# whether to perform weighting if doing DPO training. Boolean.
+dpo_use_weighting:
 
-# 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
@@ -265,8 +312,21 @@ 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
 
@@ -301,7 +361,7 @@ 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)
@@ -349,6 +409,7 @@ lr_div_factor: # Learning rate div factor
 # - adamw_torch_fused
 # - adamw_torch_xla
 # - adamw_apex_fused
+# - adopt_adamw (an EXPERIMENTAL optimizer, only for torch version >= 2.5.1)
 # - adafactor
 # - adamw_anyprecision
 # - sgd

docs/dataset-formats/conversation.qmd

@@ -6,31 +6,8 @@ order: 3
 
 ## sharegpt
 
-conversations where `from` is `human`/`gpt`. (optional: first row with role `system` to override default system prompt)
+IMPORTANT: ShareGPT is deprecated!. Please see `chat_template` section below.
 
-```{.json filename="data.jsonl"}
-{"conversations": [{"from": "...", "value": "..."}]}
-```
-
-Note: `type: sharegpt` opens special configs:
-- `conversation`: enables conversions to many Conversation types. Refer to the 'name' [here](https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py) for options.
-- `roles`: allows you to specify the roles for input and output. This is useful for datasets with custom roles such as `tool` etc to support masking.
-- `field_human`: specify the key to use instead of `human` in the conversation.
-- `field_model`: specify the key to use instead of `gpt` in the conversation.
-
-```yaml
-datasets:
-    path: ...
-    type: sharegpt
-
-    conversation: # Options (see Conversation 'name'): https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py
-    field_human: # Optional[str]. Human key to use for conversation.
-    field_model: # Optional[str]. Assistant key to use for conversation.
-    # Add additional keys from your dataset as input or output roles
-    roles:
-      input: # Optional[List[str]]. These will be masked based on train_on_input
-      output: # Optional[List[str]].
-```
 
 ## pygmalion
 
@@ -38,26 +15,137 @@ datasets:
 {"conversations": [{"role": "...", "value": "..."}]}
 ```
 
-## sharegpt.load_role
 
-conversations where `role` is used instead of `from`
+## 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": "...", "value": "..."}]}
+{"conversations": [{"role": "...", "content": "..."}]}
 ```
 
-## sharegpt.load_guanaco
+See `config.qmd` for full configs and supported templates.
 
-conversations where `from` is `prompter` `assistant` instead of default sharegpt
+### 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": "...", "value": "..."}]}
+{
+  "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}
+  ]
+}
 ```
 
-## sharegpt_jokes
+The configuration would look like:
 
-creates a chat where bot is asked to tell a joke, then explain why the joke is funny
-
-```{.json filename="data.jsonl"}
-{"conversations": [{"title": "...", "text": "...", "explanation": "..."}]}
+```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/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
@@ -185,14 +185,14 @@ style="border-radius: 10px; display: block; margin: auto;" width="560" height="3
 
 ## Debugging With Docker
 
-Using [official Axolotl Docker images](https://hub.docker.com/r/winglian/axolotl/tags) is a great way to debug your code, and is a very popular way to use Axolotl.  Attaching VSCode to Docker takes a few more steps.
+Using [official Axolotl Docker images](https://hub.docker.com/r/axolotlai/axolotl/tags) is a great way to debug your code, and is a very popular way to use Axolotl.  Attaching VSCode to Docker takes a few more steps.
 
 ### Setup
 
 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
 ```
 
@@ -202,11 +202,11 @@ cd axolotl
 Next, run the desired docker image and mount the current directory. Below is a docker command you can run to do this:[^2]
 
 ```bash
-docker run --privileged --gpus '"all"' --shm-size 10g --rm -it --name axolotl --ipc=host --ulimit memlock=-1 --ulimit stack=67108864 --mount type=bind,src="${PWD}",target=/workspace/axolotl -v ${HOME}/.cache/huggingface:/root/.cache/huggingface winglian/axolotl:main-py3.10-cu118-2.0.1
+docker run --privileged --gpus '"all"' --shm-size 10g --rm -it --name axolotl --ipc=host --ulimit memlock=-1 --ulimit stack=67108864 --mount type=bind,src="${PWD}",target=/workspace/axolotl -v ${HOME}/.cache/huggingface:/root/.cache/huggingface axolotlai/axolotl:main-py3.10-cu118-2.0.1
 ```
 
 >[!Tip]
-> To understand which containers are available, see the [Docker section of the README](../README.md#docker) and the [DockerHub repo](https://hub.docker.com/r/winglian/axolotl/tags).  For details of how the Docker containers are built, see axolotl's [Docker CI builds](../.github/workflows/main.yml).
+> To understand which containers are available, see the [Docker section of the README](../README.md#docker) and the [DockerHub repo](https://hub.docker.com/r/axolotlai/axolotl/tags).  For details of how the Docker containers are built, see axolotl's [Docker CI builds](../.github/workflows/main.yml).
 
 You will now be in the container.  Next, perform an editable install of 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/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 +1,47 @@
+---
+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 the correct unsloth and extras from source.
+
+```bash
+python scripts/unsloth_install.py | sh
+```
+
+### 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/colab-notebooks/colab-axolotl-example.ipynb

@@ -2,19 +2,15 @@
  "cells": [
   {
    "cell_type": "markdown",
-   "metadata": {
-    "id": "AKjdG7tbTb-n"
-   },
+   "metadata": {},
    "source": [
-    "# Example notebook for running Axolotl on google colab"
+    "## Setting up"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "metadata": {
-    "id": "RcbNpOgWRcii"
-   },
+   "metadata": {},
    "outputs": [],
    "source": [
     "import torch\n",
@@ -22,83 +18,76 @@
     "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"
-   },
+   "metadata": {},
    "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\"!pip install mlflow==\"2.13.0\""
+    "!pip install axolotl[deepspeed]"
    ]
   },
   {
    "cell_type": "markdown",
-   "metadata": {
-    "id": "BW2MFr7HTjub"
-   },
+   "metadata": {},
    "source": [
-    "## Create an yaml config file"
+    "## Hugging Face login (optional)"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "metadata": {
-    "id": "9pkF2dSoQEUN"
-   },
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from huggingface_hub import notebook_login\n",
+    "notebook_login()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Example configuration"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
    "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",
+    "base_model: NousResearch/Meta-Llama-3.1-8B\n",
     "\n",
     "load_in_8bit: false\n",
     "load_in_4bit: true\n",
     "strict: false\n",
     "\n",
     "datasets:\n",
-    "  - path: mhenrichsen/alpaca_2k_test\n",
+    "  - path: tatsu-lab/alpaca\n",
     "    type: alpaca\n",
-    "dataset_prepared_path:\n",
+    "dataset_prepared_path: last_run_prepared\n",
     "val_set_size: 0.05\n",
-    "output_dir: ./outputs/qlora-out\n",
+    "output_dir: ./outputs/lora-out\n",
+    "\n",
+    "sequence_len: 2048\n",
+    "sample_packing: true\n",
+    "eval_sample_packing: true\n",
+    "pad_to_sequence_len: true\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",
+    "lora_modules_to_save:\n",
+    "  - embed_tokens\n",
+    "  - lm_head\n",
     "\n",
     "wandb_project:\n",
     "wandb_entity:\n",
@@ -106,12 +95,12 @@
     "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",
+    "gradient_accumulation_steps: 2\n",
+    "micro_batch_size: 1\n",
+    "num_epochs: 1\n",
+    "optimizer: paged_adamw_8bit\n",
     "lr_scheduler: cosine\n",
-    "learning_rate: 0.0002\n",
+    "learning_rate: 2e-5\n",
     "\n",
     "train_on_inputs: false\n",
     "group_by_length: false\n",
@@ -122,13 +111,15 @@
     "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",
+    "flash_attention: false\n",
+    "sdp_attention: true\n",
     "\n",
-    "warmup_steps: 10\n",
-    "evals_per_epoch: 4\n",
+    "warmup_steps: 1\n",
+    "max_steps: 25\n",
+    "evals_per_epoch: 1\n",
+    "eval_table_size:\n",
     "saves_per_epoch: 1\n",
     "debug:\n",
     "deepspeed:\n",
@@ -136,9 +127,10 @@
     "fsdp:\n",
     "fsdp_config:\n",
     "special_tokens:\n",
-    "\n",
+    "  pad_token: <|end_of_text|>\n",
     "\"\"\"\n",
     "\n",
+    "\n",
     "# Convert the YAML string to a Python dictionary\n",
     "yaml_dict = yaml.safe_load(yaml_string)\n",
     "\n",
@@ -147,31 +139,124 @@
     "\n",
     "# Write the YAML file\n",
     "with open(file_path, 'w') as file:\n",
-    "    yaml.dump(yaml_dict, file)\n"
+    "    yaml.dump(yaml_dict, file)"
    ]
   },
   {
    "cell_type": "markdown",
-   "metadata": {
-    "id": "bidoj8YLTusD"
-   },
+   "metadata": {},
    "source": [
-    "## Launch the training"
+    "Above we have a configuration file with base LLM model and datasets specified, among many other things. Axolotl can automatically detect whether the specified datasets are on HuggingFace repo or local machine.\n",
+    "\n",
+    "The Axolotl configuration options encompass model and dataset selection, data pre-processing, and training. Let's go through them line by line:\n",
+    "\n",
+    "*   \"base model\": String value, specifies the underlying pre-trained LLM that will be used for finetuning\n",
+    "\n",
+    "Next we have options for model weights quantization. Quantization allows for reduction in occupied memory on GPUs.\n",
+    "\n",
+    "*   \"load_in_8bit\": Boolean value, whether to quantize the model weights into 8-bit integer.\n",
+    "\n",
+    "*   \"load_in_4bit\": Boolean value, whether to quantize the model weights into 4-bit integer.\n",
+    "\n",
+    "*   \"strict\": Boolean value. If false, it allows for overriding established configuration options in the yaml file when executing in command-line interface.\n",
+    "\n",
+    "*   \"datasets\": a list of dicts that contain path and type of data sets as well as other optional configurations where datasets are concerned. Supports multiple datasets.\n",
+    "\n",
+    "*   \"val_set_size\": Either a float value less than one or an integer less than the total size of dataset. Sets the size of validation set from the whole dataset. If float, sets the proportion of the dataset assigned for validation. If integer, sets the direct size of validation set.\n",
+    "\n",
+    "*   \"output_dir\": String value. Path of trained model.\n",
+    "\n",
+    "For data preprocessing:\n",
+    "\n",
+    "*   \"sequence_len\": Integer. Specifies the maximum sequence length of the input. Typically 2048 or less.\n",
+    "\n",
+    "*   \"pad_to_sequence_len\": Boolean. Padding input to maximum sequence length.\n",
+    "\n",
+    "*   \"sample_packing\": Boolean. Specifies whether to use multi-packing with block diagonal attention.\n",
+    "\n",
+    "*   \"special_tokens\": Python dict, optional. Allows users to specify the additional special tokens to be ignored by the tokenizer.\n",
+    "\n",
+    "For LoRA configuration and its hyperparamters:\n",
+    "\n",
+    "*   \"adapter\": String. Either \"lora\" or \"qlora\", depending on user's choice.\n",
+    "\n",
+    "*   \"lora_model_dir\": String, Optional. Path to directory that contains LoRA model, if there is already a trained LoRA model the user would like to use.\n",
+    "\n",
+    "*   \"lora_r\": Integer. Refers to the rank of LoRA decomposition matrices. Higher value will reduce LoRA efficiency. Recommended to be set to 8.\n",
+    "\n",
+    "*   \"lora_alpha\": Integer. Scale the weight matrices by $\\frac{\\text{lora_alpha}}{\\text{lora_r}}$Recommended to be fixed at 16.\n",
+    "\n",
+    "*   \"lora_dropout\": Float that is 1 or less. The dropout probability of a lora layer.\n",
+    "\n",
+    "*   \"lora_target_linear\": Boolean. If true, lora will target all linear modules in the transformers architecture.\n",
+    "\n",
+    "*   \"lora_modules_to_save\": If you added new tokens to the tokenizer, you may need to save some LoRA modules because they need to know the new tokens.\n",
+    "\n",
+    "See [LoRA](https://arxiv.org/abs/2106.09685) for detailed explanation of LoRA implementation.\n",
+    "\n",
+    "For the training configurations:\n",
+    "\n",
+    "*   \"gradient_accumulation_steps\": Integer. The number of steps over which to accumulate gradient for batch training. E.g. if 2, backprop is performed every two steps.\n",
+    "\n",
+    "*   \"micro_batch_size\": Integer. Batch size per gpu / gradient_accumulation_steps\n",
+    "\n",
+    "*   \"num_epochs\": Integer. Number of epochs. One epoch is when training has looped over every batch in the whole data set once.\n",
+    "\n",
+    "*   \"optimizer\": The optimizer to use for the training.\n",
+    "\n",
+    "*   \"learning_rate\": The learning rate.\n",
+    "\n",
+    "*   \"lr_scheduler\": The learning rate scheduler to use for adjusting learning rate during training.\n",
+    "\n",
+    "*   \"train_on_inputs\": Boolean. Whether to ignore or include the user's prompt from the training labels.\n",
+    "\n",
+    "*   \"group_by_length\": Boolean. Whether to group similarly sized data to minimize padding.\n",
+    "\n",
+    "*   \"bf16\": Either \"auto\", \"true\", or \"false\". Whether to use CUDA bf16 floating point format. If set to \"auto\", will automatically apply bf16 should the gpu supports it.\n",
+    "\n",
+    "*   \"fp16\": Optional. Specifies whether to use CUDA fp16. Automatically set to true if \"bf16\" is set to true. Otherwise false.\n",
+    "\n",
+    "*   \"tf32\": Boolean. Whether to use CUDA tf32. Will override bf16.\n",
+    "\n",
+    "*   \"gradient_checkpointing\": Boolean. Whether to use gradient checkpointing https://huggingface.co/docs/transformers/v4.18.0/en/performance#gradient-checkpointing\n",
+    "\n",
+    "*   \"gradient_checkpointing_kwargs\": Python Dict. Fed into the trainer.\n",
+    "\n",
+    "*   \"logging_steps\": Integer. Log training information over every specified number of steps.\n",
+    "\n",
+    "*   \"flash_attention\": Boolean. Whether to use the [flash attention](https://github.com/Dao-AILab/flash-attention) mechanism.\n",
+    "\n",
+    "*   \"sdp_attention\": Boolean. Whether to use the Scaled Dot Product attention mechanism (the attention mechanism in the [original implementation](https://arxiv.org/abs/1706.03762) of transformers.)\n",
+    "\n",
+    "*   \"warmup_steps\": Integer. The number of pre-training steps where a very low learning rate is used.\n",
+    "\n",
+    "*   \"evals_per_epoch\": Integer. Number of evaluations to be performed within one training epoch.\n",
+    "\n",
+    "*   \"saves_per_epoch\": Integer. Number of times the model is saved in one training epoch.\n",
+    "\n",
+    "*   \"weight_decay\": Positive Float. Sets the \"strength\" of weight decay (i.e. setting the coefficient of L2 regularization)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "The above is but a snippet aiming to get users familiarized with the types of streamlined configuration options axolotl provides. For a full list of configuration options, see [here](https://axolotl-ai-cloud.github.io/axolotl/docs/config.html)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Train the model"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "metadata": {
-    "colab": {
-     "base_uri": "https://localhost:8080/"
-    },
-    "id": "ydTI2Jk2RStU",
-    "outputId": "d6d0df17-4b53-439c-c802-22c0456d301b"
-   },
+   "metadata": {},
    "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"
    ]
   },
@@ -179,7 +264,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## Play with inference"
+    "Predict with trained model"
    ]
   },
   {
@@ -188,36 +273,85 @@
    "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"
+    "    --lora_model_dir=\"./outputs/lora-out\" --gradio"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Deeper Dive"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "It is also helpful to gain some familiarity over some of the core inner workings of axolotl"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Configuration Normalization"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Axolotl uses a custom Dict class, called ```DictDefault```\n",
+    "to store configurations specified in the yaml configuration file (into a Python variable named ```cfg```). The definition for this custom Dict can be found in the [utils/dict.py](https://github.com/axolotl-ai-cloud/axolotl/blob/main/src/axolotl/utils/dict.py)\n",
+    "\n",
+    "```DictDefault``` is amended such that calling a missing key from it will result in a ```None``` return type. This is important because if some configuration options aren't specified by the user, the ```None``` type allows Axolotl to perform boolean operations to determine the default settings for missing configurations. For more examples on how this is done, check out [utils/config/__init__.py](https://github.com/axolotl-ai-cloud/axolotl/blob/main/src/axolotl/utils/config/__init__.py)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Loading Models, Tokenizers, and Trainer"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "If we inspect [cli.train.py](https://github.com/axolotl-ai-cloud/axolotl/blob/main/src/axolotl/cli/train.py), we will find that most of the heavy lifting were done by the function ```train()``` which is itself imported from [src/axolotl/train.py](https://github.com/axolotl-ai-cloud/axolotl/blob/main/src/axolotl/train.py).\n",
+    "\n",
+    "```train()``` takes care of loading the appropriate tokenizer and pre-trained model through ```load_model()``` and ```load_tokenizer()``` from [src/axolotl/utils/models.py](https://github.com/axolotl-ai-cloud/axolotl/blob/main/src/axolotl/utils/models.py) respectively.\n",
+    "\n",
+    "```load_tokenizer()``` loads in the appropriate tokenizer given the desired model, as well as chat templates.\n",
+    "\n",
+    "```ModelLoader``` class follows after tokenizer has been selected. It will automatically discern the base model type, load in the desired model, as well as applying model-appropriate attention mechanism modifications (e.g. flash attention). Depending on which base model the user chooses in the configuration, ```ModelLoader``` will utilize the corresponding \"attention hijacking\" script. For example, if the user specified the base model to be ```NousResearch/Meta-Llama-3.1-8B```, which is of llama type, and set ```flash_attn``` to ```True```, ```ModelLoader``` will load in [llama_attn_hijack_flash.py](https://github.com/axolotl-ai-cloud/axolotl/blob/main/src/axolotl/monkeypatch/llama_attn_hijack_flash.py). For a list of supported attention hijacking, please refer to the directory [/src/axolotl/monkeypatch/](https://github.com/axolotl-ai-cloud/axolotl/tree/main/src/axolotl/monkeypatch)\n",
+    "\n",
+    "Another important operation encompassed in ```train()``` is setting up the training that takes into account of user-specified traning configurations (e.g. num_epochs, optimizer) through the use of ```setup_trainer()``` from [/src/axolotl/utils/trainer.py](https://github.com/axolotl-ai-cloud/axolotl/blob/main/src/axolotl/utils/trainer.py), which in turn relies on modules from [/src/axolotl/core/trainer_builder.py](https://github.com/axolotl-ai-cloud/axolotl/blob/main/src/axolotl/core/trainer_builder.py).\n",
+    "```trainer_builder.py``` provides a list of trainer object options bespoke for the task type (Causal or Reinforcement learning ('dpo', 'ipo', 'kto') )"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Monkey patch\n",
+    "\n",
+    "The [Monkey patch directory](https://github.com/axolotl-ai-cloud/axolotl/tree/main/src/axolotl/monkeypatch) is where model architecture/optimization patching scripts are stored (these are modifications that are not implemented in the official releases, hence the name monkey patch). It includes attention jacking, ReLoRA, and unsloth optimization."
    ]
   }
  ],
  "metadata": {
-  "accelerator": "GPU",
-  "colab": {
-   "gpuType": "T4",
-   "provenance": []
-  },
   "kernelspec": {
-   "display_name": "Python 3 (ipykernel)",
+   "display_name": "Python 3",
    "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"
+   "version": "3.9.6"
   }
  },
  "nbformat": 4,
- "nbformat_minor": 4
+ "nbformat_minor": 2
 }

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_glu_activation: 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/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/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_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/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/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_glu_activation: 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

@@ -1,6 +1,4 @@
-base_model: meta-llama/Meta-Llama-3-8B
-model_type: LlamaForCausalLM
-tokenizer_type: AutoTokenizer
+base_model: NousResearch/Meta-Llama-3.1-8B
 
 load_in_8bit: false
 load_in_4bit: false

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

@@ -1,4 +1,4 @@
-base_model: meta-llama/Meta-Llama-3-8B-Instruct
+base_model: NousResearch/Meta-Llama-3-8B-Instruct
 model_type: LlamaForCausalLM
 tokenizer_type: AutoTokenizer
 
@@ -10,7 +10,6 @@ chat_template: llama3
 datasets:
   - path: fozziethebeat/alpaca_messages_2k_test
     type: chat_template
-    chat_template: llama3
     field_messages: messages
     message_field_role: role
     message_field_content: content
@@ -74,3 +73,5 @@ deepspeed:
 weight_decay: 0.0
 fsdp:
 fsdp_config:
+special_tokens:
+   pad_token: <|end_of_text|>

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

@@ -0,0 +1,95 @@
+base_model: meta-llama/Llama-3.2-1B
+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
+  - 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_exact_deduplication: true
+dataset_prepared_path:
+val_set_size: 0
+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/lora-1b-deduplicate-sft.yml

@@ -0,0 +1,76 @@
+base_model: meta-llama/Llama-3.2-1B
+model_type: LlamaForCausalLM
+tokenizer_type: AutoTokenizer
+
+load_in_8bit: true
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: mhenrichsen/alpaca_2k_test
+    type: alpaca
+  - path: mhenrichsen/alpaca_2k_test
+    type: alpaca
+dataset_prepared_path:
+val_set_size: 0.0
+output_dir: ./outputs/lora-out
+
+dataset_exact_deduplication: true
+test_value: true
+
+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/lora-8b.yml

@@ -1,4 +1,4 @@
-base_model: meta-llama/Meta-Llama-3-8B
+base_model: NousResearch/Meta-Llama-3-8B
 model_type: LlamaForCausalLM
 tokenizer_type: AutoTokenizer
 
@@ -15,6 +15,7 @@ output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true
+eval_sample_packing: false
 pad_to_sequence_len: true
 
 adapter: lora

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.yml

@@ -1,4 +1,4 @@
-base_model: meta-llama/Meta-Llama-3-8B
+base_model: NousResearch/Meta-Llama-3-8B
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
 

examples/mistral/mistral-dpo-qlora.yml

@@ -0,0 +1,93 @@
+#Note that we are switching from the regular chat template to chatml.
+#If you experience problems with the special tokens, training for more epochs can help.
+#After training, merge the model before inference otherwise you might
+#face problems with the special tokens.
+
+base_model: mistralai/Mistral-7B-Instruct-v0.2
+model_type: MistralForCausalLM
+tokenizer_type: LlamaTokenizer
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+chat_template: chatml
+rl: dpo
+datasets:
+  - path: olivermolenschot/alpaca_messages_dpo_test
+    type: chat_template.default
+    field_messages: conversation
+    field_chosen: chosen
+    field_rejected: rejected
+    message_field_role: role
+    message_field_content: content
+
+dataset_prepared_path:
+val_set_size: 0.05
+output_dir: ./outputs/dpo-qlora
+
+sequence_len: 2048
+sample_packing: false
+pad_to_sequence_len: true
+
+adapter: qlora
+lora_model_dir:
+lora_r: 8
+lora_alpha: 16
+lora_dropout: 0.2
+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
+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: 16
+num_epochs: 6
+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: false
+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:
+  bos_token: "<|im_start|>"
+  eos_token: "<|im_end|>"

examples/phi/lora-3.5.yaml

@@ -0,0 +1,75 @@
+base_model: microsoft/Phi-3.5-mini-instruct
+model_type: AutoModelForCausalLM
+tokenizer_type: AutoTokenizer
+
+load_in_8bit: true
+load_in_4bit: false
+strict: false
+
+chat_template: phi_3
+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: 4
+num_epochs: 2
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bfloat16: true
+bf16: true
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+s2_attention:
+
+warmup_steps: 10
+evals_per_epoch: 4
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 4
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:

examples/qwen2/dpo.yaml

@@ -0,0 +1,67 @@
+base_model: Qwen/Qwen2.5-0.5B
+
+strict: false
+
+chat_template: qwen_25
+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: 0.0
+output_dir: ./outputs/dpo-out
+
+sequence_len: 2048
+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: 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:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:

examples/qwen2/qlora-fsdp.yaml

@@ -72,4 +72,5 @@ fsdp_config:
   fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
   fsdp_transformer_layer_cls_to_wrap: Qwen2DecoderLayer
   fsdp_state_dict_type: FULL_STATE_DICT
+  fsdp_sharding_strategy: FULL_SHARD
 special_tokens:

examples/tiny-llama/lora-mps.yml

@@ -1,4 +1,4 @@
-base_model: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
+base_model: TinyLlama/TinyLlama_v1.1
 model_type: LlamaForCausalLM
 tokenizer_type: LlamaTokenizer
 

examples/tiny-llama/lora.yml

@@ -1,6 +1,5 @@
-base_model: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
-model_type: LlamaForCausalLM
-tokenizer_type: LlamaTokenizer
+base_model: TinyLlama/TinyLlama_v1.1
+tokenizer_type: AutoTokenizer
 
 load_in_8bit: true
 load_in_4bit: false

examples/tiny-llama/pretrain.yml

@@ -9,9 +9,9 @@ strict: false
 
 max_steps: 200
 pretraining_dataset:
-  path: c4
-  name: en
-  type: pretrain
+  - path: allenai/c4
+    name: en
+    type: pretrain
 dataset_prepared_path:
 val_set_size: 0.0
 output_dir: ./outputs/model-out

examples/tiny-llama/qlora.yml

@@ -1,4 +1,4 @@
-base_model: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
+base_model: TinyLlama/TinyLlama_v1.1
 model_type: LlamaForCausalLM
 tokenizer_type: LlamaTokenizer
 

image/axolotl_logo_digital_black.svg

@@ -0,0 +1,19 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<svg xmlns="http://www.w3.org/2000/svg" version="1.1" viewBox="0 0 1113 283.5">
+    <path fill="#141310" d="M435,234.3l-12.1-48.8h-54.4l-12.1,48.8h-24.7l48.2-185.1h31.6l47.9,185.1h-24.5ZM417.7,164.9l-13.8-55.6c-2.7-10.7-4.8-19.7-6.3-26.9-.9-4.2-1.5-7.5-2-9.9-.5,2.5-1.2,5.8-2,9.9-1.5,7.1-3.6,16.1-6.3,26.7l-13.8,55.9h44.3Z"/>
+    <path fill="#141310" d="M568.2,234.3l-29.9-45.6c-1.2-1.9-2.4-4.1-3.5-6.5-.8-1.7-1.5-3.3-2.1-4.5-.6,1.3-1.4,2.8-2.3,4.5-1.3,2.4-2.6,4.6-4,6.5l-29.9,45.6h-28.5l49.6-71.9-46.5-67.9h28.5l27.6,43.1c1.2,1.9,2.3,3.9,3.4,6.1.7,1.4,1.4,2.7,1.9,3.8.5-1.1,1.1-2.4,1.8-3.8,1.1-2.2,2.2-4.2,3.4-6.1l27.6-43.1h28.5l-46.5,68.2,49.3,71.7h-28.5Z"/>
+    <path fill="#141310" d="M658.6,236.3c-16.7,0-30.2-5-40.1-14.8-9.9-9.8-14.9-23.7-14.9-41.3v-31.7c0-17.7,5-31.7,14.8-41.4,9.8-9.7,23.4-14.7,40.3-14.7s30.4,4.9,40.3,14.7c9.8,9.7,14.8,23.7,14.8,41.4v31.7c0,17.6-5,31.5-14.9,41.3-9.9,9.8-23.4,14.8-40.1,14.8ZM658.6,114.1c-9.5,0-17.1,2.7-22.6,8.1-5.5,5.4-8.3,13.4-8.3,23.8v36.7c0,10.5,2.8,18.5,8.3,23.8,5.5,5.4,13.1,8.1,22.6,8.1s17.3-2.7,22.7-8.1c5.4-5.4,8.2-13.4,8.2-23.9v-36.7c0-10.5-2.8-18.5-8.2-23.9-5.4-5.4-13.1-8.1-22.7-8.1Z"/>
+    <path fill="#141310" d="M860.6,236.3c-16.7,0-30.2-5-40.1-14.8-9.9-9.8-14.9-23.7-14.9-41.3v-31.7c0-17.7,5-31.7,14.8-41.4,9.8-9.7,23.4-14.7,40.3-14.7s30.4,4.9,40.3,14.7c9.8,9.7,14.8,23.7,14.8,41.4v31.7c0,17.6-5,31.5-14.9,41.3-9.9,9.8-23.4,14.8-40.1,14.8ZM860.6,114.1c-9.5,0-17.1,2.7-22.6,8.1-5.5,5.4-8.3,13.4-8.3,23.8v36.7c0,10.5,2.8,18.5,8.3,23.8,5.5,5.4,13.1,8.1,22.6,8.1s17.3-2.7,22.7-8.1c5.4-5.4,8.2-13.4,8.2-23.9v-36.7c0-10.5-2.8-18.5-8.2-23.9-5.4-5.4-13.1-8.1-22.7-8.1Z"/>
+    <path fill="#141310" d="M773.9,234c-18,0-32.6-14.6-32.6-32.6V48.8h24.1v152.6c0,4.7,3.8,8.5,8.5,8.5h16.8v24.1h-16.8Z"/>
+    <path fill="#141310" d="M1036.2,234.3V81.4c0-4.7-3.8-8.5-8.5-8.5h-16.8v-24.1h16.8c18,0,32.6,14.6,32.6,32.6v152.9h-24.1Z"/>
+    <path fill="#141310" d="M978.6,234.3c-18,0-32.6-14.6-32.6-32.6v-85.1h-20.3v-22.1h20.3v-45.3h24.1v45.3h30.2v22.1h-30.2v85.1c0,4.7,3.8,8.5,8.5,8.5h21.7v24.1h-21.7Z"/>
+    <path fill="#141310" d="M51.5,49h12.2v-20.6h-12.2c-16,0-29,13-29,29v32.8h20.6v-32.8c0-4.7,3.8-8.4,8.4-8.4Z"/>
+    <path fill="#141310" d="M92.8,49h12.2v-20.6h-12.2c-16,0-29,13-29,29v12.2h20.6v-12.2c0-4.7,3.8-8.4,8.4-8.4Z"/>
+    <path fill="#141310" d="M249.3,57.4c0-16-13-29-29-29h-12.2v20.6h12.2c4.7,0,8.4,3.8,8.4,8.4v32.8h20.6v-32.8Z"/>
+    <path fill="#141310" d="M187.4,90.2v-20.6h-103.1v20.6h-41.2v20.6h-20.6v41.2c0,11.4,9.2,20.6,20.6,20.6h185.5c11.4,0,20.6-9.2,20.6-20.6v-41.2h-20.6v-20.6h-41.2ZM166.8,141.7c0-5.7-4.6-10.3-10.3-10.3s-10.3,4.6-10.3,10.3v10.3h-20.6v-20.6c0-11.4,9.2-20.6,20.6-20.6s20.6,9.2,20.6,20.6v10.3ZM228.7,141.7c0-5.7-4.6-10.3-10.3-10.3s-10.3,4.6-10.3,10.3v10.3h-20.6v-20.6c0-11.4,9.2-20.6,20.6-20.6s20.6,9.2,20.6,20.6v10.3Z"/>
+    <path fill="#141310" d="M208,57.4c0-16-13-29-29-29h-12.2v20.6h12.2c4.7,0,8.4,3.8,8.4,8.4v12.2h20.6v-12.2Z"/>
+    <rect fill="#141310" x="22.5" y="234.5" width="41.2" height="20.6"/>
+    <rect fill="#141310" x="84.3" y="234.5" width="164.9" height="20.6"/>
+    <rect fill="#141310" x="208" y="193.3" width="41.2" height="20.6"/>
+    <rect fill="#141310" x="22.5" y="193.3" width="164.9" height="20.6"/>
+</svg>

image/axolotl_logo_digital_white.svg

@@ -0,0 +1,11 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<svg xmlns="http://www.w3.org/2000/svg" version="1.1" viewBox="0 0 1113 283.5">
+    <path fill="#fff" d="M462.9,234.2l-12.1-48.8h-54.4l-12.1,48.8h-24.7l48.2-185h31.6l47.9,185h-24.4ZM445.7,164.8l-13.8-55.6c-2.7-10.7-4.8-19.7-6.3-26.9-.9-4.2-1.5-7.5-2-9.9-.5,2.5-1.2,5.8-2,9.9-1.5,7.1-3.6,16.1-6.3,26.7l-13.8,55.9h44.3Z"/>
+    <path fill="#fff" d="M596.1,234.2l-29.9-45.6c-1.2-1.9-2.4-4.1-3.5-6.5-.8-1.7-1.5-3.3-2.1-4.5-.6,1.3-1.4,2.8-2.3,4.5-1.3,2.4-2.6,4.6-4,6.5l-29.9,45.6h-28.5l49.5-71.9-46.5-67.9h28.5l27.6,43.1c1.2,1.9,2.3,3.9,3.4,6.1.7,1.4,1.3,2.7,1.9,3.8.5-1.1,1.1-2.4,1.8-3.8,1.1-2.2,2.2-4.2,3.4-6.1l27.6-43.1h28.5l-46.5,68.1,49.3,71.6h-28.5Z"/>
+    <path fill="#fff" d="M686.4,236.2c-16.7,0-30.2-5-40.1-14.8-9.9-9.8-14.9-23.7-14.9-41.3v-31.7c0-17.7,5-31.6,14.8-41.4,9.8-9.7,23.4-14.7,40.2-14.7s30.4,4.9,40.2,14.7c9.8,9.7,14.8,23.7,14.8,41.4v31.7c0,17.6-5,31.4-14.9,41.3-9.9,9.8-23.4,14.8-40.1,14.8ZM686.4,114.1c-9.5,0-17.1,2.7-22.6,8.1-5.5,5.4-8.3,13.4-8.3,23.8v36.7c0,10.5,2.8,18.5,8.3,23.8,5.5,5.4,13.1,8.1,22.6,8.1s17.3-2.7,22.7-8.1c5.4-5.4,8.2-13.4,8.2-23.8v-36.7c0-10.5-2.8-18.5-8.2-23.8-5.4-5.4-13.1-8.1-22.7-8.1Z"/>
+    <path fill="#fff" d="M888.3,236.2c-16.7,0-30.2-5-40.1-14.8-9.9-9.8-14.9-23.7-14.9-41.3v-31.7c0-17.7,5-31.6,14.8-41.4,9.8-9.7,23.4-14.7,40.2-14.7s30.4,4.9,40.2,14.7c9.8,9.7,14.8,23.7,14.8,41.4v31.7c0,17.6-5,31.4-14.9,41.3-9.9,9.8-23.4,14.8-40.1,14.8ZM888.3,114.1c-9.5,0-17.1,2.7-22.6,8.1-5.5,5.4-8.3,13.4-8.3,23.8v36.7c0,10.5,2.8,18.5,8.3,23.8,5.5,5.4,13.1,8.1,22.6,8.1s17.3-2.7,22.7-8.1c5.4-5.4,8.2-13.4,8.2-23.8v-36.7c0-10.5-2.8-18.5-8.2-23.8-5.4-5.4-13.1-8.1-22.7-8.1Z"/>
+    <path fill="#fff" d="M801.7,234c-18,0-32.6-14.6-32.6-32.6V48.8h24.1v152.5c0,4.7,3.8,8.5,8.5,8.5h16.7v24.1h-16.7Z"/>
+    <path fill="#fff" d="M1063.8,234.2V81.4c0-4.7-3.8-8.5-8.5-8.5h-16.7v-24.1h16.7c18,0,32.6,14.6,32.6,32.6v152.8h-24.1Z"/>
+    <path fill="#fff" d="M1006.2,234.2c-18,0-32.6-14.6-32.6-32.6v-85h-20.3v-22.1h20.3v-45.2h24.1v45.2h30.2v22.1h-30.2v85c0,4.7,3.8,8.5,8.5,8.5h21.7v24.1h-21.7Z"/>
+    <path fill="#fff" d="M160.2,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM222,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM222,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM160.2,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM160.2,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM222,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM222,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM160.2,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM160.2,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM222,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM222,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM160.2,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM277.3,57.4c0-23.8-19.3-43.1-43.1-43.1h-12.2c-3.9,0-7.6,1.6-10.2,4.4-5.9-2.9-12.3-4.4-18.9-4.4h-12.2c-7.7,0-14.1,6.3-14.1,14.1v20.6c0,2.4.6,4.6,1.6,6.6h-37c1-2,1.6-4.2,1.6-6.6v-20.6c0-7.7-6.3-14.1-14.1-14.1h-12.2c-6.5,0-13,1.5-18.9,4.4-2.6-2.8-6.3-4.4-10.2-4.4h-12.2c-23.8,0-43.1,19.3-43.1,43.1v32.8c0,4.1,1.7,7.7,4.5,10.3-2.8,2.6-4.5,6.2-4.5,10.3v41.2c0,11,5.2,20.8,13.2,27.2-7.3.4-13.2,6.6-13.2,14v20.6c0,4.1,1.7,7.7,4.5,10.3-2.8,2.6-4.5,6.2-4.5,10.3v20.6c0,7.7,6.3,14.1,14.1,14.1h41.2c4.1,0,7.7-1.7,10.3-4.5,2.6,2.8,6.2,4.5,10.3,4.5h164.9c7.7,0,14.1-6.3,14.1-14.1v-20.6c0-4.1-1.7-7.7-4.5-10.3,2.8-2.6,4.5-6.2,4.5-10.3v-20.6c0-7.5-5.8-13.6-13.2-14,8-6.4,13.2-16.2,13.2-27.2v-41.2c0-4.1-1.7-7.7-4.5-10.3,2.8-2.6,4.5-6.2,4.5-10.3v-32.8ZM77.8,255.1h-41.2v-20.6h41.2v20.6ZM36.5,213.9v-20.6h164.9v20.6H36.5ZM263.3,255.1H98.4v-20.6h164.9v20.6ZM263.3,213.9h-41.2v-20.6h41.2v20.6ZM263.3,90.2h-20.6v20.6h20.6v41.2c0,11.4-9.2,20.6-20.6,20.6H57.2c-11.4,0-20.6-9.2-20.6-20.6v-41.2h20.6v-20.6h-20.6v-32.8c0-16,13-29,29-29h12.2v20.6h-12.2c-4.7,0-8.4,3.8-8.4,8.4v32.8h41.2v-20.6h-20.6v-12.2c0-16,13-29,29-29h12.2v20.6h-12.2c-4.7,0-8.4,3.8-8.4,8.4v12.2h103.1v-12.2c0-4.7-3.8-8.4-8.4-8.4h-12.2v-20.6h12.2c16,0,29,13,29,29v12.2h-20.6v20.6h41.2v-32.8c0-4.7-3.8-8.4-8.4-8.4h-12.2v-20.6h12.2c16,0,29,13,29,29v32.8ZM201.4,152h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6s-20.6,9.2-20.6,20.6v20.6ZM160.2,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM160.2,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM222,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM222,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM160.2,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM160.2,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM222,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM222,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM160.2,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM160.2,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM222,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM222,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM160.2,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6Z"/>
+</svg>

image/axolotl_symbol_digital_black.svg

@@ -0,0 +1,26 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<svg xmlns="http://www.w3.org/2000/svg" version="1.1" viewBox="0 0 283.5 283.5">
+  <defs>
+    <style>
+      .cls-1 {
+        fill: #141310;
+      }
+    </style>
+  </defs>
+  <!-- Generator: Adobe Illustrator 28.7.1, SVG Export Plug-In . SVG Version: 1.2.0 Build 142)  -->
+  <g>
+    <g id="Layer_1">
+      <g>
+        <path class="cls-1" d="M46.9,37.4h13.7V14.2h-13.7c-18,0-32.7,14.6-32.7,32.7v36.9h23.2v-36.9c0-5.2,4.2-9.5,9.5-9.5Z"/>
+        <path class="cls-1" d="M93.2,37.4h13.7V14.2h-13.7c-18,0-32.7,14.6-32.7,32.7v13.7h23.2v-13.7c0-5.2,4.2-9.5,9.5-9.5Z"/>
+        <path class="cls-1" d="M269.3,46.9c0-18-14.6-32.7-32.7-32.7h-13.7v23.2h13.7c5.2,0,9.5,4.2,9.5,9.5v36.9h23.2v-36.9Z"/>
+        <path class="cls-1" d="M199.7,83.8v-23.2h-116v23.2h-46.4v23.2H14.2v46.4c0,12.8,10.4,23.2,23.2,23.2h208.7c12.8,0,23.2-10.4,23.2-23.2v-46.4h-23.2v-23.2h-46.4ZM176.5,141.7c0-6.4-5.2-11.6-11.6-11.6s-11.6,5.2-11.6,11.6v11.6h-23.2v-23.2c0-12.8,10.4-23.2,23.2-23.2s23.2,10.4,23.2,23.2v11.6ZM246.1,141.7c0-6.4-5.2-11.6-11.6-11.6s-11.6,5.2-11.6,11.6v11.6h-23.2v-23.2c0-12.8,10.4-23.2,23.2-23.2s23.2,10.4,23.2,23.2v11.6Z"/>
+        <path class="cls-1" d="M222.9,46.9c0-18-14.6-32.7-32.7-32.7h-13.7v23.2h13.7c5.2,0,9.5,4.2,9.5,9.5v13.7h23.2v-13.7Z"/>
+        <rect class="cls-1" x="14.2" y="246.1" width="46.4" height="23.2"/>
+        <rect class="cls-1" x="83.8" y="246.1" width="185.5" height="23.2"/>
+        <rect class="cls-1" x="222.9" y="199.7" width="46.4" height="23.2"/>
+        <rect class="cls-1" x="14.2" y="199.7" width="185.5" height="23.2"/>
+      </g>
+    </g>
+  </g>
+</svg>

image/axolotl_symbol_digital_white.svg

@@ -0,0 +1,16 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<svg xmlns="http://www.w3.org/2000/svg" version="1.1" viewBox="0 0 283.5 283.5">
+  <defs>
+    <style>
+      .cls-1 {
+        fill: #fff;
+      }
+    </style>
+  </defs>
+  <!-- Generator: Adobe Illustrator 28.7.1, SVG Export Plug-In . SVG Version: 1.2.0 Build 142)  -->
+  <g>
+    <g id="Layer_1">
+      <path class="cls-1" d="M152,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM214,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM214,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM152,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM152,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM214,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM214,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM152,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM152,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM214,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM214,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM152,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM269.3,57.3c0-23.8-19.4-43.1-43.1-43.1h-12.2c-3.9,0-7.6,1.6-10.2,4.4-5.9-2.9-12.3-4.4-18.9-4.4h-12.2c-7.8,0-14.1,6.3-14.1,14.1v20.6c0,2.4.6,4.6,1.6,6.6h-37c1-2,1.6-4.2,1.6-6.6v-20.6c0-7.8-6.3-14.1-14.1-14.1h-12.2c-6.6,0-13,1.5-18.9,4.4-2.6-2.8-6.3-4.4-10.2-4.4h-12.2c-23.8,0-43.1,19.4-43.1,43.1v32.8c0,4.1,1.7,7.7,4.5,10.3-2.8,2.6-4.5,6.2-4.5,10.3v41.3c0,11,5.2,20.9,13.2,27.2-7.4.4-13.2,6.6-13.2,14v20.6c0,4.1,1.7,7.7,4.5,10.3-2.8,2.6-4.5,6.2-4.5,10.3v20.6c0,7.8,6.3,14.1,14.1,14.1h41.3c4.1,0,7.7-1.7,10.3-4.5,2.6,2.8,6.2,4.5,10.3,4.5h165.1c7.8,0,14.1-6.3,14.1-14.1v-20.6c0-4.1-1.7-7.7-4.5-10.3,2.8-2.6,4.5-6.2,4.5-10.3v-20.6c0-7.5-5.9-13.6-13.2-14,8-6.4,13.2-16.2,13.2-27.2v-41.3c0-4.1-1.7-7.7-4.5-10.3,2.8-2.6,4.5-6.2,4.5-10.3v-32.8ZM69.5,255.2H28.2v-20.6h41.3v20.6ZM28.2,214v-20.6h165.1v20.6H28.2ZM255.2,255.2H90.1v-20.6h165.1v20.6ZM255.2,214h-41.3v-20.6h41.3v20.6ZM255.2,90.1h-20.6v20.6h20.6v41.3c0,11.4-9.2,20.6-20.6,20.6H48.9c-11.4,0-20.6-9.2-20.6-20.6v-41.3h20.6v-20.6h-20.6v-32.8c0-16.1,13-29.1,29.1-29.1h12.2v20.6h-12.2c-4.7,0-8.4,3.8-8.4,8.4v32.8h41.3v-20.6h-20.6v-12.2c0-16.1,13-29.1,29.1-29.1h12.2v20.6h-12.2c-4.7,0-8.4,3.8-8.4,8.4v12.2h103.2v-12.2c0-4.7-3.8-8.4-8.4-8.4h-12.2v-20.6h12.2c16.1,0,29.1,13,29.1,29.1v12.2h-20.6v20.6h41.3v-32.8c0-4.7-3.8-8.4-8.4-8.4h-12.2v-20.6h12.2c16.1,0,29.1,13,29.1,29.1v32.8ZM193.3,152h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6s-20.6,9.2-20.6,20.6v20.6ZM152,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM152,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM214,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM214,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM152,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM152,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM214,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM214,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM152,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM152,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM214,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM214,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6ZM152,110.8c-11.4,0-20.6,9.2-20.6,20.6v20.6h20.6v-10.3c0-5.7,4.6-10.3,10.3-10.3s10.3,4.6,10.3,10.3v-10.3c0-11.4-9.2-20.6-20.6-20.6Z"/>
+    </g>
+  </g>
+</svg>

image/axolotl_wordmark_digital_black.svg

@@ -0,0 +1,17 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<svg xmlns="http://www.w3.org/2000/svg" version="1.1" viewBox="0 0 765.4 212.6">
+  <!-- Generator: Adobe Illustrator 28.7.1, SVG Export Plug-In . SVG Version: 1.2.0 Build 142)  -->
+  <g>
+    <g id="Layer_1">
+      <g>
+        <path d="M121.6,198.1l-12.1-48.8h-54.4l-12.1,48.8h-24.7L66.6,12.9h31.6l47.9,185.1h-24.5ZM104.4,128.6l-13.8-55.6c-2.7-10.7-4.8-19.7-6.3-26.9-.9-4.2-1.5-7.5-2-9.9-.5,2.5-1.2,5.8-2,9.9-1.5,7.1-3.6,16.1-6.3,26.7l-13.8,55.9h44.3Z"/>
+        <path d="M254.9,198.1l-29.9-45.6c-1.2-1.9-2.4-4.1-3.5-6.5-.8-1.7-1.5-3.3-2.1-4.5-.6,1.3-1.4,2.8-2.3,4.5-1.3,2.4-2.6,4.6-4,6.5l-29.9,45.6h-28.5l49.6-71.9-46.5-67.9h28.5l27.6,43.1c1.2,1.9,2.3,3.9,3.4,6.1.7,1.4,1.4,2.7,1.9,3.8.5-1.1,1.1-2.4,1.8-3.8,1.1-2.2,2.2-4.2,3.4-6.1l27.6-43.1h28.5l-46.5,68.2,49.3,71.7h-28.5Z"/>
+        <path d="M345.2,200.1c-16.7,0-30.2-5-40.1-14.8-9.9-9.8-14.9-23.7-14.9-41.3v-31.7c0-17.7,5-31.7,14.8-41.4,9.8-9.7,23.4-14.7,40.3-14.7s30.4,4.9,40.3,14.7c9.8,9.7,14.8,23.7,14.8,41.4v31.7c0,17.6-5,31.5-14.9,41.3-9.9,9.8-23.4,14.8-40.1,14.8ZM345.2,77.8c-9.5,0-17.1,2.7-22.6,8.1-5.5,5.4-8.3,13.4-8.3,23.8v36.7c0,10.5,2.8,18.5,8.3,23.8,5.5,5.4,13.1,8.1,22.6,8.1s17.3-2.7,22.7-8.1c5.4-5.4,8.2-13.4,8.2-23.9v-36.7c0-10.5-2.8-18.5-8.2-23.9-5.4-5.4-13.1-8.1-22.7-8.1Z"/>
+        <path d="M547.3,200.1c-16.7,0-30.2-5-40.1-14.8-9.9-9.8-14.9-23.7-14.9-41.3v-31.7c0-17.7,5-31.7,14.8-41.4,9.8-9.7,23.4-14.7,40.3-14.7s30.4,4.9,40.3,14.7c9.8,9.7,14.8,23.7,14.8,41.4v31.7c0,17.6-5,31.5-14.9,41.3-9.9,9.8-23.4,14.8-40.1,14.8ZM547.3,77.8c-9.5,0-17.1,2.7-22.6,8.1-5.5,5.4-8.3,13.4-8.3,23.8v36.7c0,10.5,2.8,18.5,8.3,23.8,5.5,5.4,13.1,8.1,22.6,8.1s17.3-2.7,22.7-8.1c5.4-5.4,8.2-13.4,8.2-23.9v-36.7c0-10.5-2.8-18.5-8.2-23.9-5.4-5.4-13.1-8.1-22.7-8.1Z"/>
+        <path d="M460.6,197.8c-18,0-32.6-14.6-32.6-32.6V12.5h24.1v152.6c0,4.7,3.8,8.5,8.5,8.5h16.8v24.1h-16.8Z"/>
+        <path d="M722.8,198.1V45.2c0-4.7-3.8-8.5-8.5-8.5h-16.8V12.5h16.8c18,0,32.6,14.6,32.6,32.6v152.9h-24.1Z"/>
+        <path d="M665.2,198.1c-18,0-32.6-14.6-32.6-32.6v-85.1h-20.3v-22.1h20.3V12.9h24.1v45.3h30.2v22.1h-30.2v85.1c0,4.7,3.8,8.5,8.5,8.5h21.7v24.1h-21.7Z"/>
+      </g>
+    </g>
+  </g>
+</svg>

image/axolotl_wordmark_digital_white.svg

@@ -0,0 +1,24 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<svg xmlns="http://www.w3.org/2000/svg" version="1.1" viewBox="0 0 765.4 212.6">
+  <defs>
+    <style>
+      .cls-1 {
+        fill: #fff;
+      }
+    </style>
+  </defs>
+  <!-- Generator: Adobe Illustrator 28.7.1, SVG Export Plug-In . SVG Version: 1.2.0 Build 142)  -->
+  <g>
+    <g id="Layer_1">
+      <g>
+        <path class="cls-1" d="M121.6,198.1l-12.1-48.8h-54.4l-12.1,48.8h-24.7L66.6,12.9h31.6l47.9,185.1h-24.5ZM104.4,128.6l-13.8-55.6c-2.7-10.7-4.8-19.7-6.3-26.9-.9-4.2-1.5-7.5-2-9.9-.5,2.5-1.2,5.8-2,9.9-1.5,7.1-3.6,16.1-6.3,26.7l-13.8,55.9h44.3Z"/>
+        <path class="cls-1" d="M254.9,198.1l-29.9-45.6c-1.2-1.9-2.4-4.1-3.5-6.5-.8-1.7-1.5-3.3-2.1-4.5-.6,1.3-1.4,2.8-2.3,4.5-1.3,2.4-2.6,4.6-4,6.5l-29.9,45.6h-28.5l49.6-71.9-46.5-67.9h28.5l27.6,43.1c1.2,1.9,2.3,3.9,3.4,6.1.7,1.4,1.4,2.7,1.9,3.8.5-1.1,1.1-2.4,1.8-3.8,1.1-2.2,2.2-4.2,3.4-6.1l27.6-43.1h28.5l-46.5,68.2,49.3,71.7h-28.5Z"/>
+        <path class="cls-1" d="M345.2,200.1c-16.7,0-30.2-5-40.1-14.8-9.9-9.8-14.9-23.7-14.9-41.3v-31.7c0-17.7,5-31.7,14.8-41.4,9.8-9.7,23.4-14.7,40.3-14.7s30.4,4.9,40.3,14.7c9.8,9.7,14.8,23.7,14.8,41.4v31.7c0,17.6-5,31.5-14.9,41.3-9.9,9.8-23.4,14.8-40.1,14.8ZM345.2,77.8c-9.5,0-17.1,2.7-22.6,8.1-5.5,5.4-8.3,13.4-8.3,23.8v36.7c0,10.5,2.8,18.5,8.3,23.8,5.5,5.4,13.1,8.1,22.6,8.1s17.3-2.7,22.7-8.1c5.4-5.4,8.2-13.4,8.2-23.9v-36.7c0-10.5-2.8-18.5-8.2-23.9-5.4-5.4-13.1-8.1-22.7-8.1Z"/>
+        <path class="cls-1" d="M547.3,200.1c-16.7,0-30.2-5-40.1-14.8-9.9-9.8-14.9-23.7-14.9-41.3v-31.7c0-17.7,5-31.7,14.8-41.4,9.8-9.7,23.4-14.7,40.3-14.7s30.4,4.9,40.3,14.7c9.8,9.7,14.8,23.7,14.8,41.4v31.7c0,17.6-5,31.5-14.9,41.3-9.9,9.8-23.4,14.8-40.1,14.8ZM547.3,77.8c-9.5,0-17.1,2.7-22.6,8.1-5.5,5.4-8.3,13.4-8.3,23.8v36.7c0,10.5,2.8,18.5,8.3,23.8,5.5,5.4,13.1,8.1,22.6,8.1s17.3-2.7,22.7-8.1c5.4-5.4,8.2-13.4,8.2-23.9v-36.7c0-10.5-2.8-18.5-8.2-23.9-5.4-5.4-13.1-8.1-22.7-8.1Z"/>
+        <path class="cls-1" d="M460.6,197.8c-18,0-32.6-14.6-32.6-32.6V12.5h24.1v152.6c0,4.7,3.8,8.5,8.5,8.5h16.8v24.1h-16.8Z"/>
+        <path class="cls-1" d="M722.8,198.1V45.2c0-4.7-3.8-8.5-8.5-8.5h-16.8V12.5h16.8c18,0,32.6,14.6,32.6,32.6v152.9h-24.1Z"/>
+        <path class="cls-1" d="M665.2,198.1c-18,0-32.6-14.6-32.6-32.6v-85.1h-20.3v-22.1h20.3V12.9h24.1v45.3h30.2v22.1h-30.2v85.1c0,4.7,3.8,8.5,8.5,8.5h21.7v24.1h-21.7Z"/>
+      </g>
+    </g>
+  </g>
+</svg>

pyproject.toml

@@ -0,0 +1,19 @@
+[build-system]
+requires = ["setuptools>=64", "wheel", "setuptools_scm>=8"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "axolotl"
+dynamic = ["version", "dependencies", "optional-dependencies"]
+description = "LLM Trainer"
+readme = "README.md"
+requires-python = ">=3.10"
+
+[project.scripts]
+axolotl = "axolotl.cli.main:main"
+
+[project.urls]
+Homepage = "https://axolotl-ai-cloud.github.io/axolotl/"
+Repository = "https://github.com/axolotl-ai-cloud/axolotl.git"
+
+[tool.setuptools_scm]

requirements-tests.txt

@@ -1 +1,5 @@
 pytest
+pytest-xdist
+pytest-retry
+pytest-sugar
+tbparse

requirements.txt

@@ -1,44 +1,56 @@
 --extra-index-url https://huggingface.github.io/autogptq-index/whl/cu118/
 packaging==23.2
-peft==0.11.1
-transformers==4.41.1
-tokenizers==0.19.1
-bitsandbytes==0.43.1
-accelerate==0.30.1
-deepspeed @ git+https://github.com/microsoft/DeepSpeed.git@bc48371c5e1fb8fd70fc79285e66201dbb65679b
+peft==0.14.0
+transformers>=4.46.3
+tokenizers>=0.20.1
+bitsandbytes==0.45.0
+accelerate==1.2.0
+datasets==3.1.0
+deepspeed==0.15.4
 pydantic==2.6.3
 addict
 fire
 PyYAML>=6.0
 requests
-datasets==2.19.1
-flash-attn==2.5.8
+flash-attn==2.7.0.post2
 sentencepiece
 wandb
 einops
-xformers==0.0.26.post1
+xformers>=0.0.23.post1
 optimum==1.16.2
 hf_transfer
 colorama
 numba
-numpy>=1.24.4
+numpy>=1.24.4,<=2.0.1
 # qlora things
 evaluate==0.4.1
 scipy
-scikit-learn==1.2.2
-pynvml
+scikit-learn==1.4.2
+nvidia-ml-py==12.560.30
 art
-fschat @ git+https://github.com/lm-sys/FastChat.git@27a05b04a35510afb1d767ae7e5990cbd278f8fe
 gradio==3.50.2
 tensorboard
+python-dotenv==1.0.1
+autoawq==0.2.7.post3
+triton>=2.3.0
+liger-kernel==0.4.2
 
 mamba-ssm==1.2.0.post1
 
 # remote filesystems
-s3fs
-gcsfs
+s3fs>=2024.5.0
+gcsfs>=2024.5.0
 # adlfs
 
-trl @ git+https://github.com/huggingface/trl.git@f18253bf2d747f68acc9cd89da95c85ebf59dbb9
+trl==0.12.1
 zstandard==0.22.0
 fastcore
+
+# lm eval harness
+lm_eval==0.4.4
+langdetect==1.0.9
+immutabledict==4.2.0
+antlr4-python3-runtime==4.13.2
+
+torchao==0.5.0
+schedulefree==1.3.0

requirements_env.txt

@@ -0,0 +1,315 @@
+accelerate==0.34.1
+addict==2.4.0
+aiofiles==23.2.1
+aiohttp==3.9.0
+aiosignal==1.3.1
+aiostream==0.5.2
+alembic==1.13.1
+annotated-types==0.6.0
+annoy==1.17.3
+ansible==6.7.0
+ansible-core==2.13.13
+ansible-vault==2.1.0
+anyio==3.7.1
+appdirs==1.4.4
+art==6.0
+asgiref==3.7.2
+async-timeout==4.0.2
+attrdict==2.0.1
+attrs==22.2.0
+awscli==1.32.75
+-e git+ssh://git@github.com/OpenAccess-AI-Collective/axolotl.git@6e354682e3c1735d3f7fb9e362280c38e922260f#egg=axolotl
+backoff==2.2.1
+base58==2.1.1
+beartype==0.17.2
+bitnet==0.2.1
+bitsandbytes==0.42.0
+bittensor==6.7.0
+black==23.7.0
+blinker==1.7.0
+boto3==1.34.75
+botocore==1.34.75
+cachetools==5.3.3
+cachy==0.1.1
+certifi==2023.7.22
+cffi==1.16.0
+cfgv==3.3.1
+chai-guanaco==1.2.4
+charset-normalizer==3.2.0
+cleo==0.6.8
+click==8.1.7
+cloudpickle==2.0.0
+cohere==4.11.2
+colorama==0.4.4
+coloredlogs==15.0.1
+CoLT5-attention==0.10.20
+contextlib2==21.6.0
+contourpy==1.2.0
+cryptography==41.0.3
+cycler==0.12.1
+cytoolz==0.12.3
+databricks-cli==0.18.0
+dataclasses-json==0.5.7
+datasets==2.11.0
+ddt==1.6.0
+decorator==5.1.1
+deepspeed==0.15.0
+# Editable Git install with no remote (dialogpt==0.1)
+-e /Users/wing/Projects/ml/dialogpt/src
+dill==0.3.6
+distlib==0.3.6
+docker==7.0.0
+docker-pycreds==0.4.0
+docstring-parser==0.15
+docutils==0.16
+ecdsa==0.18.0
+einops==0.7.0
+einops-exts==0.0.4
+einx==0.1.3
+entrypoints==0.4
+eth-hash==0.6.0
+eth-keys==0.5.0
+eth-typing==4.0.0
+eth-utils==2.3.1
+evaluate==0.4.0
+exceptiongroup==1.1.1
+fastapi==0.109.2
+fastcore==1.5.29
+ffmpy==0.4.0
+filelock==3.12.2
+-e git+https://github.com/NousResearch/finetuning-subnet.git@24e9407d6b4430a7ca39d344692f89ce5a97d27e#egg=finetuning_subnet
+fire==0.5.0
+first==2.0.2
+flake8==7.0.0
+Flask==3.0.1
+fonttools==4.47.2
+frozendict==2.4.1
+frozenlist==1.3.3
+fschat @ git+https://github.com/lm-sys/FastChat.git@27a05b04a35510afb1d767ae7e5990cbd278f8fe
+fsspec==2023.6.0
+fuzzywuzzy==0.18.0
+gitdb==4.0.10
+GitPython==3.1.31
+google-pasta==0.2.0
+gradio==4.42.0
+gradio_client==1.3.0
+greenlet==2.0.2
+grpclib==0.4.7
+gunicorn==21.2.0
+h11==0.14.0
+h2==4.1.0
+hpack==4.0.0
+httpcore==0.17.3
+httpx==0.24.1
+huggingface-hub==0.23.4
+humanfriendly==10.0
+hyperframe==6.0.1
+identify==2.5.24
+idna==3.4
+immutables==0.20
+importlib-metadata==6.7.0
+importlib-resources==6.1.1
+inflection==0.5.1
+iniconfig==2.0.0
+itsdangerous==2.1.2
+Jinja2==3.1.2
+jmespath==1.0.1
+joblib==1.3.2
+jsonlines==3.1.0
+jsonschema==2.6.0
+kiwisolver==1.4.5
+langchain==0.0.144
+Levenshtein==0.24.0
+libcst==1.1.0
+liger-kernel==0.0.0
+lion-pytorch==0.1.2
+llama-cpp-python==0.1.36
+llvmlite==0.40.1
+local-attention==1.9.0
+loguru==0.7.0
+Mako==1.3.2
+Markdown==3.5.2
+markdown-it-py==3.0.0
+markdown2==2.4.10
+MarkupSafe==2.1.2
+marshmallow==3.19.0
+marshmallow-enum==1.5.1
+matplotlib==3.8.2
+mccabe==0.7.0
+mdurl==0.1.2
+MEGABYTE-pytorch==0.0.7
+-e git+https://github.com/cg123/mergekit.git@53c5f414774a0558b8d84858fb6374bc93a8f1c1#egg=mergekit
+mlflow==2.10.0
+modal==0.62.77
+more-itertools==10.2.0
+mpmath==1.2.1
+msgpack==1.0.7
+msgpack-numpy-opentensor==0.5.0
+multidict==6.0.4
+multiprocess==0.70.14
+munch==2.5.0
+mypy==1.3.0
+mypy-extensions==1.0.0
+nest-asyncio==1.6.0
+netaddr==0.10.1
+networkx==3.0rc1
+nh3==0.2.14
+nodeenv==1.8.0
+nomic==2.0.2
+numba==0.57.1
+numexpr==2.8.4
+numpy==1.24.4
+oauthlib==3.2.2
+openai==0.27.4
+openapi==1.1.0
+openapi-schema-pydantic==1.2.4
+optimum==1.8.6
+orjson==3.10.7
+packaging==23.1
+pandas==2.0.0
+parameterized==0.9.0
+password-strength==0.0.3.post2
+pastel==0.1.1
+pathos==0.3.0
+pathspec==0.11.1
+pathtools==0.1.2
+peft==0.11.1
+pendulum==3.0.0
+Pillow==9.5.0
+pip-tools==1.11.0
+platformdirs==3.2.0
+pluggy==1.4.0
+poetry==0.7.1
+pox==0.3.2
+ppft==1.7.6.6
+pre-commit==3.3.2
+prettytable==3.10.0
+prompt-toolkit==3.0.39
+protobuf==3.20.2
+protobuf3-to-dict==0.1.5
+psutil==5.9.5
+psycopg==3.1.18
+PuLP==2.8.0
+py==1.11.0
+py-bip39-bindings==0.1.11
+py-cpuinfo==9.0.0
+py-ed25519-zebra-bindings==1.0.1
+py-sr25519-bindings==0.2.0
+pyarrow==11.0.0
+pyasn1==0.6.0
+pycodestyle==2.11.1
+pycparser==2.21
+pycryptodome==3.20.0
+pydantic==2.5.3
+pydantic_core==2.14.6
+pydub==0.25.1
+pyfiglet==0.8.post1
+pyflakes==3.2.0
+Pygments==2.15.1
+PyJWT==2.8.0
+pylev==1.4.0
+PyNaCl==1.5.0
+pynvml==11.5.0
+pyparsing==2.4.7
+pyrsistent==0.14.11
+pytest==8.0.2
+pytest-asyncio==0.23.4
+python-dateutil==2.8.2
+python-dotenv==1.0.1
+python-Levenshtein==0.24.0
+python-multipart==0.0.9
+pytz==2023.3
+PyYAML==6.0.1
+querystring-parser==1.2.4
+rapidfuzz==3.6.1
+regex==2023.6.3
+requests==2.31.0
+requests-toolbelt==0.8.0
+resolvelib==0.8.1
+responses==0.18.0
+retry==0.9.2
+rich==13.7.0
+rsa==4.7.2
+ruff==0.6.3
+s3transfer==0.10.1
+safetensors==0.4.5
+sagemaker==2.148.0
+scalecodec==1.2.7
+schedulefree==1.2.1
+schema==0.7.5
+scikit-learn==1.4.0
+scipy==1.9.3
+seaborn==0.13.2
+semantic-version==2.10.0
+sentencepiece==0.2.0
+sentry-sdk==1.19.1
+setproctitle==1.3.2
+shellingham==1.5.4
+shortuuid==1.0.11
+shtab==1.6.5
+sigtools==4.0.1
+six==1.16.0
+skypilot==0.4.1
+smdebug-rulesconfig==1.0.1
+smmap==5.0.0
+sniffio==1.3.0
+SQLAlchemy==1.4.47
+sqlparse==0.4.4
+starlette==0.36.3
+substrate-interface==1.5.2
+svgwrite==1.4.3
+sympy==1.11.1
+synchronicity==0.6.7
+tabulate==0.9.0
+tblib==1.7.0
+tenacity==8.2.2
+tensor-parallel==2.0.0
+termcolor==2.2.0
+text2art==0.2.0
+threadpoolctl==3.2.0
+tiktoken==0.6.0
+time-machine==2.14.1
+timm==0.9.16
+tokenizers==0.19.1
+tokenmonster==1.1.12
+toml==0.9.6
+tomli==2.0.1
+tomlkit==0.12.0
+toolz==0.12.1
+torch==2.2.0
+torchdata==0.6.1
+torchdiffeq==0.2.3
+TorchFix==0.4.0
+torchtext==0.15.2
+torchvision==0.17.0
+tqdm==4.66.2
+transformers==4.44.2
+trl==0.9.6
+typer==0.12.5
+types-certifi==2021.10.8.3
+types-requests==2.31.0.20240125
+types-setuptools==69.0.0.20240125
+types-toml==0.10.8.7
+typing==3.7.4.3
+typing-inspect==0.8.0
+typing_extensions==4.9.0
+tyro==0.5.18
+tzdata==2023.3
+unique-names-generator==1.0.2
+urllib3==2.2.2
+uvicorn==0.22.0
+vector_quantize_pytorch==1.14.1
+virtualenv==20.23.0
+voyager==2.0.2
+wandb==0.16.2
+watchfiles==0.21.0
+wavedrom==2.0.3.post3
+wcwidth==0.2.6
+websocket-client==1.7.0
+websockets==12.0
+Werkzeug==3.0.1
+wonderwords==2.2.0
+xxhash==3.2.0
+yarl==1.8.2
+zetascale==2.2.7
+zipp==3.15.0

scripts/chat_datasets.py

@@ -0,0 +1,60 @@
+"""
+helper script to parse chat datasets into a usable yaml
+"""
+import click
+import yaml
+from datasets import load_dataset
+
+
+@click.command()
+@click.argument("dataset", type=str)
+@click.option("--split", type=str, default="train")
+def parse_dataset(dataset=None, split="train"):
+    ds_cfg = {}
+    ds_cfg["path"] = dataset
+    ds_cfg["split"] = split
+    ds_cfg["type"] = "chat_template"
+    ds_cfg["chat_template"] = "<<<Replace based on your model>>>"
+
+    dataset = load_dataset(dataset, split=split)
+    features = dataset.features
+    feature_keys = features.keys()
+    field_messages = None
+    for key in ["conversation", "conversations", "messages"]:
+        if key in feature_keys:
+            field_messages = key
+            break
+    if not field_messages:
+        raise ValueError(
+            f'No conversation field found in dataset: {", ".join(feature_keys)}'
+        )
+    ds_cfg["field_messages"] = field_messages
+
+    message_fields = features["conversations"][0].keys()
+    message_field_role = None
+    for key in ["from", "role"]:
+        if key in message_fields:
+            message_field_role = key
+            break
+    if not message_field_role:
+        raise ValueError(
+            f'No role field found in messages: {", ".join(message_fields)}'
+        )
+    ds_cfg["message_field_role"] = message_field_role
+
+    message_field_content = None
+    for key in ["content", "text", "value"]:
+        if key in message_fields:
+            message_field_content = key
+            break
+    if not message_field_content:
+        raise ValueError(
+            f'No content field found in messages: {", ".join(message_fields)}'
+        )
+    ds_cfg["message_field_content"] = message_field_content
+
+    print(yaml.dump({"datasets": [ds_cfg]}))
+
+
+if __name__ == "__main__":
+    parse_dataset()

scripts/cloud-entrypoint.sh

@@ -2,7 +2,7 @@
 
 # Export specific ENV variables to /etc/rp_environment
 echo "Exporting environment variables..."
-printenv | grep -E '^RUNPOD_|^PATH=|^_=' | sed 's/^\(.*\)=\(.*\)$/export \1="\2"/' >> /etc/rp_environment
+printenv | grep -E '^HF_|^BNB_|^CUDA_|^NCCL_|^NV|^RUNPOD_|^PATH=|^_=' | sed 's/^\([^=]*\)=\(.*\)$/export \1="\2"/' | grep -v 'printenv' >> /etc/rp_environment
 echo 'source /etc/rp_environment' >> ~/.bashrc
 
 add_keys_to_authorized() {

scripts/cutcrossentropy_install.py

@@ -0,0 +1,28 @@
+"""Script to output the correct installation command for cut-cross-entropy."""
+import importlib.util
+import sys
+
+try:
+    import torch
+except ImportError as exc:
+    raise ImportError("Install torch via `pip install torch`") from exc
+from packaging.version import Version as V
+
+v = V(torch.__version__)
+
+# no cut-cross-entropy support for torch < 2.4.0
+if v < V("2.4.0"):
+    print("")
+    sys.exit(0)
+
+cce_spec = importlib.util.find_spec("cut_cross_entropy")
+
+UNINSTALL_PREFIX = ""
+if cce_spec:
+    if not importlib.util.find_spec("cut_cross_entropy.transformers"):
+        UNINSTALL_PREFIX = "pip uninstall -y cut-cross-entropy && "
+
+print(
+    UNINSTALL_PREFIX
+    + 'pip install "cut-cross-entropy @ git+https://github.com/apple/ml-cross-entropy.git@9c297c905f55b73594b5d650722d1e78183b77bd"'
+)

scripts/motd

@@ -11,7 +11,7 @@ Welcome to the axolotl cloud image! If the you've mounted a disk to /workspace a
 ```
 cd /workspace
 rm -rf /workspace/axolotl
-git clone https://github.com/OpenAccess-AI-Collective/axolotl.git
+git clone https://github.com/axolotl-ai-cloud/axolotl.git
 cd axolotl
 pip install --no-deps -e .
 ```

scripts/unsloth_install.py

@@ -0,0 +1,36 @@
+# noqa
+# pylint: skip-file
+try:
+    import torch
+except ImportError:
+    raise ImportError("Install torch via `pip install torch`")
+from packaging.version import Version as V
+
+v = V(torch.__version__)
+cuda = str(torch.version.cuda)
+try:
+    is_ampere = torch.cuda.get_device_capability()[0] >= 8
+except RuntimeError:
+    is_ampere = False
+if cuda != "12.1" and cuda != "11.8" and cuda != "12.4":
+    raise RuntimeError(f"CUDA = {cuda} not supported!")
+if v <= V("2.1.0"):
+    raise RuntimeError(f"Torch = {v} too old!")
+elif v <= V("2.1.1"):
+    x = "cu{}{}-torch211"
+elif v <= V("2.1.2"):
+    x = "cu{}{}-torch212"
+elif v < V("2.3.0"):
+    x = "cu{}{}-torch220"
+elif v < V("2.4.0"):
+    x = "cu{}{}-torch230"
+elif v < V("2.5.0"):
+    x = "cu{}{}-torch240"
+elif v < V("2.6.0"):
+    x = "cu{}{}-torch250"
+else:
+    raise RuntimeError(f"Torch = {v} too new!")
+x = x.format(cuda.replace(".", ""), "-ampere" if is_ampere else "")
+print(
+    f'pip install unsloth-zoo==2024.11.7 && pip install --no-deps "unsloth[{x}]==2024.11.9"'
+)

setup.py

@@ -1,5 +1,4 @@
 """setup.py for axolotl"""
-
 import platform
 import re
 from importlib.metadata import PackageNotFoundError, version
@@ -29,13 +28,20 @@ def parse_requirements():
                 _install_requires.append(line)
 
     try:
+        xformers_version = [req for req in _install_requires if "xformers" in req][0]
+        torchao_version = [req for req in _install_requires if "torchao" in req][0]
+        autoawq_version = [req for req in _install_requires if "autoawq" in req][0]
+
         if "Darwin" in platform.system():
             # don't install xformers on MacOS
-            _install_requires.pop(_install_requires.index("xformers==0.0.26.post1"))
+            _install_requires.pop(_install_requires.index(xformers_version))
         else:
             # detect the version of torch already installed
             # and set it so dependencies don't clobber the torch version
-            torch_version = version("torch")
+            try:
+                torch_version = version("torch")
+            except PackageNotFoundError:
+                torch_version = "2.5.1"
             _install_requires.append(f"torch=={torch_version}")
 
             version_match = re.match(r"^(\d+)\.(\d+)(?:\.(\d+))?", torch_version)
@@ -48,46 +54,65 @@ def parse_requirements():
             else:
                 raise ValueError("Invalid version format")
 
-            if (major, minor) >= (2, 3):
-                pass
+            if (major, minor) >= (2, 5):
+                _install_requires.pop(_install_requires.index(xformers_version))
+                if patch == 0:
+                    _install_requires.append("xformers==0.0.28.post2")
+                else:
+                    _install_requires.append("xformers==0.0.28.post3")
+                _install_requires.pop(_install_requires.index(autoawq_version))
+            elif (major, minor) >= (2, 4):
+                if patch == 0:
+                    _install_requires.pop(_install_requires.index(xformers_version))
+                    _install_requires.append("xformers>=0.0.27")
+                else:
+                    _install_requires.pop(_install_requires.index(xformers_version))
+                    _install_requires.append("xformers==0.0.28.post1")
+            elif (major, minor) >= (2, 3):
+                _install_requires.pop(_install_requires.index(torchao_version))
+                if patch == 0:
+                    _install_requires.pop(_install_requires.index(xformers_version))
+                    _install_requires.append("xformers>=0.0.26.post1")
+                else:
+                    _install_requires.pop(_install_requires.index(xformers_version))
+                    _install_requires.append("xformers>=0.0.27")
             elif (major, minor) >= (2, 2):
-                _install_requires.pop(_install_requires.index("xformers==0.0.26.post1"))
+                _install_requires.pop(_install_requires.index(torchao_version))
+                _install_requires.pop(_install_requires.index(xformers_version))
                 _install_requires.append("xformers>=0.0.25.post1")
             else:
-                _install_requires.pop(_install_requires.index("xformers==0.0.26.post1"))
+                _install_requires.pop(_install_requires.index(torchao_version))
+                _install_requires.pop(_install_requires.index(xformers_version))
                 _install_requires.append("xformers>=0.0.23.post1")
 
     except PackageNotFoundError:
         pass
-
     return _install_requires, _dependency_links
 
 
 install_requires, dependency_links = parse_requirements()
 
-
 setup(
-    name="axolotl",
-    version="0.4.1",
-    description="LLM Trainer",
-    long_description="Axolotl is a tool designed to streamline the fine-tuning of various AI models, offering support for multiple configurations and architectures.",
     package_dir={"": "src"},
-    packages=find_packages(),
+    packages=find_packages("src"),
     install_requires=install_requires,
     dependency_links=dependency_links,
+    entry_points={
+        "console_scripts": [
+            "axolotl=axolotl.cli.main:main",
+        ],
+    },
     extras_require={
         "flash-attn": [
-            "flash-attn==2.5.8",
-        ],
-        "fused-dense-lib": [
-            "fused-dense-lib  @ git+https://github.com/Dao-AILab/flash-attention@v2.5.8#subdirectory=csrc/fused_dense_lib",
+            "flash-attn==2.7.0.post2",
         ],
         "deepspeed": [
-            "deepspeed @ git+https://github.com/microsoft/DeepSpeed.git@bc48371c5e1fb8fd70fc79285e66201dbb65679b",
+            "deepspeed==0.15.4",
             "deepspeed-kernels",
         ],
         "mamba-ssm": [
             "mamba-ssm==1.2.0.post1",
+            "causal_conv1d",
         ],
         "auto-gptq": [
             "auto-gptq==0.5.1",
@@ -101,5 +126,11 @@ setup(
         "galore": [
             "galore_torch",
         ],
+        "optimizers": [
+            "galore_torch",
+            "lion-pytorch==0.1.2",
+            "lomo-optim==0.1.1",
+            "torch-optimi==0.2.1",
+        ],
     },
 )

src/axolotl/__init__.py

@@ -0,0 +1,8 @@
+"""Axolotl - Train and fine-tune large language models"""
+
+try:
+    from importlib.metadata import version
+
+    __version__ = version("axolotl")
+except ImportError:
+    __version__ = "unknown"

src/axolotl/cli/__init__.py

@@ -29,18 +29,24 @@ from transformers.utils.import_utils import _is_package_available
 from axolotl.common.cli import TrainerCliArgs, load_model_and_tokenizer
 from axolotl.logging_config import configure_logging
 from axolotl.train import TrainDatasetMeta
+from axolotl.utils.chat_templates import (
+    get_chat_template,
+    get_chat_template_from_config,
+)
+from axolotl.utils.comet_ import setup_comet_env_vars
 from axolotl.utils.config import (
     normalize_cfg_datasets,
     normalize_config,
+    prepare_plugins,
     validate_config,
 )
 from axolotl.utils.data import load_prepare_dpo_datasets, prepare_dataset
 from axolotl.utils.dict import DictDefault
 from axolotl.utils.distributed import is_main_process
 from axolotl.utils.mlflow_ import setup_mlflow_env_vars
-from axolotl.utils.models import load_tokenizer
+from axolotl.utils.models import load_processor, load_tokenizer
 from axolotl.utils.tokenization import check_dataset_labels
-from axolotl.utils.trainer import prepare_optim_env
+from axolotl.utils.trainer import prepare_opinionated_env, prepare_optim_env
 from axolotl.utils.wandb_ import setup_wandb_env_vars
 
 project_root = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
@@ -52,8 +58,22 @@ LOG = logging.getLogger("axolotl.scripts")
 
 os.environ["HF_HUB_ENABLE_HF_TRANSFER"] = "1"
 
+AXOLOTL_LOGO = """
+     #@@ #@@      @@# @@#
+    @@  @@          @@  @@           =@@#                               @@                 #@    =@@#.
+    @@    #@@@@@@@@@    @@           #@#@=                              @@                 #@     .=@@
+      #@@@@@@@@@@@@@@@@@            =@# @#     ##=     ##    =####=+    @@      =#####+  =#@@###.   @@
+    @@@@@@@@@@/  +@@/  +@@          #@  =@=     #@=   @@   =@#+  +#@#   @@    =@#+  +#@#   #@.      @@
+    @@@@@@@@@@  ##@@  ##@@         =@#   @#      =@# @#    @@      @@   @@    @@      #@   #@       @@
+     @@@@@@@@@@@@@@@@@@@@          #@=+++#@=      =@@#     @@      @@   @@    @@      #@   #@       @@
+                                  =@#=====@@     =@# @#    @@      @@   @@    @@      #@   #@       @@
+    @@@@@@@@@@@@@@@@  @@@@        #@      #@=   #@=  +@@   #@#    =@#   @@.   =@#    =@#   #@.      @@
+                                 =@#       @#  #@=     #@   =#@@@@#=    +#@@=  +#@@@@#=    .##@@+   @@
+    @@@@  @@@@@@@@@@@@@@@@
+"""
 
-def print_axolotl_text_art(suffix=None):
+
+def print_legacy_axolotl_text_art(suffix=None):
     font = "nancyj"
     ascii_text = "  axolotl"
     if suffix:
@@ -66,6 +86,13 @@ def print_axolotl_text_art(suffix=None):
     print_dep_versions()
 
 
+def print_axolotl_text_art(
+    **kwargs,  # pylint: disable=unused-argument
+):
+    if is_main_process():
+        print(AXOLOTL_LOGO)
+
+
 def print_dep_versions():
     packages = ["accelerate", "peft", "transformers", "trl", "torch", "bitsandbytes"]
     max_len = max(len(pkg) for pkg in packages)
@@ -73,8 +100,8 @@ def print_dep_versions():
         print("*" * 40)
         print("**** Axolotl Dependency Versions *****")
         for pkg in packages:
-            version = _is_package_available(pkg, return_version=True)
-            print(f"{pkg: >{max_len}}: {version[1]: <15}")
+            pkg_version = _is_package_available(pkg, return_version=True)
+            print(f"{pkg: >{max_len}}: {pkg_version[1]: <15}")
         print("*" * 40)
 
 
@@ -112,7 +139,7 @@ def check_remote_config(config: Union[str, Path]):
         with open(output_path, "wb") as file:
             file.write(content)
         LOG.info(
-            f"Using the following config obtained from {config}:\n\n{content.decode('utf-8')}\n"
+            f"Using the following config obtained from {config}: \n\n{content.decode('utf-8')}\n"
         )
         return output_path
 
@@ -166,18 +193,19 @@ def do_inference(
 ):
     model, tokenizer = load_model_and_tokenizer(cfg=cfg, cli_args=cli_args)
     prompter = cli_args.prompter
-    default_tokens = {"unk_token": "<unk>", "bos_token": "<s>", "eos_token": "</s>"}
-
-    for token, symbol in default_tokens.items():
-        # If the token isn't already specified in the config, add it
-        if not (cfg.special_tokens and token in cfg.special_tokens):
-            tokenizer.add_special_tokens({token: symbol})
 
     prompter_module = None
+    chat_template_str = None
     if prompter:
         prompter_module = getattr(
             importlib.import_module("axolotl.prompters"), prompter
         )
+    elif cfg.chat_template:
+        chat_template_str = get_chat_template(cfg.chat_template)
+    elif cfg.datasets[0].type == "chat_template":
+        chat_template_str = get_chat_template_from_config(
+            cfg=cfg, ds_cfg=cfg.datasets[0], tokenizer=tokenizer
+        )
 
     model = model.to(cfg.device, dtype=cfg.torch_dtype)
 
@@ -187,13 +215,31 @@ def do_inference(
         instruction = get_multi_line_input()
         if not instruction:
             return
+
         if prompter_module:
             prompt: str = next(
                 prompter_module().build_prompt(instruction=instruction.strip("\n"))
             )
         else:
             prompt = instruction.strip()
-        batch = tokenizer(prompt, return_tensors="pt", add_special_tokens=True)
+
+        if chat_template_str:
+            batch = tokenizer.apply_chat_template(
+                [
+                    {
+                        "role": "user",
+                        "content": prompt,
+                    }
+                ],
+                return_tensors="pt",
+                add_special_tokens=True,
+                add_generation_prompt=True,
+                chat_template=chat_template_str,
+                tokenize=True,
+                return_dict=True,
+            )
+        else:
+            batch = tokenizer(prompt, return_tensors="pt", add_special_tokens=True)
 
         print("=" * 40)
         model.eval()
@@ -233,18 +279,15 @@ def do_inference_gradio(
 
     model, tokenizer = load_model_and_tokenizer(cfg=cfg, cli_args=cli_args)
     prompter = cli_args.prompter
-    default_tokens = {"unk_token": "<unk>", "bos_token": "<s>", "eos_token": "</s>"}
-
-    for token, symbol in default_tokens.items():
-        # If the token isn't already specified in the config, add it
-        if not (cfg.special_tokens and token in cfg.special_tokens):
-            tokenizer.add_special_tokens({token: symbol})
 
     prompter_module = None
+    chat_template_str = None
     if prompter:
         prompter_module = getattr(
             importlib.import_module("axolotl.prompters"), prompter
         )
+    elif cfg.chat_template:
+        chat_template_str = get_chat_template(cfg.chat_template, tokenizer=tokenizer)
 
     model = model.to(cfg.device, dtype=cfg.torch_dtype)
 
@@ -258,7 +301,24 @@ def do_inference_gradio(
             )
         else:
             prompt = instruction.strip()
-        batch = tokenizer(prompt, return_tensors="pt", add_special_tokens=True)
+
+        if chat_template_str:
+            batch = tokenizer.apply_chat_template(
+                [
+                    {
+                        "role": "user",
+                        "content": prompt,
+                    }
+                ],
+                return_tensors="pt",
+                add_special_tokens=True,
+                add_generation_prompt=True,
+                chat_template=chat_template_str,
+                tokenize=True,
+                return_dict=True,
+            )
+        else:
+            batch = tokenizer(prompt, return_tensors="pt", add_special_tokens=True)
 
         model.eval()
         with torch.no_grad():
@@ -281,6 +341,7 @@ def do_inference_gradio(
             streamer = TextIteratorStreamer(tokenizer)
             generation_kwargs = {
                 "inputs": batch["input_ids"].to(cfg.device),
+                "attention_mask": batch["attention_mask"].to(cfg.device),
                 "generation_config": generation_config,
                 "streamer": streamer,
             }
@@ -319,7 +380,7 @@ def choose_config(path: Path):
 
     if len(yaml_files) == 1:
         print(f"Using default YAML file '{yaml_files[0]}'")
-        return yaml_files[0]
+        return str(yaml_files[0])
 
     print("Choose a YAML file:")
     for idx, file in enumerate(yaml_files):
@@ -330,7 +391,7 @@ def choose_config(path: Path):
         try:
             choice = int(input("Enter the number of your choice: "))
             if 1 <= choice <= len(yaml_files):
-                chosen_file = yaml_files[choice - 1]
+                chosen_file = str(yaml_files[choice - 1])
             else:
                 print("Invalid choice. Please choose a number from the list.")
         except ValueError:
@@ -371,17 +432,24 @@ def load_cfg(config: Union[str, Path] = Path("examples/"), **kwargs):
     except:  # pylint: disable=bare-except # noqa: E722
         gpu_version = None
 
+    prepare_plugins(cfg)
+
     cfg = validate_config(
         cfg,
         capabilities={
             "bf16": is_torch_bf16_gpu_available(),
-            "n_gpu": os.environ.get("WORLD_SIZE", 1),
+            "n_gpu": int(os.environ.get("WORLD_SIZE", 1)),
             "compute_capability": gpu_version,
         },
+        env_capabilities={
+            "torch_version": str(torch.__version__).split("+", maxsplit=1)[0],
+        },
     )
 
     prepare_optim_env(cfg)
 
+    prepare_opinionated_env(cfg)
+
     normalize_config(cfg)
 
     normalize_cfg_datasets(cfg)
@@ -390,6 +458,8 @@ def load_cfg(config: Union[str, Path] = Path("examples/"), **kwargs):
 
     setup_mlflow_env_vars(cfg)
 
+    setup_comet_env_vars(cfg)
+
     return cfg
 
 
@@ -399,12 +469,20 @@ def load_datasets(
     cli_args: TrainerCliArgs,
 ) -> TrainDatasetMeta:
     tokenizer = load_tokenizer(cfg)
+    processor = load_processor(cfg, tokenizer=tokenizer) if cfg.processor_type else None
 
     train_dataset, eval_dataset, total_num_steps, prompters = prepare_dataset(
-        cfg, tokenizer
+        cfg,
+        tokenizer,
+        processor=processor,
     )
 
-    if cli_args.debug or cfg.debug:
+    if (
+        cli_args.debug
+        or cfg.debug
+        or cli_args.debug_text_only
+        or int(cli_args.debug_num_examples) > 0
+    ):
         LOG.info("check_dataset_labels...")
         check_dataset_labels(
             train_dataset.select(

src/axolotl/cli/inference.py

@@ -2,9 +2,11 @@
 CLI to run inference on a trained model
 """
 from pathlib import Path
+from typing import Union
 
 import fire
 import transformers
+from dotenv import load_dotenv
 
 from axolotl.cli import (
     do_inference,
@@ -15,10 +17,10 @@ from axolotl.cli import (
 from axolotl.common.cli import TrainerCliArgs
 
 
-def do_cli(config: Path = Path("examples/"), gradio=False, **kwargs):
+def do_cli(config: Union[Path, str] = Path("examples/"), gradio=False, **kwargs):
     # pylint: disable=duplicate-code
     print_axolotl_text_art()
-    parsed_cfg = load_cfg(config, **kwargs)
+    parsed_cfg = load_cfg(config, inference=True, **kwargs)
     parsed_cfg.sample_packing = False
     parser = transformers.HfArgumentParser((TrainerCliArgs))
     parsed_cli_args, _ = parser.parse_args_into_dataclasses(
@@ -33,4 +35,5 @@ def do_cli(config: Path = Path("examples/"), gradio=False, **kwargs):
 
 
 if __name__ == "__main__":
+    load_dotenv()
     fire.Fire(do_cli)

src/axolotl/cli/main.py

@@ -0,0 +1,233 @@
+"""CLI definition for various axolotl commands."""
+# pylint: disable=redefined-outer-name
+import subprocess  # nosec B404
+from typing import Optional
+
+import click
+
+import axolotl
+from axolotl.cli.utils import (
+    add_options_from_config,
+    add_options_from_dataclass,
+    build_command,
+    fetch_from_github,
+)
+from axolotl.common.cli import PreprocessCliArgs, TrainerCliArgs
+from axolotl.utils.config.models.input.v0_4_1 import AxolotlInputConfig
+
+
+@click.group()
+@click.version_option(version=axolotl.__version__, prog_name="axolotl")
+def cli():
+    """Axolotl CLI - Train and fine-tune large language models"""
+
+
+@cli.command()
+@click.argument("config", type=click.Path(exists=True, path_type=str))
+@add_options_from_dataclass(PreprocessCliArgs)
+@add_options_from_config(AxolotlInputConfig)
+def preprocess(config: str, **kwargs):
+    """Preprocess datasets before training."""
+    kwargs = {k: v for k, v in kwargs.items() if v is not None}
+
+    from axolotl.cli.preprocess import do_cli
+
+    do_cli(config=config, **kwargs)
+
+
+@cli.command()
+@click.argument("config", type=click.Path(exists=True, path_type=str))
+@click.option(
+    "--accelerate/--no-accelerate",
+    default=True,
+    help="Use accelerate launch for multi-GPU training",
+)
+@add_options_from_dataclass(TrainerCliArgs)
+@add_options_from_config(AxolotlInputConfig)
+def train(config: str, accelerate: bool, **kwargs):
+    """Train or fine-tune a model."""
+    kwargs = {k: v for k, v in kwargs.items() if v is not None}
+
+    if accelerate:
+        base_cmd = ["accelerate", "launch", "-m", "axolotl.cli.train"]
+        if config:
+            base_cmd.append(config)
+        cmd = build_command(base_cmd, kwargs)
+        subprocess.run(cmd, check=True)  # nosec B603
+    else:
+        from axolotl.cli.train import do_cli
+
+        do_cli(config=config, **kwargs)
+
+
+@cli.command()
+@click.argument("config", type=click.Path(exists=True, path_type=str))
+@click.option(
+    "--accelerate/--no-accelerate",
+    default=True,
+    help="Use accelerate launch for multi-GPU inference",
+)
+@click.option(
+    "--lora-model-dir",
+    type=click.Path(exists=True, path_type=str),
+    help="Directory containing LoRA model",
+)
+@click.option(
+    "--base-model",
+    type=click.Path(exists=True, path_type=str),
+    help="Path to base model for non-LoRA models",
+)
+@click.option("--gradio", is_flag=True, help="Launch Gradio interface")
+@click.option("--load-in-8bit", is_flag=True, help="Load model in 8-bit mode")
+@add_options_from_dataclass(TrainerCliArgs)
+@add_options_from_config(AxolotlInputConfig)
+def inference(
+    config: str,
+    accelerate: bool,
+    lora_model_dir: Optional[str] = None,
+    base_model: Optional[str] = None,
+    **kwargs,
+):
+    """Run inference with a trained model."""
+    kwargs = {k: v for k, v in kwargs.items() if v is not None}
+    del kwargs["inference"]  # interferes with inference.do_cli
+
+    if lora_model_dir:
+        kwargs["lora_model_dir"] = lora_model_dir
+    if base_model:
+        kwargs["output_dir"] = base_model
+
+    if accelerate:
+        base_cmd = ["accelerate", "launch", "-m", "axolotl.cli.inference"]
+        if config:
+            base_cmd.append(config)
+        cmd = build_command(base_cmd, kwargs)
+        subprocess.run(cmd, check=True)  # nosec B603
+    else:
+        from axolotl.cli.inference import do_cli
+
+        do_cli(config=config, **kwargs)
+
+
+@cli.command()
+@click.argument("config", type=click.Path(exists=True, path_type=str))
+@click.option(
+    "--accelerate/--no-accelerate",
+    default=False,
+    help="Use accelerate launch for multi-GPU operations",
+)
+@click.option(
+    "--model-dir",
+    type=click.Path(exists=True, path_type=str),
+    help="Directory containing model weights to shard",
+)
+@click.option(
+    "--save-dir",
+    type=click.Path(path_type=str),
+    help="Directory to save sharded weights",
+)
+@add_options_from_dataclass(TrainerCliArgs)
+@add_options_from_config(AxolotlInputConfig)
+def shard(config: str, accelerate: bool, **kwargs):
+    """Shard model weights."""
+    kwargs = {k: v for k, v in kwargs.items() if v is not None}
+
+    if accelerate:
+        base_cmd = ["accelerate", "launch", "-m", "axolotl.cli.shard"]
+        if config:
+            base_cmd.append(config)
+        cmd = build_command(base_cmd, kwargs)
+        subprocess.run(cmd, check=True)  # nosec B603
+    else:
+        from axolotl.cli.shard import do_cli
+
+        do_cli(config=config, **kwargs)
+
+
+@cli.command()
+@click.argument("config", type=click.Path(exists=True, path_type=str))
+@click.option(
+    "--accelerate/--no-accelerate",
+    default=True,
+    help="Use accelerate launch for weight merging",
+)
+@click.option(
+    "--model-dir",
+    type=click.Path(exists=True, path_type=str),
+    help="Directory containing sharded weights",
+)
+@click.option(
+    "--save-path", type=click.Path(path_type=str), help="Path to save merged weights"
+)
+@add_options_from_dataclass(TrainerCliArgs)
+@add_options_from_config(AxolotlInputConfig)
+def merge_sharded_fsdp_weights(config: str, accelerate: bool, **kwargs):
+    """Merge sharded FSDP model weights."""
+    kwargs = {k: v for k, v in kwargs.items() if v is not None}
+
+    if accelerate:
+        base_cmd = [
+            "accelerate",
+            "launch",
+            "-m",
+            "axolotl.cli.merge_sharded_fsdp_weights",
+        ]
+        if config:
+            base_cmd.append(config)
+        cmd = build_command(base_cmd, kwargs)
+        subprocess.run(cmd, check=True)  # nosec B603
+    else:
+        from axolotl.cli.merge_sharded_fsdp_weights import do_cli
+
+        do_cli(config=config, **kwargs)
+
+
+@cli.command()
+@click.argument("config", type=click.Path(exists=True, path_type=str))
+@click.option(
+    "--lora-model-dir",
+    type=click.Path(exists=True, path_type=str),
+    help="Directory containing the LoRA model to merge",
+)
+@click.option(
+    "--output-dir",
+    type=click.Path(path_type=str),
+    help="Directory to save the merged model",
+)
+def merge_lora(
+    config: str,
+    lora_model_dir: Optional[str] = None,
+    output_dir: Optional[str] = None,
+):
+    """Merge a trained LoRA into a base model"""
+    kwargs = {}
+    if lora_model_dir:
+        kwargs["lora_model_dir"] = lora_model_dir
+    if output_dir:
+        kwargs["output_dir"] = output_dir
+
+    from axolotl.cli.merge_lora import do_cli
+
+    do_cli(config=config, **kwargs)
+
+
+@cli.command()
+@click.argument("directory", type=click.Choice(["examples", "deepspeed_configs"]))
+@click.option("--dest", help="Destination directory")
+def fetch(directory: str, dest: Optional[str]):
+    """
+    Fetch example configs or other resources.
+
+    Available directories:
+    - examples: Example configuration files
+    - deepspeed_configs: DeepSpeed configuration files
+    """
+    fetch_from_github(f"{directory}/", dest)
+
+
+def main():
+    cli()
+
+
+if __name__ == "__main__":
+    main()

src/axolotl/cli/merge_lora.py

@@ -2,15 +2,17 @@
 CLI to run merge a trained LoRA into a base model
 """
 from pathlib import Path
+from typing import Union
 
 import fire
 import transformers
+from dotenv import load_dotenv
 
 from axolotl.cli import do_merge_lora, load_cfg, print_axolotl_text_art
 from axolotl.common.cli import TrainerCliArgs
 
 
-def do_cli(config: Path = Path("examples/"), **kwargs):
+def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs):
     # pylint: disable=duplicate-code
     print_axolotl_text_art()
     parser = transformers.HfArgumentParser((TrainerCliArgs))
@@ -48,4 +50,5 @@ def do_cli(config: Path = Path("examples/"), **kwargs):
 
 
 if __name__ == "__main__":
+    load_dotenv()
     fire.Fire(do_cli)

src/axolotl/cli/merge_sharded_fsdp_weights.py

@@ -0,0 +1,204 @@
+"""
+This module provides a CLI to merge sharded FSDP model checkpoints into a single combined checkpoint
+"""
+import json
+import logging
+import os
+import shutil
+from pathlib import Path
+from typing import Dict, Union
+
+import fire
+import torch
+import torch.distributed.checkpoint as dist_cp
+import torch.distributed.checkpoint.format_utils as dist_cp_format_utils
+import transformers
+from accelerate.utils import (
+    SAFE_WEIGHTS_INDEX_NAME,
+    SAFE_WEIGHTS_NAME,
+    WEIGHTS_INDEX_NAME,
+    WEIGHTS_NAME,
+    is_torch_version,
+)
+from dotenv import load_dotenv
+from huggingface_hub import split_torch_state_dict_into_shards
+from safetensors.torch import save_file as safe_save_file
+from torch.distributed.checkpoint.format_utils import _EmptyStateDictLoadPlanner
+
+from axolotl.cli import load_cfg, print_axolotl_text_art
+from axolotl.common.cli import TrainerCliArgs
+
+LOG = logging.getLogger("axolotl.cli.merge_sharded_fsdp_weights")
+
+
+class BFloat16CastPlanner(_EmptyStateDictLoadPlanner):
+    """
+    A custom planner to cast tensors to bfloat16 on the fly during loading.
+    """
+
+    def commit_tensor(self, read_item, tensor):  # pylint: disable=unused-argument
+        tensor.copy_(tensor.to(torch.bfloat16))
+
+
+def _distributed_checkpoint_to_merged_weights(
+    checkpoint_dir: Union[str, Path],
+    save_path: str,
+    safe_serialization: bool = False,
+    max_shard_size: str = "5GB",
+):
+    """
+    Passthrough to `torch.distributed.checkpoint.format_utils.dcp_to_torch_save`
+
+    Will save under `save_path` as either `model.safetensors` or `pytorch_model.bin`.
+    """
+
+    state_dict: Dict = {}
+    save_path_ = Path(save_path)
+    save_path_.mkdir(exist_ok=True)
+    dist_cp_format_utils._load_state_dict(  # pylint: disable=protected-access
+        state_dict,
+        storage_reader=dist_cp.FileSystemReader(checkpoint_dir),
+        planner=BFloat16CastPlanner(),  # pylint: disable=protected-access
+        no_dist=True,
+    )
+
+    # To handle if state is a dict like {model: {...}}
+    if len(state_dict.keys()) == 1:
+        state_dict = state_dict[list(state_dict)[0]]
+
+    # Ensure all tensors are in bfloat16
+    for key, value in state_dict.items():
+        if isinstance(value, torch.Tensor) and value.dtype != torch.bfloat16:
+            state_dict[key] = value.to(torch.bfloat16)
+
+    weights_name = SAFE_WEIGHTS_NAME if safe_serialization else WEIGHTS_NAME
+
+    filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(
+        ".safetensors", "{suffix}.safetensors"
+    )
+    state_dict_split = split_torch_state_dict_into_shards(
+        state_dict, filename_pattern=filename_pattern, max_shard_size=max_shard_size
+    )
+    # Save index if sharded
+    index = None
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+
+    # Save the model
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+
+    for shard_file, tensors in filename_to_tensors:
+        shard = {tensor: state_dict[tensor] for tensor in tensors}
+
+        if safe_serialization:
+            safe_save_file(
+                shard, os.path.join(save_path_, shard_file), metadata={"format": "pt"}
+            )
+        else:
+            torch.save(shard, os.path.join(save_path_, shard_file))
+
+    if index is not None:
+        save_index_file = (
+            SAFE_WEIGHTS_INDEX_NAME if safe_serialization else WEIGHTS_INDEX_NAME
+        )
+        save_index_file = os.path.join(save_path_, save_index_file)
+        # Save the index as well
+        with open(save_index_file, "w", encoding="utf-8") as fout:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            fout.write(content)
+
+    return save_path_
+
+
+def merge_fsdp_weights(
+    checkpoint_dir: str,
+    output_path: str,
+    safe_serialization: bool = False,
+    remove_checkpoint_dir: bool = False,
+):
+    """
+    Merge the weights from sharded FSDP model checkpoints into a single combined checkpoint. Should be used if
+    `SHARDED_STATE_DICT` was used for the model. Weights will be saved to `{output_path}/model.safetensors` if
+    `safe_serialization` else `pytorch_model.bin`.
+
+    Note: this is a CPU-bound process.
+
+    Args:
+        checkpoint_dir (`str`):
+            The directory containing the FSDP checkpoints (can be either the model or optimizer).
+        output_path (`str`):
+            The path to save the merged checkpoint.
+        safe_serialization (`bool`, *optional*, defaults to `True`):
+            Whether to save the merged weights with safetensors (recommended).
+        remove_checkpoint_dir (`bool`, *optional*, defaults to `False`):
+            Whether to remove the checkpoint directory after merging.
+    """
+    checkpoint_dir_ = Path(checkpoint_dir)
+    from accelerate.state import PartialState
+
+    if not is_torch_version(">=", "2.3.0"):
+        raise ValueError("`merge_fsdp_weights` requires PyTorch >= 2.3.0`")
+
+    # Verify that the checkpoint directory exists
+    if not checkpoint_dir_.exists():
+        model_path_exists = (checkpoint_dir_ / "pytorch_model_fsdp_0").exists()
+        optimizer_path_exists = (checkpoint_dir_ / "optimizer_0").exists()
+        err = f"Tried to load from {checkpoint_dir_} but couldn't find a valid metadata file."
+        if model_path_exists and optimizer_path_exists:
+            err += (
+                " However, potential model and optimizer checkpoint directories exist."
+            )
+            err += f"Please pass in either {checkpoint_dir_}/pytorch_model_fsdp_0 or {checkpoint_dir_}/optimizer_0"
+            err += "instead."
+        elif model_path_exists:
+            err += " However, a potential model checkpoint directory exists."
+            err += (
+                f"Please try passing in {checkpoint_dir_}/pytorch_model_fsdp_0 instead."
+            )
+        elif optimizer_path_exists:
+            err += " However, a potential optimizer checkpoint directory exists."
+            err += f"Please try passing in {checkpoint_dir_}/optimizer_0 instead."
+        raise ValueError(err)
+
+    # To setup `save` to work
+    state = PartialState()
+    if state.is_main_process:
+        LOG.info(f"Merging FSDP weights from {checkpoint_dir_}")
+        save_path = _distributed_checkpoint_to_merged_weights(
+            checkpoint_dir_, output_path, safe_serialization
+        )
+        LOG.info(f"Successfully merged FSDP weights and saved to {save_path}")
+        if remove_checkpoint_dir:
+            LOG.info(f"Removing old checkpoint directory {checkpoint_dir_}")
+            shutil.rmtree(checkpoint_dir_)
+    state.wait_for_everyone()
+
+
+def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs):
+    # pylint: disable=duplicate-code
+    print_axolotl_text_art()
+    parser = transformers.HfArgumentParser((TrainerCliArgs))
+    parsed_cli_args, _ = parser.parse_args_into_dataclasses(
+        return_remaining_strings=True
+    )
+    parsed_cli_args.merge_lora = True
+
+    parsed_cfg = load_cfg(
+        config,
+        **kwargs,
+    )
+
+    fsdp_dir = Path(parsed_cfg.output_dir) / "pytorch_model_fsdp_0"
+    merge_fsdp_weights(
+        checkpoint_dir=str(fsdp_dir),
+        output_path=str(Path(parsed_cfg.output_dir) / "merged"),
+        safe_serialization=True,
+    )
+
+
+if __name__ == "__main__":
+    load_dotenv()
+    fire.Fire(do_cli)

src/axolotl/cli/preprocess.py

@@ -2,6 +2,7 @@
 CLI to run training on a model
 """
 import logging
+import warnings
 from pathlib import Path
 from typing import Union
 
@@ -9,6 +10,7 @@ import fire
 import transformers
 from accelerate import init_empty_weights
 from colorama import Fore
+from dotenv import load_dotenv
 from transformers import AutoModelForCausalLM
 
 from axolotl.cli import (
@@ -21,10 +23,7 @@ from axolotl.cli import (
 )
 from axolotl.common.cli import PreprocessCliArgs
 from axolotl.common.const import DEFAULT_DATASET_PREPARED_PATH
-from axolotl.prompt_strategies.sharegpt import (
-    register_chatml_template,
-    register_llama3_template,
-)
+from axolotl.utils.trainer import disable_datasets_caching
 
 LOG = logging.getLogger("axolotl.cli.preprocess")
 
@@ -41,23 +40,6 @@ def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs):
         return_remaining_strings=True
     )
 
-    if parsed_cfg.chat_template == "chatml":
-        if parsed_cfg.default_system_message:
-            LOG.info(
-                f"ChatML set. Adding default system message: {parsed_cfg.default_system_message}"
-            )
-            register_chatml_template(parsed_cfg.default_system_message)
-        else:
-            register_chatml_template()
-    elif parsed_cfg.chat_template == "llama3":
-        if parsed_cfg.default_system_message:
-            LOG.info(
-                f"LLaMA-3 set. Adding default system message: {parsed_cfg.default_system_message}"
-            )
-            register_llama3_template(parsed_cfg.default_system_message)
-        else:
-            register_llama3_template()
-
     if not parsed_cfg.dataset_prepared_path:
         msg = (
             Fore.RED
@@ -68,15 +50,27 @@ def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs):
         LOG.warning(msg)
         parsed_cfg.dataset_prepared_path = DEFAULT_DATASET_PREPARED_PATH
 
-    if parsed_cfg.rl:  # and parsed_cfg.rl != "orpo":
-        load_rl_datasets(cfg=parsed_cfg, cli_args=parsed_cli_args)
-    else:
-        load_datasets(cfg=parsed_cfg, cli_args=parsed_cli_args)
+    with disable_datasets_caching():
+        if parsed_cfg.rl:  # and parsed_cfg.rl != "orpo":
+            load_rl_datasets(cfg=parsed_cfg, cli_args=parsed_cli_args)
+        else:
+            load_datasets(cfg=parsed_cfg, cli_args=parsed_cli_args)
 
     if parsed_cli_args.download:
         model_name = parsed_cfg.base_model
-        with init_empty_weights():
-            AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=True)
+        with warnings.catch_warnings():
+            # there are a bunch of useless UserWarnings about
+            # "copying from a non-meta parameter in the checkpoint to a meta parameter in the current model"
+            warnings.simplefilter("ignore")
+            with init_empty_weights(include_buffers=True):
+                # fmt: off
+                try:
+                    AutoModelForCausalLM.from_pretrained(
+                        model_name, trust_remote_code=True
+                    )
+                except Exception as exc:  # pylint: disable=broad-exception-caught,unused-variable  # nosec B110  # noqa F841
+                    pass
+                # fmt: on
 
     LOG.info(
         Fore.GREEN
@@ -86,4 +80,5 @@ def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs):
 
 
 if __name__ == "__main__":
+    load_dotenv()
     fire.Fire(do_cli)

src/axolotl/cli/shard.py

@@ -7,6 +7,7 @@ from typing import Union
 
 import fire
 import transformers
+from dotenv import load_dotenv
 
 from axolotl.cli import load_cfg, print_axolotl_text_art
 from axolotl.common.cli import TrainerCliArgs, load_model_and_tokenizer
@@ -40,4 +41,5 @@ def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs):
 
 
 if __name__ == "__main__":
+    load_dotenv()
     fire.Fire(do_cli)

src/axolotl/cli/train.py

@@ -3,12 +3,11 @@ CLI to run training on a model
 """
 import logging
 from pathlib import Path
-from typing import Tuple, Union
+from typing import Union
 
 import fire
+from dotenv import load_dotenv
 from transformers.hf_argparser import HfArgumentParser
-from transformers.modeling_utils import PreTrainedModel
-from transformers.tokenization_utils import PreTrainedTokenizer
 
 from axolotl.cli import (
     check_accelerate_default_config,
@@ -19,10 +18,7 @@ from axolotl.cli import (
     print_axolotl_text_art,
 )
 from axolotl.common.cli import TrainerCliArgs
-from axolotl.prompt_strategies.sharegpt import (
-    register_chatml_template,
-    register_llama3_template,
-)
+from axolotl.integrations.base import PluginManager
 from axolotl.train import train
 
 LOG = logging.getLogger("axolotl.cli.train")
@@ -38,33 +34,25 @@ def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs):
     return do_train(parsed_cfg, parsed_cli_args)
 
 
-def do_train(cfg, cli_args) -> Tuple[PreTrainedModel, PreTrainedTokenizer]:
+def do_train(cfg, cli_args) -> None:
     print_axolotl_text_art()
     check_accelerate_default_config()
     check_user_token()
-    if cfg.chat_template == "chatml" and cfg.default_system_message:
-        LOG.info(
-            f"ChatML set. Adding default system message: {cfg.default_system_message}"
-        )
-        register_chatml_template(cfg.default_system_message)
-    else:
-        register_chatml_template()
-
-    if cfg.chat_template == "llama3" and cfg.default_system_message:
-        LOG.info(
-            f"LLaMA-3 set. Adding default system message: {cfg.default_system_message}"
-        )
-        register_llama3_template(cfg.default_system_message)
-    else:
-        register_llama3_template()
 
     if cfg.rl:  # and cfg.rl != "orpo":
         dataset_meta = load_rl_datasets(cfg=cfg, cli_args=cli_args)
     else:
         dataset_meta = load_datasets(cfg=cfg, cli_args=cli_args)
 
-    return train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
+    model, tokenizer = train(cfg=cfg, cli_args=cli_args, dataset_meta=dataset_meta)
+    plugin_manager = PluginManager.get_instance()
+
+    del model
+    del tokenizer
+
+    plugin_manager.post_train_unload(cfg)
 
 
 if __name__ == "__main__":
+    load_dotenv()
     fire.Fire(do_cli)

src/axolotl/cli/utils.py

@@ -0,0 +1,218 @@
+"""Utility methods for axoltl CLI."""
+import concurrent.futures
+import dataclasses
+import hashlib
+import json
+import logging
+from pathlib import Path
+from types import NoneType
+from typing import Any, Dict, List, Optional, Tuple, Type, Union, get_args, get_origin
+
+import click
+import requests
+from pydantic import BaseModel
+
+LOG = logging.getLogger("axolotl.cli.utils")
+
+
+def add_options_from_dataclass(config_class: Type[Any]):
+    """Create Click options from the fields of a dataclass."""
+
+    def decorator(function):
+        # Process dataclass fields in reverse order for correct option ordering
+        for field in reversed(dataclasses.fields(config_class)):
+            field_type = field.type
+
+            if get_origin(field_type) is Union and type(None) in get_args(field_type):
+                field_type = next(
+                    t for t in get_args(field_type) if not isinstance(t, NoneType)
+                )
+
+            if field_type == bool:
+                field_name = field.name.replace("_", "-")
+                option_name = f"--{field_name}/--no-{field_name}"
+                function = click.option(
+                    option_name,
+                    default=field.default,
+                    help=field.metadata.get("description"),
+                )(function)
+            else:
+                option_name = f"--{field.name.replace('_', '-')}"
+                function = click.option(
+                    option_name,
+                    type=field_type,
+                    default=field.default,
+                    help=field.metadata.get("description"),
+                )(function)
+        return function
+
+    return decorator
+
+
+def add_options_from_config(config_class: Type[BaseModel]):
+    """Create Click options from the fields of a Pydantic model."""
+
+    def decorator(function):
+        # Process model fields in reverse order for correct option ordering
+        for name, field in reversed(config_class.model_fields.items()):
+            if field.annotation == bool:
+                field_name = name.replace("_", "-")
+                option_name = f"--{field_name}/--no-{field_name}"
+                function = click.option(
+                    option_name, default=None, help=field.description
+                )(function)
+            else:
+                option_name = f"--{name.replace('_', '-')}"
+                function = click.option(
+                    option_name, default=None, help=field.description
+                )(function)
+        return function
+
+    return decorator
+
+
+def build_command(base_cmd: List[str], options: Dict[str, Any]) -> List[str]:
+    """Build command list from base command and options."""
+    cmd = base_cmd.copy()
+
+    for key, value in options.items():
+        if value is None:
+            continue
+
+        key = key.replace("_", "-")
+
+        if isinstance(value, bool):
+            if value:
+                cmd.append(f"--{key}")
+        else:
+            cmd.extend([f"--{key}", str(value)])
+
+    return cmd
+
+
+def download_file(
+    file_info: tuple, raw_base_url: str, dest_path: Path, dir_prefix: str
+) -> Tuple[str, str]:
+    """
+    Download a single file and return its processing status.
+
+    Args:
+        file_info: Tuple of (file_path, remote_sha)
+        raw_base_url: Base URL for raw GitHub content
+        dest_path: Local destination directory
+        dir_prefix: Directory prefix to filter files
+
+    Returns:
+        Tuple of (file_path, status) where status is 'new', 'updated', or 'unchanged'
+    """
+    file_path, remote_sha = file_info
+    raw_url = f"{raw_base_url}/{file_path}"
+    dest_file = dest_path / file_path.split(dir_prefix)[-1]
+
+    # Check if file exists and needs updating
+    if dest_file.exists():
+        with open(dest_file, "rb") as file:
+            content = file.read()
+            # Calculate git blob SHA
+            blob = b"blob " + str(len(content)).encode() + b"\0" + content
+            local_sha = hashlib.sha1(blob, usedforsecurity=False).hexdigest()
+
+        if local_sha == remote_sha:
+            print(f"Skipping {file_path} (unchanged)")
+            return file_path, "unchanged"
+
+        print(f"Updating {file_path}")
+        status = "new"
+    else:
+        print(f"Downloading {file_path}")
+        status = "new"
+
+    # Create directories if needed
+    dest_file.parent.mkdir(parents=True, exist_ok=True)
+
+    # Download and save file
+    try:
+        response = requests.get(raw_url, timeout=30)
+        response.raise_for_status()
+
+        with open(dest_file, "wb") as file:
+            file.write(response.content)
+
+        return file_path, status
+    except (requests.RequestException, IOError) as request_error:
+        print(f"Error downloading {file_path}: {str(request_error)}")
+        return file_path, "error"
+
+
+def fetch_from_github(
+    dir_prefix: str, dest_dir: Optional[str] = None, max_workers: int = 5
+) -> None:
+    """
+    Sync files from a specific directory in the GitHub repository.
+    Only downloads files that don't exist locally or have changed.
+
+    Args:
+        dir_prefix: Directory prefix to filter files (e.g., 'examples/', 'deepspeed_configs/')
+        dest_dir: Local destination directory
+        max_workers: Maximum number of concurrent downloads
+    """
+    api_url = "https://api.github.com/repos/axolotl-ai-cloud/axolotl/git/trees/main?recursive=1"
+    raw_base_url = "https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main"
+
+    # Get repository tree with timeout
+    response = requests.get(api_url, timeout=30)
+    response.raise_for_status()
+    tree = json.loads(response.text)
+
+    # Filter for files and get their SHA
+    files = {
+        item["path"]: item["sha"]
+        for item in tree["tree"]
+        if item["type"] == "blob" and item["path"].startswith(dir_prefix)
+    }
+
+    if not files:
+        raise click.ClickException(f"No files found in {dir_prefix}")
+
+    # Default destination directory is the last part of dir_prefix
+    default_dest = Path(dir_prefix.rstrip("/"))
+    dest_path = Path(dest_dir) if dest_dir else default_dest
+
+    # Keep track of processed files for summary
+    files_processed: Dict[str, List[str]] = {
+        "new": [],
+        "updated": [],
+        "unchanged": [],
+        "error": [],
+    }
+
+    # Process files in parallel using ThreadPoolExecutor
+    with concurrent.futures.ThreadPoolExecutor(max_workers=max_workers) as executor:
+        future_to_file = {
+            executor.submit(
+                download_file,
+                (file_path, remote_sha),
+                raw_base_url,
+                dest_path,
+                dir_prefix,
+            ): file_path
+            for file_path, remote_sha in files.items()
+        }
+
+        # Process completed tasks as they finish
+        for future in concurrent.futures.as_completed(future_to_file):
+            file_path = future_to_file[future]
+            try:
+                file_path, status = future.result()
+                files_processed[status].append(file_path)
+            except (requests.RequestException, IOError) as request_error:
+                print(f"Error processing {file_path}: {str(request_error)}")
+                files_processed["error"].append(file_path)
+
+    # Log summary
+    LOG.info("\nSync Summary:")
+    LOG.info(f"New files: {len(files_processed['new'])}")
+    LOG.info(f"Updated files: {len(files_processed['updated'])}")
+    LOG.info(f"Unchanged files: {len(files_processed['unchanged'])}")
+    if files_processed["error"]:
+        LOG.info(f"Failed files: {len(files_processed['error'])}")

src/axolotl/common/architectures.py

@@ -0,0 +1,15 @@
+"""
+Common architecture specific constants
+"""
+
+MOE_ARCH_BLOCK = {
+    "dbrx": "DbrxFFN",
+    "jamba": "JambaSparseMoeBlock",
+    "jetmoe": [
+        "JetMoeMoA",
+        "JetMoeMoE",
+    ],
+    "mixtral": "MixtralSparseMoeBlock",
+    "qwen2_moe": "Qwen2MoeSparseMoeBlock",
+    "deepseek_v2": "DeepseekV2MoE",
+}