Update __init__.py

* fix: add multiple tips about eos_token masking

* fix: format dataset preprocessing doc

* Update docs/dataset-formats/conversation.qmd

Co-authored-by: salman <salman.mohammadi@outlook.com>

---------

Co-authored-by: salman <salman.mohammadi@outlook.com>

.github/CONTRIBUTING.md

@@ -15,7 +15,7 @@ First of all, thank you for your interest in contributing to axolotl! We appreci
   - [Commit Messages](#commit-messages)
 - [Additional Resources](#additional-resources)
 
-## Code of Conductcode
+## Code of Conduct
 
 All contributors are expected to adhere to our [Code of Conduct](CODE_OF_CONDUCT.md). Please read it before participating in the axolotl community.
 

.github/workflows/base.yml

@@ -22,24 +22,6 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - cuda: "121"
-            cuda_version: 12.1.1
-            cudnn_version: 8
-            python_version: "3.10"
-            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: 12.1.1
-            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: 2.4.1
-            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
           - cuda: "124"
             cuda_version: 12.4.1
             cudnn_version: ""
@@ -52,6 +34,12 @@ jobs:
             python_version: "3.11"
             pytorch: 2.5.1
             torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+          - cuda: "124"
+            cuda_version: 12.4.1
+            cudnn_version: ""
+            python_version: "3.11"
+            pytorch: 2.6.0
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
     steps:
       - name: Checkout
         uses: actions/checkout@v4

.github/workflows/docs.yml

@@ -19,7 +19,7 @@ jobs:
         - name: Setup Python
           uses: actions/setup-python@v5
           with:
-            python-version: '3.10'
+            python-version: '3.11'
         - name: install dependencies
           run: |
             python3 -m pip install jupyter

.github/workflows/lint.yml

@@ -1,6 +1,7 @@
 name: lint
 on:
   # check on PRs, and manual triggers
+  merge_group:
   pull_request:
       paths:
        - '**.py'
@@ -18,6 +19,6 @@ jobs:
       - uses: actions/checkout@v4
       - uses: actions/setup-python@v5
         with:
-          python-version: "3.10"
+          python-version: "3.11"
           cache: 'pip' # caching pip dependencies
       - uses: pre-commit/action@v3.0.1

.github/workflows/main.yml

@@ -15,17 +15,6 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - cuda: 121
-            cuda_version: 12.1.1
-            python_version: "3.10"
-            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: 124
             cuda_version: 12.4.1
             python_version: "3.11"
@@ -35,6 +24,12 @@ jobs:
             cuda_version: 12.4.1
             python_version: "3.11"
             pytorch: 2.5.1
+            axolotl_extras: vllm
+            is_latest: true
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.6.0
             axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:
@@ -82,17 +77,6 @@ jobs:
     strategy:
       matrix:
         include:
-          - cuda: 121
-            cuda_version: 12.1.1
-            python_version: "3.10"
-            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: 124
             cuda_version: 12.4.1
             python_version: "3.11"
@@ -103,6 +87,7 @@ jobs:
             python_version: "3.11"
             pytorch: 2.5.1
             axolotl_extras:
+            is_latest: true
     runs-on: axolotl-gpu-runner
     steps:
       - name: Checkout
@@ -145,10 +130,10 @@ jobs:
     strategy:
       matrix:
         include:
-          - cuda: 121
-            cuda_version: 12.1.1
+          - cuda: 124
+            cuda_version: 12.4.1
             python_version: "3.11"
-            pytorch: 2.3.1
+            pytorch: 2.4.1
             axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:

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

@@ -4,6 +4,10 @@ on:
   pull_request:
     paths:
       - 'tests/e2e/multigpu/*.py'
+      - 'requirements.txt'
+      - 'setup.py'
+      - 'pyproject.toml'
+      - '.github/workflows/multi-gpu-e2e.yml'
   workflow_dispatch:
   schedule:
     - cron: '0 0 * * 1,4'  # Runs at 00:00 UTC every monday & thursday
@@ -20,23 +24,25 @@ jobs:
       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:
+            axolotl_extras:  # no vllm support for 2.4.1
             num_gpus: 2
             nightly_build: "true"
           - cuda: 124
             cuda_version: 12.4.1
             python_version: "3.11"
             pytorch: 2.5.1
+            axolotl_extras: vllm
+            num_gpus: 2
+            nightly_build: "true"
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.6.0
+            # awaiting vllm#12721
             axolotl_extras:
             num_gpus: 2
             nightly_build: "true"
@@ -48,11 +54,11 @@ jobs:
       - name: Install Python
         uses: actions/setup-python@v5
         with:
-          python-version: "3.10"
+          python-version: "3.11"
       - name: Install Modal
         run: |
           python -m pip install --upgrade pip
-          pip install modal==0.63.64 jinja2
+          pip install modal==0.71.8 jinja2
       - name: Update env vars
         run: |
           echo "BASE_TAG=main-base-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}" >> $GITHUB_ENV

.github/workflows/nightlies.yml

@@ -12,17 +12,6 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - cuda: 121
-            cuda_version: 12.1.1
-            python_version: "3.10"
-            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: 124
             cuda_version: 12.4.1
             python_version: "3.11"
@@ -33,6 +22,11 @@ jobs:
             python_version: "3.11"
             pytorch: 2.5.1
             axolotl_extras:
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.6.0
+            axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:
       - name: Checkout
@@ -76,17 +70,6 @@ jobs:
     strategy:
       matrix:
         include:
-          - cuda: 121
-            cuda_version: 12.1.1
-            python_version: "3.10"
-            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: 124
             cuda_version: 12.4.1
             python_version: "3.11"

.github/workflows/pypi.yml

@@ -13,10 +13,13 @@ jobs:
     permissions:
       contents: write
     steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+
       - 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
+        run: gh release create "$GITHUB_REF_NAME" --generate-notes
   pypi-publish:
     name: Upload release to PyPI
     runs-on: ubuntu-latest
@@ -33,12 +36,12 @@ jobs:
       - name: Setup Python
         uses: actions/setup-python@v5
         with:
-          python-version: "3.10"
+          python-version: "3.11"
 
       - name: Install dependencies
         run: |
           pip3 install wheel packaging
-          pip3 install -e .
+          pip3 install --no-build-isolation -e .
           pip3 install -r requirements-dev.txt -r requirements-tests.txt
 
       - name: Extract tag name

.github/workflows/tests-nightly.yml

@@ -12,7 +12,7 @@ jobs:
       - uses: actions/checkout@v4
       - uses: actions/setup-python@v5
         with:
-          python-version: "3.10"
+          python-version: "3.11"
           cache: 'pip' # caching pip dependencies
       - uses: pre-commit/action@v3.0.1
         env:
@@ -25,13 +25,8 @@ jobs:
       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"
+        python_version: ["3.11"]
+        pytorch_version: ["2.4.1", "2.5.1", "2.6.0"]
     timeout-minutes: 20
 
     steps:
@@ -44,6 +39,11 @@ jobs:
           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 }} --index-url https://download.pytorch.org/whl/cpu
@@ -60,11 +60,15 @@ jobs:
         run: |
           pip3 install --upgrade pip
           pip3 install --upgrade packaging
-          pip3 install -U -e .
+          pip3 install --no-build-isolation -U -e .
           python scripts/unsloth_install.py | sh
           python scripts/cutcrossentropy_install.py | sh
           pip3 install -r requirements-dev.txt -r requirements-tests.txt
 
+      - name: Make sure PyTorch version wasn't clobbered
+        run: |
+          python -c "import torch; assert '${{ matrix.pytorch_version }}' in torch.__version__"
+
       - name: Ensure axolotl CLI was installed
         run: |
           axolotl --help
@@ -89,13 +93,6 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - cuda: 121
-            cuda_version: 12.1.1
-            python_version: "3.10"
-            pytorch: 2.3.1
-            num_gpus: 1
-            axolotl_extras: mamba-ssm
-            nightly_build: "true"
           - cuda: 124
             cuda_version: 12.4.1
             python_version: "3.11"
@@ -110,17 +107,24 @@ jobs:
             num_gpus: 1
             axolotl_extras:
             nightly_build: "true"
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.6.0
+            num_gpus: 1
+            axolotl_extras:
+            nightly_build: "true"
     steps:
       - name: Checkout
         uses: actions/checkout@v4
       - name: Install Python
         uses: actions/setup-python@v5
         with:
-          python-version: "3.10"
+          python-version: "3.11"
       - name: Install Modal
         run: |
           python -m pip install --upgrade pip
-          pip install modal==0.63.64 jinja2
+          pip install modal==0.71.8 jinja2
       - name: Update env vars
         run: |
           echo "BASE_TAG=main-base-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}" >> $GITHUB_ENV

.github/workflows/tests.yml

@@ -1,6 +1,7 @@
 name: Tests
 on:
   # check on push/merge to main, PRs, and manual triggers
+  merge_group:
   push:
     branches:
       - "main"
@@ -34,7 +35,7 @@ jobs:
       - uses: actions/checkout@v4
       - uses: actions/setup-python@v5
         with:
-          python-version: "3.10"
+          python-version: "3.11"
           cache: 'pip' # caching pip dependencies
       - uses: pre-commit/action@v3.0.1
         env:
@@ -47,19 +48,23 @@ jobs:
       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"
+        python_version: ["3.11"]
+        pytorch_version: ["2.4.1", "2.5.1", "2.6.0"]
     timeout-minutes: 20
 
     steps:
       - name: Check out repository code
         uses: actions/checkout@v4
 
+      - name: Restore HF cache
+        id: hf-cache-restore
+        uses: actions/cache/restore@v4
+        with:
+          path: |
+            /home/runner/.cache/huggingface/hub/datasets--*
+            /home/runner/.cache/huggingface/hub/models--*
+          key: ${{ runner.os }}-hf-hub-cache-${{ hashFiles('**/conftest.py') }}
+
       - name: Setup Python
         uses: actions/setup-python@v5
         with:
@@ -78,24 +83,37 @@ jobs:
       - name: Install dependencies
         run: |
           pip3 show torch
-          pip3 install -U -e .
+          pip3 install --no-build-isolation -U -e .
           python scripts/unsloth_install.py | sh
           python scripts/cutcrossentropy_install.py | sh
           pip3 install -r requirements-dev.txt -r requirements-tests.txt
 
+      - name: Make sure PyTorch version wasn't clobbered
+        run: |
+          python -c "import torch; assert '${{ matrix.pytorch_version }}' in torch.__version__"
+
       - name: Ensure axolotl CLI was installed
         run: |
           axolotl --help
 
       - name: Run tests
         run: |
-          pytest -n8 --dist loadfile --ignore=tests/e2e/ --ignore=tests/patched/ tests/
-          pytest tests/patched/
+          pytest -v -n8 --dist loadfile --ignore=tests/e2e/ --ignore=tests/patched/ tests/
+          pytest -v tests/patched/
 
       - name: cleanup pip cache
         run: |
           find "$(pip cache dir)/http-v2" -type f -mtime +14 -exec rm {} \;
 
+      - name: Save HF cache
+        id: hf-cache
+        uses: actions/cache/save@v4
+        with:
+          path: |
+            /home/runner/.cache/huggingface/hub/datasets--*
+            /home/runner/.cache/huggingface/hub/models--*
+          key: ${{ steps.hf-cache-restore.outputs.cache-primary-key }}
+
   pytest-sdist:
     name: PyTest from Source Dist
     runs-on: ubuntu-latest
@@ -104,13 +122,22 @@ jobs:
       max-parallel: 1
       matrix:
         python_version: ["3.11"]
-        pytorch_version: ["2.4.1", "2.5.1"]
+        pytorch_version: ["2.4.1", "2.5.1", "2.6.0"]
     timeout-minutes: 20
 
     steps:
       - name: Check out repository code
         uses: actions/checkout@v4
 
+      - name: Restore HF cache
+        id: hf-cache-restore
+        uses: actions/cache/restore@v4
+        with:
+          path: |
+            /home/runner/.cache/huggingface/hub/datasets--*
+            /home/runner/.cache/huggingface/hub/models--*
+          key: ${{ runner.os }}-hf-hub-cache-${{ hashFiles('**/conftest.py') }}
+
       - name: Setup Python
         uses: actions/setup-python@v5
         with:
@@ -120,7 +147,7 @@ jobs:
       - name: upgrade pip
         run: |
           pip3 install --upgrade pip
-          pip3 install --upgrade packaging setuptools wheel
+          pip3 install --upgrade packaging setuptools setuptools_scm build wheel
 
       - name: Install PyTorch
         run: |
@@ -129,25 +156,38 @@ jobs:
       - name: Install dependencies
         run: |
           pip3 show torch
-          python3 setup.py sdist
-          pip3 install dist/axolotl*.tar.gz
+          python -m build --no-isolation --sdist
+          pip3 install --no-build-isolation 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: Make sure PyTorch version wasn't clobbered
+        run: |
+          python -c "import torch; assert '${{ matrix.pytorch_version }}' in torch.__version__"
+
       - name: Ensure axolotl CLI was installed
         run: |
           axolotl --help
 
       - name: Run tests
         run: |
-          pytest -n8 --dist loadfile --ignore=tests/e2e/ --ignore=tests/patched/ tests/
-          pytest tests/patched/
+          pytest -v -n8 --dist loadfile --ignore=tests/e2e/ --ignore=tests/patched/ tests/
+          pytest -v tests/patched/
 
       - name: cleanup pip cache
         run: |
           find "$(pip cache dir)/http-v2" -type f -mtime +14 -exec rm {} \;
 
+      - name: Save HF cache
+        id: hf-cache
+        uses: actions/cache/save@v4
+        with:
+          path: |
+            /home/runner/.cache/huggingface/hub/datasets--*
+            /home/runner/.cache/huggingface/hub/models--*
+          key: ${{ steps.hf-cache-restore.outputs.cache-primary-key }}
+
   docker-e2e-tests-1st:
     if: ${{ ! contains(github.event.commits[0].message, '[skip e2e]') && github.repository_owner == 'axolotl-ai-cloud' }}
     # this job needs to be run on self-hosted GPU runners...
@@ -162,20 +202,20 @@ jobs:
           - cuda: 124
             cuda_version: 12.4.1
             python_version: "3.11"
-            pytorch: 2.4.1
+            pytorch: 2.5.1
             num_gpus: 1
-            axolotl_extras:
+            axolotl_extras: vllm
     steps:
       - name: Checkout
         uses: actions/checkout@v4
       - name: Install Python
         uses: actions/setup-python@v5
         with:
-          python-version: "3.10"
+          python-version: "3.11"
       - name: Install Modal
         run: |
           python -m pip install --upgrade pip
-          pip install modal==0.63.64 jinja2
+          pip install modal==0.71.8 jinja2
       - name: Update env vars
         run: |
           echo "BASE_TAG=main-base-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}" >> $GITHUB_ENV
@@ -183,6 +223,7 @@ jobs:
           echo "AXOLOTL_ARGS=${{ matrix.axolotl_args}}" >> $GITHUB_ENV
           echo "AXOLOTL_EXTRAS=${{ matrix.axolotl_extras}}" >> $GITHUB_ENV
           echo "CUDA=${{ matrix.cuda }}" >> $GITHUB_ENV
+          echo "MODAL_IMAGE_BUILDER_VERSION=2024.10" >> $GITHUB_ENV
           echo "N_GPUS=${{ matrix.num_gpus }}" >> $GITHUB_ENV
       - name: Run tests job on Modal
         run: |
@@ -199,16 +240,16 @@ jobs:
       fail-fast: false
       matrix:
         include:
-          - cuda: 121
-            cuda_version: 12.1.1
-            python_version: "3.10"
-            pytorch: 2.3.1
-            num_gpus: 1
-            axolotl_extras: mamba-ssm
           - cuda: 124
             cuda_version: 12.4.1
             python_version: "3.11"
-            pytorch: 2.5.1
+            pytorch: 2.4.1
+            num_gpus: 1
+            axolotl_extras:
+          - cuda: 124
+            cuda_version: 12.4.1
+            python_version: "3.11"
+            pytorch: 2.6.0
             num_gpus: 1
             axolotl_extras:
     steps:
@@ -217,11 +258,11 @@ jobs:
       - name: Install Python
         uses: actions/setup-python@v5
         with:
-          python-version: "3.10"
+          python-version: "3.11"
       - name: Install Modal
         run: |
           python -m pip install --upgrade pip
-          pip install modal==0.63.64 jinja2
+          pip install modal==0.71.8 jinja2
       - name: Update env vars
         run: |
           echo "BASE_TAG=main-base-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}" >> $GITHUB_ENV
@@ -229,6 +270,7 @@ jobs:
           echo "AXOLOTL_ARGS=${{ matrix.axolotl_args}}" >> $GITHUB_ENV
           echo "AXOLOTL_EXTRAS=${{ matrix.axolotl_extras}}" >> $GITHUB_ENV
           echo "CUDA=${{ matrix.cuda }}" >> $GITHUB_ENV
+          echo "MODAL_IMAGE_BUILDER_VERSION=2024.10" >> $GITHUB_ENV
           echo "N_GPUS=${{ matrix.num_gpus }}" >> $GITHUB_ENV
       - name: Run tests job on Modal
         run: |

.gitignore

@@ -1,6 +1,7 @@
 **/axolotl.egg-info
 configs
 last_run_prepared/
+outputs
 .vscode
 _site/
 

.pre-commit-config.yaml

@@ -19,11 +19,11 @@ repos:
     hooks:
       - id: isort
 -   repo: https://github.com/PyCQA/flake8
-    rev: 6.0.0
+    rev: 6.1.0
     hooks:
     - id: flake8
 -   repo: https://github.com/PyCQA/pylint
-    rev: v2.17.4
+    rev: v3.3.0
     hooks:
     - id: pylint
 -   repo: https://github.com/pre-commit/mirrors-mypy

.pylintrc

@@ -1,5 +1,5 @@
 [MASTER]
-init-hook="from pylint.config import find_pylintrc; import os, sys; sys.path.append(os.path.dirname(find_pylintrc()))"
+init-hook="from pylint.config import find_default_config_files; import sys; sys.path.append(next(find_default_config_files()).parent.as_posix())"
 
 [TYPECHECK]
 
@@ -12,3 +12,4 @@ generated-members=numpy.*, torch.*
 disable=missing-function-docstring, line-too-long, import-error,
     too-many-arguments, too-many-locals, too-many-statements, too-many-branches, too-few-public-methods,
     too-many-instance-attributes, fixme, import-outside-toplevel, logging-fstring-interpolation,
+    too-many-positional-arguments, possibly-used-before-assignment

MANIFEST.in

@@ -1,4 +1,5 @@
 include requirements.txt
 include README.md
 include LICENSE
+include src/setuptools_axolotl_dynamic_dependencies.py
 recursive-include axolotl *.py

README.md

@@ -1,8 +1,8 @@
 <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%;">
+        <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/887513285d98132142bf5db2a74eb5e0928787f1/image/axolotl_logo_digital_white.svg">
+        <source media="(prefers-color-scheme: light)" srcset="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/887513285d98132142bf5db2a74eb5e0928787f1/image/axolotl_logo_digital_black.svg">
+        <img alt="Axolotl" src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/887513285d98132142bf5db2a74eb5e0928787f1/image/axolotl_logo_digital_black.svg" width="400" height="104" style="max-width: 100%;">
     </picture>
 </p>
 
@@ -10,218 +10,109 @@
     <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>
+    <br/>
+    <a href="https://github.com/axolotl-ai-cloud/axolotl/graphs/contributors"><img src="https://img.shields.io/github/contributors-anon/axolotl-ai-cloud/axolotl?color=yellow&style=flat-square" alt="contributors" style="height: 20px;"></a>
     <img src="https://img.shields.io/github/stars/axolotl-ai-cloud/axolotl" alt="GitHub Repo stars">
-</p>
-<p align="center">
+    <br/>
+    <a href="https://discord.com/invite/HhrNrHJPRb"><img src="https://img.shields.io/badge/discord-7289da.svg?style=flat-square&logo=discord" alt="discord" style="height: 20px;"></a>
+    <a href="https://twitter.com/axolotl_ai"><img src="https://img.shields.io/twitter/follow/axolotl_ai?style=social" alt="twitter" style="height: 20px;"></a>
+    <br/>
     <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">
+    <a href="https://www.phorm.ai/query?projectId=e315ba4a-4e14-421f-ab05-38a1f9076f25">
+    <img alt="phorm.ai" src="https://img.shields.io/badge/Phorm-Ask_AI-%23F2777A.svg?&logo=data:image/svg+xml;base64,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">
+  </a>
 </p>
 
-Axolotl is a tool designed to streamline the fine-tuning of various AI models, offering support for multiple configurations and architectures.
+Axolotl is a tool designed to streamline post-training for various AI models.
+Post-training refers to any modifications or additional training performed on
+pre-trained models - including full model fine-tuning, parameter-efficient tuning (like
+LoRA and QLoRA), supervised fine-tuning (SFT), instruction tuning, and alignment
+techniques. With support for multiple model architectures and training configurations,
+Axolotl makes it easy to get started with these techniques.
+
+Axolotl is designed to work with YAML config files that contain everything you need to
+preprocess a dataset, train or fine-tune a model, run model inference or evaluation,
+and much more.
 
 Features:
+
 - Train various Huggingface models such as llama, pythia, falcon, mpt
 - Supports fullfinetune, lora, qlora, relora, and gptq
 - Customize configurations using a simple yaml file or CLI overwrite
 - Load different dataset formats, use custom formats, or bring your own tokenized datasets
-- Integrated with xformer, flash attention, [liger kernel](https://github.com/linkedin/Liger-Kernel), rope scaling, and multipacking
+- Integrated with [xformers](https://github.com/facebookresearch/xformers), flash attention, [liger kernel](https://github.com/linkedin/Liger-Kernel), rope scaling, and multipacking
 - Works with single GPU or multiple GPUs via FSDP or Deepspeed
 - Easily run with Docker locally or on the cloud
 - Log results and optionally checkpoints to wandb, mlflow or Comet
 - And more!
 
-<a href="https://www.phorm.ai/query?projectId=e315ba4a-4e14-421f-ab05-38a1f9076f25">
-  <img alt="phorm.ai" src="https://img.shields.io/badge/Phorm-Ask_AI-%23F2777A.svg?&logo=data:image/svg+xml;base64,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">
-</a>
+## 🚀 Quick Start
 
-<table>
-<tr>
-<td>
+**Requirements**:
 
-## Table of Contents
-- [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-)
+- NVIDIA GPU (Ampere or newer for `bf16` and Flash Attention) or AMD GPU
+- Python 3.11
+- PyTorch ≥2.4.1
 
-</td>
-<td>
-
-<div align="center">
-  <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>
-    </p>
-    <p>
-      Go ahead and Axolotl questions!!
-    </p>
-    <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>
-
-</td>
-</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.
+### Installation
 
 ```bash
-git clone https://github.com/axolotl-ai-cloud/axolotl
-cd axolotl
+pip3 install --no-build-isolation axolotl[flash-attn,deepspeed]
 
-pip3 install packaging ninja
-pip3 install -e '.[flash-attn,deepspeed]'
+# Download example axolotl configs, deepspeed configs
+axolotl fetch examples
+axolotl fetch deepspeed_configs  # OPTIONAL
 ```
 
-### Usage
-```bash
-# preprocess datasets - optional but recommended
-CUDA_VISIBLE_DEVICES="0" python -m axolotl.cli.preprocess examples/openllama-3b/lora.yml
+Other installation approaches are described [here](https://axolotl-ai-cloud.github.io/axolotl/docs/installation.html).
 
-# 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:
+### Your First Fine-tune
 
 ```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)
+# Fetch axolotl examples
 axolotl fetch examples
 
-# Fetch deepspeed config files (stores in "deepspeed_configs/" folder)
-axolotl fetch deepspeed_configs
-
-# Optionally, specify a destination folder
+# Or, specify a custom path
 axolotl fetch examples --dest path/to/folder
+
+# Train a model using LoRA
+axolotl train examples/llama-3/lora-1b.yml
 ```
 
-## Badge ❤🏷️
+That's it! Check out our [Getting Started Guide](https://axolotl-ai-cloud.github.io/axolotl/docs/getting-started.html) for a more detailed walkthrough.
 
-Building something cool with Axolotl? Consider adding a badge to your model card.
+## ✨ Key Features
 
-```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)
-```
+- **Multiple Model Support**: Train various models like LLaMA, Mistral, Mixtral, Pythia, and more
+- **Training Methods**: Full fine-tuning, LoRA, QLoRA, and more
+- **Easy Configuration**: Simple YAML files to control your training setup
+- **Performance Optimizations**: Flash Attention, xformers, multi-GPU training
+- **Flexible Dataset Handling**: Use various formats and custom datasets
+- **Cloud Ready**: Run on cloud platforms or local hardware
 
-[<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)
+## 📚 Documentation
 
-## Sponsors 🤝❤
+- [Installation Options](https://axolotl-ai-cloud.github.io/axolotl/docs/installation.html) - Detailed setup instructions for different environments
+- [Configuration Guide](https://axolotl-ai-cloud.github.io/axolotl/docs/config.html) - Full configuration options and examples
+- [Dataset Guide](https://axolotl-ai-cloud.github.io/axolotl/docs/dataset-formats/) - Supported formats and how to use them
+- [Multi-GPU Training](https://axolotl-ai-cloud.github.io/axolotl/docs/multi-gpu.html)
+- [Multi-Node Training](https://axolotl-ai-cloud.github.io/axolotl/docs/multi-node.html)
+- [Multipacking](https://axolotl-ai-cloud.github.io/axolotl/docs/multipack.html)
+- [FAQ](https://axolotl-ai-cloud.github.io/axolotl/docs/faq.html) - Frequently asked questions
 
-If you love axolotl, consider sponsoring the project by reaching out directly to [wing@axolotl.ai](mailto:wing@axolotl.ai).
+## 🤝 Getting Help
 
----
+- Join our [Discord community](https://discord.gg/HhrNrHJPRb) for support
+- Check out our [Examples](https://github.com/axolotl-ai-cloud/axolotl/tree/main/examples/) directory
+- Read our [Debugging Guide](https://axolotl-ai-cloud.github.io/axolotl/docs/debugging.html)
+- Need dedicated support? Please contact [✉️wing@axolotl.ai](mailto:wing@axolotl.ai) for options
 
-- [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
 
----
+Contributions are welcome! Please see our [Contributing Guide](https://github.com/axolotl-ai-cloud/axolotl/blob/main/.github/CONTRIBUTING.md) for details.
 
-## 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
+## Supported Models
 
 |             | fp16/fp32 | lora | qlora | gptq | gptq w/flash attn | flash attn | xformers attn |
 |-------------|:----------|:-----|-------|------|-------------------|------------|--------------|
@@ -246,523 +137,16 @@ Thanks to all of our contributors to date. Help drive open source AI progress fo
 ❌: not supported
 ❓: untested
 
-## Advanced Setup
+## ❤️ Sponsors
 
-### Environment
+Thank you to our sponsors who help make Axolotl possible:
 
-#### Docker
+- [Modal](https://www.modal.com?utm_source=github&utm_medium=github&utm_campaign=axolotl) - Modal lets you run
+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 large language models, run protein folding simulations, and much more.
 
-  ```bash
-  docker run --gpus '"all"' --rm -it axolotlai/axolotl:main-latest
-  ```
+Interested in sponsoring? Contact us at [wing@axolotl.ai](mailto:wing@axolotl.ai)
 
-  Or run on the current files for development:
+## 📜 License
 
-  ```sh
-  docker compose up -d
-  ```
-
->[!Tip]
-> If you want to debug axolotl or prefer to use Docker as your development environment, see the [debugging guide's section on Docker](docs/debugging.qmd#debugging-with-docker).
-
-  <details>
-
-  <summary>Docker advanced</summary>
-
-  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 axolotlai/axolotl:main-latest
-  ```
-
-  It additionally:
-  * Prevents memory issues when running e.g. deepspeed (e.g. you could hit SIGBUS/signal 7 error) through `--ipc` and `--ulimit` args.
-  * Persists the downloaded HF data (models etc.) and your modifications to axolotl code through `--mount`/`-v` args.
-  * The `--name` argument simply makes it easier to refer to the container in vscode (`Dev Containers: Attach to Running Container...`) or in your terminal.
-  * The `--privileged` flag gives all capabilities to the container.
-  * The `--shm-size 10g` argument increases the shared memory size. Use this if you see `exitcode: -7` errors using deepspeed.
-
-  [More information on nvidia website](https://docs.nvidia.com/deeplearning/frameworks/user-guide/index.html#setincshmem)
-
-  </details>
-
-#### Conda/Pip venv
-  1. Install python >=**3.10**
-
-  2. Install pytorch stable https://pytorch.org/get-started/locally/
-
-  3. Install Axolotl along with python dependencies
-        ```bash
-        pip3 install packaging
-        pip3 install -e '.[flash-attn,deepspeed]'
-        ```
-  4. (Optional) Login to Huggingface to use gated models/datasets.
-        ```bash
-        huggingface-cli login
-        ```
-        Get the token at huggingface.co/settings/tokens
-
-#### Cloud GPU
-
-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)
-- on RunPod use this [direct link](https://runpod.io/gsc?template=v2ickqhz9s&ref=6i7fkpdz)
-
-#### Bare Metal Cloud GPU
-
-##### LambdaLabs
-
-  <details>
-
-  <summary>Click to Expand</summary>
-
-  1. Install python
-  ```bash
-  sudo apt update
-  sudo apt install -y python3.10
-
-  sudo update-alternatives --install /usr/bin/python python /usr/bin/python3.10 1
-  sudo update-alternatives --config python # pick 3.10 if given option
-  python -V # should be 3.10
-
-  ```
-
-  2. Install pip
-  ```bash
-  wget https://bootstrap.pypa.io/get-pip.py
-  python get-pip.py
-  ```
-
-  3. Install Pytorch https://pytorch.org/get-started/locally/
-
-  4. Follow instructions on quickstart.
-
-  5. Run
-  ```bash
-  pip3 install protobuf==3.20.3
-  pip3 install -U --ignore-installed requests Pillow psutil scipy
-  ```
-
-  6. Set path
-  ```bash
-  export LD_LIBRARY_PATH=/usr/lib/x86_64-linux-gnu:$LD_LIBRARY_PATH
-  ```
-  </details>
-
-##### GCP
-
-<details>
-
-<summary>Click to Expand</summary>
-
-Use a Deeplearning linux OS with cuda and pytorch installed. Then follow instructions on quickstart.
-
-Make sure to run the below to uninstall xla.
-```bash
-pip uninstall -y torch_xla[tpu]
-```
-
-</details>
-
-#### Windows
-Please use WSL or Docker!
-
-#### Mac
-
-Use the below instead of the install method in QuickStart.
-```
-pip3 install -e '.'
-```
-More info: [mac.md](/docs/mac.qmd)
-
-#### Google Colab
-
-Please use this example [notebook](examples/colab-notebooks/colab-axolotl-example.ipynb).
-
-#### Launching on public clouds via SkyPilot
-To launch on GPU instances (both on-demand and spot instances) on 7+ clouds (GCP, AWS, Azure, OCI, and more), you can use [SkyPilot](https://skypilot.readthedocs.io/en/latest/index.html):
-
-```bash
-pip install "skypilot-nightly[gcp,aws,azure,oci,lambda,kubernetes,ibm,scp]"  # choose your clouds
-sky check
-```
-
-Get the [example YAMLs](https://github.com/skypilot-org/skypilot/tree/master/llm/axolotl) of using Axolotl to finetune `mistralai/Mistral-7B-v0.1`:
-```
-git clone https://github.com/skypilot-org/skypilot.git
-cd skypilot/llm/axolotl
-```
-
-Use one command to launch:
-```bash
-# On-demand
-HF_TOKEN=xx sky launch axolotl.yaml --env HF_TOKEN
-
-# Managed spot (auto-recovery on preemption)
-HF_TOKEN=xx BUCKET=<unique-name> sky spot launch axolotl-spot.yaml --env HF_TOKEN --env BUCKET
-```
-
-#### Launching on public clouds via dstack
-To launch on GPU instance (both on-demand and spot instances) on public clouds (GCP, AWS, Azure, Lambda Labs, TensorDock, Vast.ai, and CUDO), you can use [dstack](https://dstack.ai/).
-
-Write a job description in YAML as below:
-
-```yaml
-# dstack.yaml
-type: task
-
-image: axolotlai/axolotl-cloud:main-latest
-
-env:
-  - HUGGING_FACE_HUB_TOKEN
-  - WANDB_API_KEY
-
-commands:
-  - accelerate launch -m axolotl.cli.train config.yaml
-
-ports:
-  - 6006
-
-resources:
-  gpu:
-    memory: 24GB..
-    count: 2
-```
-
-then, simply run the job with `dstack run` command. Append `--spot` option if you want spot instance. `dstack run` command will show you the instance with cheapest price across multi cloud services:
-
-```bash
-pip install dstack
-HUGGING_FACE_HUB_TOKEN=xxx WANDB_API_KEY=xxx dstack run . -f dstack.yaml # --spot
-```
-
-For further and fine-grained use cases, please refer to the official [dstack documents](https://dstack.ai/docs/) and the detailed description of [axolotl example](https://github.com/dstackai/dstack/tree/master/examples/fine-tuning/axolotl) on the official repository.
-
-### Dataset
-
-Axolotl supports a variety of dataset formats.  It is recommended to use a JSONL.  The schema of the JSONL depends upon the task and the prompt template you wish to use.  Instead of a JSONL, you can also use a HuggingFace dataset with columns for each JSONL field.
-
-See [the documentation](https://axolotl-ai-cloud.github.io/axolotl/docs/dataset-formats/) for more information on how to use different dataset formats.
-
-### Config
-
-See [examples](examples) for quick start. It is recommended to duplicate and modify to your needs. The most important options are:
-
-- model
-  ```yaml
-  base_model: ./llama-7b-hf # local or huggingface repo
-  ```
-  Note: The code will load the right architecture.
-
-- dataset
-  ```yaml
-  datasets:
-      # huggingface repo
-    - path: vicgalle/alpaca-gpt4
-      type: alpaca
-
-      # huggingface repo with specific configuration/subset
-    - path: EleutherAI/pile
-      name: enron_emails
-      type: completion # format from earlier
-      field: text # Optional[str] default: text, field to use for completion data
-
-      # huggingface repo with multiple named configurations/subsets
-    - path: bigcode/commitpackft
-      name:
-        - ruby
-        - python
-        - typescript
-      type: ... # unimplemented custom format
-
-      # chat_template https://axolotl-ai-cloud.github.io/axolotl/docs/dataset-formats/conversation.html#chat_template
-    - path: ...
-      type: chat_template
-      chat_template: chatml # defaults to tokenizer's chat_template
-
-      # local
-    - path: data.jsonl # or json
-      ds_type: json # see other options below
-      type: alpaca
-
-      # dataset with splits, but no train split
-    - path: knowrohit07/know_sql
-      type: context_qa.load_v2
-      train_on_split: validation
-
-      # loading from s3 or gcs
-      # s3 creds will be loaded from the system default and gcs only supports public access
-    - path: s3://path_to_ds # Accepts folder with arrow/parquet or file path like above. Supports s3, gcs.
-      ...
-
-      # Loading Data From a Public URL
-      # - The file format is `json` (which includes `jsonl`) by default. For different formats, adjust the `ds_type` option accordingly.
-    - path: https://some.url.com/yourdata.jsonl # The URL should be a direct link to the file you wish to load. URLs must use HTTPS protocol, not HTTP.
-      ds_type: json # this is the default, see other options below.
-  ```
-
-- loading
-  ```yaml
-  load_in_4bit: true
-  load_in_8bit: true
-
-  bf16: auto # require >=ampere, auto will detect if your GPU supports this and choose automatically.
-  fp16: # leave empty to use fp16 when bf16 is 'auto'. set to false if you want to fallback to fp32
-  tf32: true # require >=ampere
-
-  bfloat16: true # require >=ampere, use instead of bf16 when you don't want AMP (automatic mixed precision)
-  float16: true # use instead of fp16 when you don't want AMP
-  ```
-  Note: Repo does not do 4-bit quantization.
-
-- lora
-  ```yaml
-  adapter: lora # 'qlora' or leave blank for full finetune
-  lora_r: 8
-  lora_alpha: 16
-  lora_dropout: 0.05
-  lora_target_modules:
-    - q_proj
-    - v_proj
-  ```
-
-#### All Config Options
-
-See [these docs](docs/config.qmd) for all config options.
-
-### Train
-
-Run
-```bash
-accelerate launch -m axolotl.cli.train your_config.yml
-```
-
-> [!TIP]
-> You can also reference a config file that is hosted on a public URL, for example `accelerate launch -m axolotl.cli.train https://yourdomain.com/your_config.yml`
-
-#### Preprocess dataset
-
-You can optionally pre-tokenize dataset with the following before finetuning.
-This is recommended for large datasets.
-
-- Set `dataset_prepared_path:` to a local folder for saving and loading pre-tokenized dataset.
-- (Optional): Set `push_dataset_to_hub: hf_user/repo` to push it to Huggingface.
-- (Optional): Use `--debug` to see preprocessed examples.
-
-```bash
-python -m axolotl.cli.preprocess your_config.yml
-```
-
-#### Multi-GPU
-
-Below are the options available in axolotl for training with multiple GPUs. Note that DeepSpeed
-is the recommended multi-GPU option currently because FSDP may experience
-[loss instability](https://github.com/huggingface/transformers/issues/26498).
-
-##### DeepSpeed
-
-Deepspeed is an optimization suite for multi-gpu systems allowing you to train much larger models than you
-might typically be able to fit into your GPU's VRAM. More information about the various optimization types
-for deepspeed is available at https://huggingface.co/docs/accelerate/main/en/usage_guides/deepspeed#what-is-integrated
-
-We provide several default deepspeed JSON configurations for ZeRO stage 1, 2, and 3.
-
-```yaml
-deepspeed: deepspeed_configs/zero1.json
-```
-
-```shell
-accelerate launch -m axolotl.cli.train examples/llama-2/config.yml --deepspeed deepspeed_configs/zero1.json
-```
-
-##### FSDP
-
-- llama FSDP
-```yaml
-fsdp:
-  - full_shard
-  - auto_wrap
-fsdp_config:
-  fsdp_offload_params: true
-  fsdp_state_dict_type: FULL_STATE_DICT
-  fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer
-```
-
-##### FSDP + QLoRA
-
-Axolotl supports training with FSDP and QLoRA, see [these docs](docs/fsdp_qlora.qmd) for more information.
-
-##### Weights & Biases Logging
-
-Make sure your `WANDB_API_KEY` environment variable is set (recommended) or you login to wandb with `wandb login`.
-
-- wandb options
-```yaml
-wandb_mode:
-wandb_project:
-wandb_entity:
-wandb_watch:
-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:
-
-```yml
-special_tokens:
-  bos_token: "<s>"
-  eos_token: "</s>"
-  unk_token: "<unk>"
-tokens: # these are delimiters
-  - "<|im_start|>"
-  - "<|im_end|>"
-```
-
-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.
-The config file is the same config file used for training.
-
-Pass the appropriate flag to the inference command, depending upon what kind of model was trained:
-
-- Pretrained LORA:
-  ```bash
-  python -m axolotl.cli.inference examples/your_config.yml --lora_model_dir="./lora-output-dir"
-  ```
-- Full weights finetune:
-  ```bash
-  python -m axolotl.cli.inference examples/your_config.yml --base_model="./completed-model"
-  ```
-- Full weights finetune w/ a prompt from a text file:
-  ```bash
-  cat /tmp/prompt.txt | python -m axolotl.cli.inference examples/your_config.yml \
-    --base_model="./completed-model" --prompter=None --load_in_8bit=True
-  ```
--- With gradio hosting
-  ```bash
-  python -m axolotl.cli.inference examples/your_config.yml --gradio
-  ```
-
-Please use `--sample_packing False` if you have it on and receive the error similar to below:
-
-> RuntimeError: stack expects each tensor to be equal size, but got [1, 32, 1, 128] at entry 0 and [1, 32, 8, 128] at entry 1
-
-### Merge LORA to base
-
-The following command will merge your LORA adapater with your base model. You can optionally pass the argument `--lora_model_dir` to specify the directory where your LORA adapter was saved, otherwhise, this will be inferred from `output_dir` in your axolotl config file.  The merged model is saved in the sub-directory `{lora_model_dir}/merged`.
-
-```bash
-python3 -m axolotl.cli.merge_lora your_config.yml --lora_model_dir="./completed-model"
-```
-
-You may need to use the `gpu_memory_limit` and/or `lora_on_cpu` config options to avoid running out of memory. If you still run out of CUDA memory, you can try to merge in system RAM with
-
-```bash
-CUDA_VISIBLE_DEVICES="" python3 -m axolotl.cli.merge_lora ...
-```
-
-although this will be very slow, and using the config options above are recommended instead.
-
-## Common Errors 🧰
-
-See also the [FAQ's](./docs/faq.qmd) and [debugging guide](docs/debugging.qmd).
-
-> If you encounter a 'Cuda out of memory' error, it means your GPU ran out of memory during the training process. Here's how to resolve it:
-
-Please reduce any below
-  - `micro_batch_size`
-  - `eval_batch_size`
-  - `gradient_accumulation_steps`
-  - `sequence_len`
-
-If it does not help, try running without deepspeed and without accelerate (replace "accelerate launch" with "python") in the command.
-
-Using adamw_bnb_8bit might also save you some memory.
-
-> `failed (exitcode: -9)`
-
-Usually means your system has run out of system memory.
-Similarly, you should consider reducing the same settings as when you run out of VRAM.
-Additionally, look into upgrading your system RAM which should be simpler than GPU upgrades.
-
-> RuntimeError: expected scalar type Float but found Half
-
-Try set `fp16: true`
-
-> NotImplementedError: No operator found for `memory_efficient_attention_forward` ...
-
-Try to turn off xformers.
-
-> accelerate config missing
-
-It's safe to ignore it.
-
-> NCCL Timeouts during training
-
-See the [NCCL](docs/nccl.qmd) guide.
-
-
-### Tokenization Mismatch b/w Inference & Training
-
-For many formats, Axolotl constructs prompts by concatenating token ids _after_ tokenizing strings.  The reason for concatenating token ids rather than operating on strings is to maintain precise accounting for attention masks.
-
-If you decode a prompt constructed by axolotl, you might see spaces between tokens (or lack thereof) that you do not expect, especially around delimiters and special tokens.  When you are starting out with a new format, you should always do the following:
-
-1. Materialize some data using `python -m axolotl.cli.preprocess your_config.yml --debug`, and then decode the first few rows with your model's tokenizer.
-2. During inference, right before you pass a tensor of token ids to your model, decode these tokens back into a string.
-3. Make sure the inference string from #2 looks **exactly** like the data you fine tuned on from #1, including spaces and new lines.  If they aren't the same, adjust your inference server accordingly.
-4. As an additional troubleshooting step, you can look at the token ids between 1 and 2 to make sure they are identical.
-
-Having misalignment between your prompts during training and inference can cause models to perform very poorly, so it is worth checking this.  See [this blog post](https://hamel.dev/notes/llm/finetuning/05_tokenizer_gotchas.html) for a concrete example.
-
-## Debugging Axolotl
-
-See [this debugging guide](docs/debugging.qmd) for tips on debugging Axolotl, along with an example configuration for debugging with VSCode.
-
-## Need help? 🙋
-
-Join our [Discord server](https://discord.gg/HhrNrHJPRb) where our community members can help you.
-
-Need dedicated support? Please contact us at [✉️wing@axolotl.ai](ailto:wing@axolotl.ai) for dedicated support options.
+This project is licensed under the Apache 2.0 License - see the [LICENSE](LICENSE) file for details.

_quarto.yml

@@ -3,10 +3,12 @@ project:
 
 website:
   title: "Axolotl"
-  description: "Fine-tuning"
+  description: "We make fine-tuning accessible, scalable, and fun"
   favicon: favicon.jpg
+
   navbar:
-    title: Axolotl
+    logo: image/axolotl_logo_digital_white.svg
+    title: false
     background: dark
     pinned: false
     collapse: false
@@ -25,29 +27,58 @@ website:
       contents:
         - text: Home
           href: index.qmd
-        - section: "How-To Guides"
+
+        - section: "Getting Started"
           contents:
-          # TODO Edit folder structure after we have more docs.
-            - docs/debugging.qmd
-            - docs/multipack.qmd
-            - docs/fsdp_qlora.qmd
-            - docs/input_output.qmd
-            - docs/rlhf.qmd
-            - docs/nccl.qmd
-            - docs/mac.qmd
-            - docs/multi-node.qmd
-            - docs/unsloth.qmd
-            - docs/amd_hpc.qmd
+            - docs/getting-started.qmd
+            - docs/installation.qmd
+            - docs/cli.qmd
+            - docs/inference.qmd
+
         - section: "Dataset Formats"
           contents: docs/dataset-formats/*
+
+        - section: "Deployments"
+          contents:
+            - docs/multi-gpu.qmd
+            - docs/multi-node.qmd
+            - docs/ray-integration.qmd
+            - docs/amd_hpc.qmd
+            - docs/mac.qmd
+
+        - section: "How To Guides"
+          contents:
+            - docs/multimodal.qmd
+            - docs/rlhf.qmd
+            - docs/reward_modelling.qmd
+            - docs/lr_groups.qmd
+            - docs/lora_optims.qmd
+
+        - section: "Core Concepts"
+          contents:
+            - docs/batch_vs_grad.qmd
+            - docs/dataset_preprocessing.qmd
+            - docs/multipack.qmd
+
+        - section: "Advanced Features"
+          contents:
+            - docs/fsdp_qlora.qmd
+            - docs/unsloth.qmd
+            - docs/torchao.qmd
+            - docs/custom_integrations.qmd
+
+        - section: "Troubleshooting"
+          contents:
+            - docs/faq.qmd
+            - docs/debugging.qmd
+            - docs/nccl.qmd
+
         - section: "Reference"
           contents:
             - docs/config.qmd
-        - docs/faq.qmd
-
 
 format:
   html:
-    theme: materia
+    theme: darkly
     css: styles.css
     toc: true

cicd/Dockerfile.jinja

@@ -8,6 +8,7 @@ ENV PYTORCH_VERSION="{{ PYTORCH_VERSION }}"
 ENV GITHUB_REF="{{ GITHUB_REF }}"
 ENV GITHUB_SHA="{{ GITHUB_SHA }}"
 ENV NIGHTLY_BUILD="{{ NIGHTLY_BUILD }}"
+ENV HF_HOME="{{ HF_HOME }}"
 
 RUN apt-get update && \
     apt-get install -y --allow-change-held-packages vim curl nano libnccl2 libnccl-dev
@@ -31,9 +32,9 @@ RUN if [ "$NIGHTLY_BUILD" = "true" ] ; then \
     fi
 
 RUN if [ "$AXOLOTL_EXTRAS" != "" ] ; then \
-        pip install -e .[deepspeed,flash-attn,optimizers,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
+        pip install --no-build-isolation -e .[deepspeed,flash-attn,optimizers,ray,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
     else \
-        pip install -e .[deepspeed,flash-attn,optimizers] $AXOLOTL_ARGS; \
+        pip install --no-build-isolation -e .[deepspeed,flash-attn,optimizers,ray] $AXOLOTL_ARGS; \
     fi
 
 RUN python scripts/unsloth_install.py | sh

cicd/cicd.sh

@@ -1,7 +1,11 @@
 #!/bin/bash
 set -e
 
+python -c "import torch; assert '$PYTORCH_VERSION' in torch.__version__"
+
 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/
+pytest -v --durations=10 /workspace/axolotl/tests/e2e/patched/lora_kernels  # running these with the other patches causes a failure
+pytest -v --durations=10 --ignore=tests/e2e/patched/lora_kernels /workspace/axolotl/tests/e2e/patched
+pytest -v --durations=10 -n1 /workspace/axolotl/tests/e2e/solo/
+pytest -v --durations=10 /workspace/axolotl/tests/e2e/integrations/
+pytest -v --durations=10 --ignore=tests/e2e/solo/ --ignore=tests/e2e/patched/ --ignore=tests/e2e/multigpu/ --ignore=tests/e2e/integrations/ /workspace/axolotl/tests/e2e/

cicd/multigpu.py

@@ -23,11 +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", "2.3.1"),
-    "BASE_TAG": os.environ.get("BASE_TAG", "main-base-py3.11-cu121-2.3.1"),
+    "PYTORCH_VERSION": os.environ.get("PYTORCH_VERSION", "2.4.1"),
+    "BASE_TAG": os.environ.get("BASE_TAG", "main-base-py3.11-cu121-2.4.1"),
     "CUDA": os.environ.get("CUDA", "121"),
     "GITHUB_REF": os.environ.get("GITHUB_REF", "refs/heads/main"),
     "GITHUB_SHA": os.environ.get("GITHUB_SHA", ""),
+    "HF_HOME": "/workspace/data/huggingface-cache/hub",
 }
 
 dockerfile_contents = df_template.render(**df_args)
@@ -36,18 +37,20 @@ 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")
-)
+cicd_image = Image.from_dockerfile(
+    pathlib.Path(temp_dir) / "Dockerfile",
+    force_build=True,
+    gpu="A10G",
+).env(df_args)
 
 app = App("Axolotl CI/CD", secrets=[])
 
+hf_cache_volume = modal.Volume.from_name(
+    "axolotl-ci-hf-hub-cache", create_if_missing=True
+)
+VOLUME_CONFIG = {
+    "/workspace/data/huggingface-cache/hub": hf_cache_volume,
+}
 
 N_GPUS = int(os.environ.get("N_GPUS", 2))
 GPU_CONFIG = modal.gpu.H100(count=N_GPUS)
@@ -67,6 +70,7 @@ def run_cmd(cmd: str, run_folder: str):
     timeout=60 * 60,
     cpu=8.0,
     memory=131072 * N_GPUS,
+    volumes=VOLUME_CONFIG,
 )
 def cicd_pytest():
     run_cmd("./cicd/multigpu.sh", "/workspace/axolotl")

cicd/tests.py

@@ -1,6 +1,4 @@
-"""
- modal application to run axolotl gpu tests in Modal
- """
+"""Modal app to run axolotl GPU tests"""
 # pylint: disable=duplicate-code
 
 import os
@@ -23,12 +21,13 @@ 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"),
+    "PYTORCH_VERSION": os.environ.get("PYTORCH_VERSION", "2.4.1"),
+    "BASE_TAG": os.environ.get("BASE_TAG", "main-base-py3.11-cu121-2.4.1"),
     "CUDA": os.environ.get("CUDA", "121"),
     "GITHUB_REF": os.environ.get("GITHUB_REF", "refs/heads/main"),
     "GITHUB_SHA": os.environ.get("GITHUB_SHA", ""),
     "NIGHTLY_BUILD": os.environ.get("NIGHTLY_BUILD", ""),
+    "HF_HOME": "/workspace/data/huggingface-cache/hub",
 }
 
 dockerfile_contents = df_template.render(**df_args)
@@ -37,22 +36,24 @@ temp_dir = tempfile.mkdtemp()
 with open(pathlib.Path(temp_dir) / "Dockerfile", "w", encoding="utf-8") as f:
     f.write(dockerfile_contents)
 
-cicd_image = (
-    Image.from_dockerfile(
-        pathlib.Path(temp_dir) / "Dockerfile",
-        context_mount=None,
-        force_build=True,
-        gpu="A10G",
-    )
-    .env(df_args)
-    .pip_install("fastapi==0.110.0", "pydantic==2.6.3")
-)
+cicd_image = Image.from_dockerfile(
+    pathlib.Path(temp_dir) / "Dockerfile",
+    context_mount=None,
+    force_build=True,
+    gpu="A10G",
+).env(df_args)
 
 app = App("Axolotl CI/CD", secrets=[])
 
+hf_cache_volume = modal.Volume.from_name(
+    "axolotl-ci-hf-hub-cache", create_if_missing=True
+)
+VOLUME_CONFIG = {
+    "/workspace/data/huggingface-cache/hub": hf_cache_volume,
+}
 
 N_GPUS = int(os.environ.get("N_GPUS", 1))
-GPU_CONFIG = modal.gpu.A10G(count=N_GPUS)
+GPU_CONFIG = modal.gpu.L40S(count=N_GPUS)
 
 
 def run_cmd(cmd: str, run_folder: str):
@@ -69,6 +70,7 @@ def run_cmd(cmd: str, run_folder: str):
     timeout=60 * 60,
     cpu=8.0,
     memory=131072,
+    volumes=VOLUME_CONFIG,
 )
 def cicd_pytest():
     run_cmd("./cicd/cicd.sh", "/workspace/axolotl")

deepspeed_configs/zero1_torch_compile.json

@@ -0,0 +1,27 @@
+{
+  "zero_optimization": {
+    "stage": 1,
+    "overlap_comm": true
+  },
+  "bf16": {
+    "enabled": "auto"
+  },
+  "fp16": {
+    "enabled": "auto",
+    "auto_cast": false,
+    "loss_scale": 0,
+    "initial_scale_power": 32,
+    "loss_scale_window": 1000,
+    "hysteresis": 2,
+    "min_loss_scale": 1
+  },
+  "compile": {
+    "disable": false,
+    "backend": "inductor"
+  },
+  "gradient_accumulation_steps": "auto",
+  "gradient_clipping": "auto",
+  "train_batch_size": "auto",
+  "train_micro_batch_size_per_gpu": "auto",
+  "wall_clock_breakdown": false
+}

docker/Dockerfile

@@ -20,9 +20,9 @@ WORKDIR /workspace/axolotl
 
 # If AXOLOTL_EXTRAS is set, append it in brackets
 RUN if [ "$AXOLOTL_EXTRAS" != "" ] ; then \
-        pip install -e .[deepspeed,flash-attn,optimizers,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
+        pip install --no-build-isolation -e .[deepspeed,flash-attn,optimizers,ray,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
     else \
-        pip install -e .[deepspeed,flash-attn,optimizers] $AXOLOTL_ARGS; \
+        pip install --no-build-isolation -e .[deepspeed,flash-attn,optimizers,ray] $AXOLOTL_ARGS; \
     fi
 
 RUN python scripts/unsloth_install.py | sh

docker/Dockerfile-cloud

@@ -20,7 +20,8 @@ RUN apt install --yes --no-install-recommends openssh-server tmux && \
     printf "\n[[ -z \"\$TMUX\"  ]] && { tmux attach-session -t ssh_tmux || tmux new-session -s ssh_tmux; exit; }\n" >> ~/.bashrc && \
     printf "[ ! -z \"\$TERM\" -a -r /etc/motd ] && cat /etc/motd\n" >> ~/.bashrc && \
     chmod +x /workspace/axolotl/scripts/cloud-entrypoint.sh && \
-    chmod +x /root/cloud-entrypoint.sh
+    chmod +x /root/cloud-entrypoint.sh && \
+    echo 'set-option -g history-limit 5000' >> ~/.tmux.conf
 
 ENTRYPOINT ["/root/cloud-entrypoint.sh"]
 CMD ["sleep", "infinity"]

docker/Dockerfile-tests

@@ -24,9 +24,9 @@ RUN git fetch origin +$GITHUB_REF && \
 
 # If AXOLOTL_EXTRAS is set, append it in brackets
 RUN if [ "$AXOLOTL_EXTRAS" != "" ] ; then \
-        pip install -e .[deepspeed,flash-attn,mamba-ssm,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
+        pip install --no-build-isolation -e .[deepspeed,flash-attn,mamba-ssm,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
     else \
-        pip install -e .[deepspeed,flash-attn,mamba-ssm] $AXOLOTL_ARGS; \
+        pip install --no-build-isolation -e .[deepspeed,flash-attn,mamba-ssm] $AXOLOTL_ARGS; \
     fi
 
 # So we can test the Docker image

docs/amd_hpc.qmd

@@ -1,5 +1,5 @@
 ---
-title: Training with AMD GPUs on HPC Systems
+title: AMD GPUs on HPC Systems
 description: A comprehensive guide for using Axolotl on distributed systems with AMD GPUs
 ---
 
@@ -52,7 +52,7 @@ 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 .
+pip install --no-build-isolation .
 ```
 
 ### 6. Install Axolotl
@@ -63,7 +63,7 @@ Clone and install Axolotl:
 git clone https://github.com/axolotl-ai-cloud/axolotl
 cd axolotl
 pip install packaging ninja
-pip install -e .
+pip install --no-build-isolation -e .
 ```
 
 ### 7. Apply xformers Workaround

docs/cli.qmd

@@ -0,0 +1,288 @@
+---
+title: "CLI Reference"
+format:
+  html:
+    toc: true
+    toc-expand: 2
+    toc-depth: 3
+execute:
+  enabled: false
+---
+
+The Axolotl CLI provides a streamlined interface for training and fine-tuning large language models. This guide covers
+the CLI commands, their usage, and common examples.
+
+
+## Basic Commands
+
+All Axolotl commands follow this general structure:
+
+```bash
+axolotl <command> [config.yml] [options]
+```
+
+The config file can be local or a URL to a raw YAML file.
+
+## Command Reference
+
+### fetch
+
+Downloads example configurations and deepspeed configs to your local machine.
+
+```bash
+# Get example YAML files
+axolotl fetch examples
+
+# Get deepspeed config files
+axolotl fetch deepspeed_configs
+
+# Specify custom destination
+axolotl fetch examples --dest path/to/folder
+```
+
+### preprocess
+
+Preprocesses and tokenizes your dataset before training. This is recommended for large datasets.
+
+```bash
+# Basic preprocessing
+axolotl preprocess config.yml
+
+# Preprocessing with one GPU
+CUDA_VISIBLE_DEVICES="0" axolotl preprocess config.yml
+
+# Debug mode to see processed examples
+axolotl preprocess config.yml --debug
+
+# Debug with limited examples
+axolotl preprocess config.yml --debug --debug-num-examples 5
+```
+
+Configuration options:
+
+```yaml
+dataset_prepared_path: Local folder for saving preprocessed data
+push_dataset_to_hub: HuggingFace repo to push preprocessed data (optional)
+```
+
+### train
+
+Trains or fine-tunes a model using the configuration specified in your YAML file.
+
+```bash
+# Basic training
+axolotl train config.yml
+
+# Train and set/override specific options
+axolotl train config.yml \
+    --learning-rate 1e-4 \
+    --micro-batch-size 2 \
+    --num-epochs 3
+
+# Training without accelerate
+axolotl train config.yml --no-accelerate
+
+# Resume training from checkpoint
+axolotl train config.yml --resume-from-checkpoint path/to/checkpoint
+```
+
+It is possible to run sweeps over multiple hyperparameters by passing in a sweeps config.
+
+```bash
+# Basic training with sweeps
+axolotl train config.yml --sweep path/to/sweep.yaml
+```
+
+Example sweep config:
+```yaml
+_:
+  # This section is for dependent variables we need to fix
+  - load_in_8bit: false
+    load_in_4bit: false
+    adapter: lora
+  - load_in_8bit: true
+    load_in_4bit: false
+    adapter: lora
+
+# These are independent variables
+learning_rate: [0.0003, 0.0006]
+lora_r:
+  - 16
+  - 32
+lora_alpha:
+  - 16
+  - 32
+  - 64
+```
+
+
+
+### inference
+
+Runs inference using your trained model in either CLI or Gradio interface mode.
+
+```bash
+# CLI inference with LoRA
+axolotl inference config.yml --lora-model-dir="./outputs/lora-out"
+
+# CLI inference with full model
+axolotl inference config.yml --base-model="./completed-model"
+
+# Gradio web interface
+axolotl inference config.yml --gradio \
+    --lora-model-dir="./outputs/lora-out"
+
+# Inference with input from file
+cat prompt.txt | axolotl inference config.yml \
+    --base-model="./completed-model"
+```
+
+### merge-lora
+
+Merges trained LoRA adapters into the base model.
+
+```bash
+# Basic merge
+axolotl merge-lora config.yml
+
+# Specify LoRA directory (usually used with checkpoints)
+axolotl merge-lora config.yml --lora-model-dir="./lora-output/checkpoint-100"
+
+# Merge using CPU (if out of GPU memory)
+CUDA_VISIBLE_DEVICES="" axolotl merge-lora config.yml
+```
+
+Configuration options:
+
+```yaml
+gpu_memory_limit: Limit GPU memory usage
+lora_on_cpu: Load LoRA weights on CPU
+```
+
+### merge-sharded-fsdp-weights
+
+Merges sharded FSDP model checkpoints into a single combined checkpoint.
+
+```bash
+# Basic merge
+axolotl merge-sharded-fsdp-weights config.yml
+```
+
+### evaluate
+
+Evaluates a model's performance using metrics specified in the config.
+
+```bash
+# Basic evaluation
+axolotl evaluate config.yml
+```
+
+### lm-eval
+
+Runs LM Evaluation Harness on your model.
+
+```bash
+# Basic evaluation
+axolotl lm-eval config.yml
+```
+
+Configuration options:
+
+```yaml
+# List of tasks to evaluate
+lm_eval_tasks:
+  - arc_challenge
+  - hellaswag
+lm_eval_batch_size: # Batch size for evaluation
+output_dir: # Directory to save evaluation results
+```
+
+## Legacy CLI Usage
+
+While the new Click-based CLI is preferred, Axolotl still supports the legacy module-based CLI:
+
+```bash
+# Preprocess
+python -m axolotl.cli.preprocess config.yml
+
+# Train
+accelerate launch -m axolotl.cli.train config.yml
+
+# Inference
+accelerate launch -m axolotl.cli.inference config.yml \
+    --lora_model_dir="./outputs/lora-out"
+
+# Gradio interface
+accelerate launch -m axolotl.cli.inference config.yml \
+    --lora_model_dir="./outputs/lora-out" --gradio
+```
+
+::: {.callout-important}
+When overriding CLI parameters in the legacy CLI, use same notation as in yaml file (e.g., `--lora_model_dir`).
+
+**Note:** This differs from the new Click-based CLI, which uses dash notation (e.g., `--lora-model-dir`). Keep this in mind if you're referencing newer documentation or switching between CLI versions.
+:::
+
+## Remote Compute with Modal Cloud
+
+Axolotl supports running training and inference workloads on Modal cloud infrastructure. This is configured using a
+cloud YAML file alongside your regular Axolotl config.
+
+### Cloud Configuration
+
+Create a cloud config YAML with your Modal settings:
+
+```yaml
+# cloud_config.yml
+provider: modal
+gpu: a100  # Supported: l40s, a100-40gb, a100-80gb, a10g, h100, t4, l4
+gpu_count: 1    # Number of GPUs to use
+timeout: 86400  # Maximum runtime in seconds (24 hours)
+branch: main    # Git branch to use (optional)
+
+volumes:        # Persistent storage volumes
+  - name: axolotl-cache
+    mount: /workspace/cache
+  - name: axolotl-data
+    mount: /workspace/data
+  - name: axolotl-artifacts
+    mount: /workspace/artifacts
+
+env:            # Environment variables
+  - WANDB_API_KEY
+  - HF_TOKEN
+```
+
+### Running on Modal Cloud
+
+Commands that support the --cloud flag:
+
+```bash
+# Preprocess on cloud
+axolotl preprocess config.yml --cloud cloud_config.yml
+
+# Train on cloud
+axolotl train config.yml --cloud cloud_config.yml
+
+# Train without accelerate on cloud
+axolotl train config.yml --cloud cloud_config.yml --no-accelerate
+
+# Run lm-eval on cloud
+axolotl lm-eval config.yml --cloud cloud_config.yml
+```
+
+### Cloud Configuration Options
+
+```yaml
+provider: # compute provider, currently only `modal` is supported
+gpu: # GPU type to use
+gpu_count: # Number of GPUs (default: 1)
+memory: # RAM in GB (default: 128)
+timeout: # Maximum runtime in seconds
+timeout_preprocess: # Preprocessing timeout
+branch: # Git branch to use
+docker_tag: # Custom Docker image tag
+volumes: # List of persistent storage volumes
+env: # Environment variables to pass
+secrets: # Secrets to inject
+```

docs/config.qmd

@@ -46,6 +46,10 @@ overrides_of_model_config:
     type: # linear | dynamic
     factor: # float
 
+# optional overrides the base model loading from_pretrained
+overrides_of_model_kwargs:
+  # use_cache: False
+
 # optional overrides to the bnb 4bit quantization configuration
 # https://huggingface.co/docs/transformers/main/main_classes/quantization#transformers.BitsAndBytesConfig
 bnb_config_kwargs:
@@ -74,6 +78,9 @@ tf32: true # require >=ampere
 bfloat16: true # require >=ampere
 float16: true
 
+# Use Tensor parallel
+tensor_parallel: true # require multi-gGPU
+
 # Limit the memory for all available GPUs to this amount (if an integer, expressed in gigabytes); default: unset
 gpu_memory_limit: 20GiB
 # Do the LoRA/PEFT loading on CPU -- this is required if the base model is so large it takes up most or all of the available GPU VRAM, e.g. during a model and LoRA merge
@@ -87,7 +94,12 @@ datasets:
     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
+
+    shards: # Optional[int] split dataset into N pieces (use with shards_idx)
+    shards_idx: # Optional[int] = 0 the index of sharded dataset to use
+
+    preprocess_shards: # Optional[int] process dataset in N sequential chunks for memory efficiency (exclusive with `shards`)
+
     name: # Optional[str] name of dataset configuration to load
     train_on_split: train # Optional[str] name of dataset split to load from
     revision: # Optional[str] The specific revision of the dataset to use when loading from the Hugging Face Hub. This can be a commit hash, tag, or branch name. If not specified, the latest version will be used. This parameter is ignored for local datasets.
@@ -127,34 +139,50 @@ datasets:
     # - 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`).
+
+    # Custom jinja chat template. Used only if `chat_template: jinja` or empty.
     chat_template_jinja:
-    # The key in the data example that contains the messages. Default is "messages".
+
+    # Key containing the messages (default: "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.
+
+    # Mapping of properties from the input dataset to the chat template.
+    # (default: message_property_mappings={'role':'role', 'content':'content'})
+    # If a property exists in the template but not in this mapping, the system will attempt
+    # to load it directly from the message using the property name as the key.
+    # Example: In the mapping below, 'from' is loaded from input dataset and used as 'role',
+    # while 'value' is loaded and used as 'content' in the chat template.
+    message_property_mappings:
+      role: from
+      content: value
+      # ...
+
+    message_property_mappings:
+
+    # Optional[Dict[str, List]]. Roles mapping in the messages. The default is:
     roles:
       user: ["human", "user"]
-      assistant: ["gpt", "assistant", "ai"]
+      assistant: ["gpt", "assistant"]
       system: ["system"]
+      tool: ["tool"]
 
-    ## NOTE: Leaving the below empty will default to using the simple legacy tokenization strategy where only last message is trained on.
+    # IMPORTANT: The following fields determine which parts of the conversation to train on.
+    # Priority order: message_field_training > message_field_training_detail > train_on_inputs or role in roles_to_train
+    # See examples at `docs/dataset-formats/conversation.qmd`
+    # Note: If the below 4 fields are set to empty, defaults to training only on the last message.
 
     # Optional[List[str]]. Roles to train on. The tokens from these roles will be considered for the loss.
-    roles_to_train: ["gpt", "assistant"]
+    roles_to_train: ["assistant"]  # default
     # Optional[str]. Which EOS tokens to train on in the conversation. Possible values are:
     # - all: train on all EOS tokens
-    # - turn: train on the EOS token at the end of each trainable turn
+    # - turn (default): train on the EOS token at the end of each trainable turn
     # - last: train on the last EOS token in the conversation
+    # TIP: Please make sure that your `tokenizer.eos_token` is same as EOS/EOT token in template. Otherwise, set `eos_token` under `special_tokens`.
     train_on_eos: 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
 
 
@@ -181,6 +209,12 @@ rl:
 # whether to perform weighting if doing DPO training. Boolean.
 dpo_use_weighting:
 
+# reward modelling: `True` or `False`
+reward_model:
+
+# process reward modelling: `True` or `False`
+process_reward_model:
+
 # The name of the chat template to use for training, following values are supported:
 # - tokenizer_default: Uses the chat template that is available in the tokenizer_config.json. If the chat template is not available in the tokenizer, it will raise an error. This is the default value.
 # - alpaca/inst/chatml/gemma/cohere/llama3/phi_3/deepseek_v2/jamba: These chat templates are available in the axolotl codebase at src/axolotl/utils/chat_templates.py
@@ -238,6 +272,11 @@ total_num_tokens:
 sample_packing_group_size: 100000
 # The number of samples which can be packed into one sequence. Increase if using a large sequence_len with many short samples.
 sample_packing_bin_size: 200
+# whether to concatenate samples during pretraining
+pretraining_sample_concatenation:
+
+# Use batch flattening for speedups when not using sample_packing
+batch_flattening:
 
 # Passed through to transformers when loading the model when launched without accelerate
 # Use `sequential` when training w/ model parallelism to limit memory
@@ -279,6 +318,13 @@ lora_modules_to_save:
 
 lora_fan_in_fan_out: false
 
+# Apply custom LoRA autograd functions and activation function Triton kernels for
+# speed and memory savings
+# See: https://axolotl-ai-cloud.github.io/axolotl/docs/lora_optims.html
+lora_mlp_kernel: true
+lora_qkv_kernel: true
+lora_o_kernel: true
+
 # LoRA+ hyperparameters
 # For more details about the following options, see:
 # https://arxiv.org/abs/2402.12354  and `src/axolotl/core/train_builder.py`
@@ -327,11 +373,15 @@ comet_mode: # Create a new experiment ("create") or log to an existing one ("get
 comet_online: # Set to True to log data to Comet server, or False for offline storage. Default is True.
 comet_experiment_config: # Dictionary for additional configuration settings, see the doc for more details.
 
+# Tensorboard
+use_tensorboard: # Optional[bool]
+
 # Where to save the full-finetuned model to
 output_dir: ./completed-model
 
 # Whether to use torch.compile and which backend to use
-torch_compile:  # bool
+# setting to `auto` will enable torch compile when torch>=2.5.1
+torch_compile:  # Optional[Union[Literal["auto"], bool]]
 torch_compile_backend:  # Optional[str]
 
 # Training hyperparameters
@@ -348,10 +398,11 @@ warmup_ratio: 0.05  # cannot use with warmup_steps
 learning_rate: 0.00003
 lr_quadratic_warmup:
 logging_steps:
-eval_steps: # Leave empty to eval at each epoch, integers for every N steps. decimal for fraction of total steps
+eval_steps: # Leave empty to eval at each epoch, integer for every N steps. float for fraction of total steps
 evals_per_epoch: # number of times per epoch to run evals, mutually exclusive with eval_steps
-save_strategy: # Set to `"no"` to skip checkpoint saves
-save_steps: # Leave empty to save at each epoch
+eval_strategy: # Set to `"no"` to skip evaluation, `"epoch"` at end of each epoch, leave empty to infer from `eval_steps`.
+save_strategy: # Set to `"no"` to skip checkpoint saves, `"epoch"` at end of each epoch, `"best"` when better result is achieved, leave empty to infer from `save_steps`.
+save_steps: # Leave empty to save at each epoch, integer for every N steps. float for fraction of total steps
 saves_per_epoch: # number of times per epoch to save a checkpoint, mutually exclusive with save_steps
 save_total_limit: # Checkpoints saved at a time
 # Maximum number of iterations to train for. It precedes num_epochs which means that
@@ -359,10 +410,20 @@ save_total_limit: # Checkpoints saved at a time
 # e.g., when 1 epoch is 1000 steps => `num_epochs: 2` and `max_steps: 100` will train for 100 steps
 max_steps:
 
+# bool of whether to include tokens trainer per second in the training metrics. This iterates over the entire dataset once, so it takes some time.
+include_tokens_per_second: # Optional[bool]
+
+# whether to find batch size that fits in memory. Passed to underlying transformers Trainer
+auto_find_batch_size: # Optional[bool]
+
 eval_table_size: # Approximate number of predictions sent to wandb depending on batch size. Enabled above 0. Default is 0
 eval_max_new_tokens: # Total number of tokens generated for predictions sent to wandb. Default is 128
 eval_causal_lm_metrics: # HF evaluate metrics used during evaluation. Default is ["sacrebleu", "comet", "ter", "chrf", "perplexity"]
 
+profiler_steps: # enable the pytorch profiler to capture the first N steps of training to the output_dir.
+                # see https://pytorch.org/blog/understanding-gpu-memory-1/ for more information
+                # snapshots can be visualized @ https://pytorch.org/memory_viz
+
 loss_watchdog_threshold: # High loss value, indicating the learning has broken down (a good estimate is ~2 times the loss at the start of training)
 loss_watchdog_patience: # Number of high-loss steps in a row before the trainer aborts (default: 3)
 

docs/custom_integrations.qmd

@@ -0,0 +1,57 @@
+---
+title: Custom Integrations
+toc: true
+toc-depth: 3
+---
+
+```{python}
+#| echo: false
+
+import re
+
+def process_readme(integration_name):
+    try:
+        path = f'../src/axolotl/integrations/{integration_name}/README.md'
+        with open(path, 'r') as f:
+            txt = f.read()
+            # Remove h1 headings
+            txt = re.sub(r'^# .*\n?', '', txt, flags=re.MULTILINE)
+            # Convert h2 to h3
+            txt = re.sub(r'^## ', '### ', txt, flags=re.MULTILINE)
+            return txt
+    except FileNotFoundError:
+        return None
+
+def print_section(name, folder_name):
+    output = f"\n## {name}\n"
+    content = process_readme(folder_name)
+    if content:
+        output += content
+    output += f"\nPlease see reference [here](https://github.com/axolotl-ai-cloud/axolotl/tree/main/src/axolotl/integrations/{folder_name})\n"
+    return output
+```
+
+```{python}
+#| output: asis
+#| echo: false
+
+# Introduction text
+print("""
+Axolotl adds custom features through `integrations`. They are located within the `src/axolotl/integrations` directory.
+
+To enable them, please check the respective documentations.
+""")
+
+# Sections
+sections = [
+    ("Cut Cross Entropy", "cut_cross_entropy"),
+    ("Grokfast", "grokfast"),
+    ("Knowledge Distillation (KD)", "kd"),
+    ("Liger Kernels", "liger"),
+    ("Language Model Evaluation Harness (LM Eval)", "lm_eval"),
+    ("Spectrum", "spectrum")
+]
+
+for section_name, folder_name in sections:
+    print(print_section(section_name, folder_name))
+```

docs/dataset-formats/conversation.qmd

@@ -6,8 +6,9 @@ order: 3
 
 ## sharegpt
 
-IMPORTANT: ShareGPT is deprecated!. Please see `chat_template` section below.
-
+::: {.callout-important}
+ShareGPT is deprecated!. Please see [chat_template](#chat_template) section below.
+:::
 
 ## pygmalion
 
@@ -15,7 +16,6 @@ IMPORTANT: ShareGPT is deprecated!. Please see `chat_template` section below.
 {"conversations": [{"role": "...", "value": "..."}]}
 ```
 
-
 ## 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.
@@ -24,7 +24,7 @@ Chat Template strategy uses a jinja2 template that converts a list of messages i
 {"conversations": [{"role": "...", "content": "..."}]}
 ```
 
-See `config.qmd` for full configs and supported templates.
+See [configs](../config.qmd) for full configs and supported templates.
 
 ### Migrating from sharegpt
 
@@ -44,8 +44,9 @@ datasets:
     type: chat_template
 
     field_messages: conversations
-    message_field_role: from
-    message_field_content: value
+    message_property_mappings:
+      role: from
+      content: value
 
 # new (if setting a new chat_template like chatml, gemma, etc)
 chat_template: chatml
@@ -54,8 +55,9 @@ datasets:
     type: chat_template
 
     field_messages: conversations
-    message_field_role: from
-    message_field_content: value
+    message_property_mappings:
+      role: from
+      content: value
 ```
 
 We recommend checking the below examples for other usecases.
@@ -68,6 +70,8 @@ We recommend checking the below examples for other usecases.
 datasets:
   - path: ...
     type: chat_template
+    roles_to_train:
+    train_on_eos:
 ```
 
 2. Using the `gemma` chat template to override the tokenizer_config.json's chat template on OpenAI messages format, training on all assistant messages.
@@ -77,7 +81,7 @@ chat_template: gemma # this overwrites the tokenizer's chat_template
 datasets:
   - path: ...
     type: chat_template
-    roles_to_train: ["assistant"]
+    roles_to_train: ["assistant"]  # default value
 ```
 
 3. Using the tokenizer_config.json's chat template or `chatml` as fallback if the former's chat template does not exist, on OpenAI messages format, training on all assistant messages.
@@ -87,7 +91,6 @@ chat_template: tokenizer_default_fallback_chatml # this overwrites the tokenizer
 datasets:
   - path: ...
     type: chat_template
-    roles_to_train: ["assistant"]
 ```
 
 4. Using a custom jinja template on OpenAI messages format, training on all assistant messages.
@@ -99,9 +102,12 @@ chat_template_jinja: "{{ bos_token }}{% for message in messages %}{% if (message
 datasets:
   - path: ...
     type: chat_template
-    roles_to_train: ["assistant"]
 ```
 
+::: {.callout-important}
+Please make sure that your `tokenizer.eos_token` is same as EOS/EOT token in template. Otherwise, set `eos_token` under `special_tokens`.
+:::
+
 5. (Advanced) Using fine-grained control over tokens and turns to train in a conversation
 
 For a data sample that looks like:
@@ -140,12 +146,15 @@ datasets:
     type: chat_template
     chat_template: tokenizer_default
     field_messages: conversations
-    message_field_role: from
-    message_field_content: value
+    message_property_mappings:
+      role: from
+      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.
+::: {.callout-tip}
+It is not necessary to set both `message_field_training` and `message_field_training_detail` at once.
+:::

docs/dataset-formats/index.qmd

@@ -1,14 +1,491 @@
 ---
 title: Dataset Formats
-description: Supported dataset formats.
-listing:
-  fields: [title, description]
-  type: table
-  sort-ui: false
-  filter-ui: false
-  max-description-length: 250
+description: Guide to Dataset Formats in Axolotl
+back-to-top-navigation: true
+toc: true
+toc-depth: 5
 ---
 
-Axolotl supports a variety of dataset formats.  It is recommended to use a JSONL format.  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.
 
-Below are these various formats organized by task:
+Axolotl is a training framework that aims to make the process convenient yet flexible to users by simply passing a config yaml file.
+
+As there are a lot of available options in Axolotl, this guide aims to provide an simplify the user experience to choosing the proper choice.
+
+Axolotl supports 3 kinds of training methods: pre-training, supervised fine-tuning, and preference-based post-training (e.g. DPO, ORPO, PRMs). Each method has their own dataset format which are described below.
+
+## Pre-training
+
+When aiming to train on large corpora of text datasets, pre-training is your go-to choice. Due to the size of these datasets, downloading the entire-datasets before beginning training would be prohibitively time-consuming. Axolotl supports [streaming](https://huggingface.co/docs/datasets/en/stream) to only load batches into memory at a time.
+
+A sample format for a pre-training dataset is as follows:
+
+```json
+{"text": "first row"}
+{"text": "second row"}
+...
+```
+
+It is typically recommended to save your dataset as `.jsonl` due to its flexibility and simplicity.
+
+Axolotl supports loading from a Hugging Face hub repo or from local files.
+
+::: {.callout-important}
+For pre-training only, Axolotl would split texts if it exceeds the context length into multiple smaller prompts.
+:::
+
+### Pre-training from Hugging Face hub datasets
+
+As an example, to train using a Hugging Face dataset `hf_org/name`, you can pass the following config:
+
+```yaml
+pretraining_dataset: hf_org/name
+```
+
+### Pre-training from local dataset files
+
+Given a few corpus files: `A.jsonl`, `B.jsonl`, and `C.jsonl`, your config will look like the below:
+
+```yaml
+pretraining_dataset:
+  - path: json
+    data_files:
+      - A.jsonl
+      - B.jsonl
+      - C.jsonl
+```
+
+While we recommend `.jsonl`, you can also use the other formats (`csv`, `parquet`, `arrow`, `SQL`, `Webdataset`) that are supported by [`Dataset.load_dataset`](https://huggingface.co/docs/datasets/loading#local-and-remote-files)
+
+### Pre-training without streaming
+
+On the rare case that the dataset is small and can be loaded entirely into memory, another approach to running pre-training is to use the `completion` format. This would mean that the entire dataset is pre-tokenized instead of on-demand in streaming.
+
+One benefit of this is that the tokenization can be performed separately on a CPU-only machine, and then transferred to a GPU machine for training to save costs.
+
+From Hugging Face:
+
+```yaml
+datasets:
+  - path: hf_org/name
+    type: completion
+```
+
+From local files (either example works):
+
+```yaml
+datasets:
+  - path: A.jsonl
+    type: completion
+
+  - path: json
+    data_files: ["A.jsonl", "B.jsonl", "C.jsonl"]
+    type: completion
+```
+
+### Pre-training dataset configuration tips
+
+#### Setting max_steps
+
+When using streaming for large datasets, Axolotl does not know in advance how large the dataset is and does not know when to stop.
+
+Therefore, it is necessary to set `max_steps: int` in your config for pre-training to run, so that Axolotl knows when to stop training.
+
+One step is equal to `sequence_len * micro_batch_size * gradient_accumulation_steps * total_num_gpus` tokens.
+
+#### Group_by_length
+
+It is recommended to leave this off if downloading from Hugging Face hub as it would download the entire dataset which can be very large.
+
+### Reference
+
+Please see docs [here](pretraining.qmd).
+
+## Supervised fine-tuning (SFT)
+
+Supervised fine-tuning is the process of training models to respond to an instruction or chat input.
+
+As there are a wide variety of dataset formats, Axolotl tries to support a majority of the formats available in public datasets.
+
+Axolotl provides four approaches for loading datasets, however, it's easier to work backwards from the dataset you have available to figure out which approach to use.
+
+A flow chart is as follows:
+
+1. Do you already have the dataset tokenized? If yes, check [Pre-Tokenized Dataset](#pre-tokenized-dataset).
+
+2. Do you want to format the dataset yourself and manually choose each section to mask? If yes, check [Template Free Dataset](#template-free-dataset)
+
+3. Is your dataset in a "conversation" format, containing a `list[messages]`? If yes, check [Conversation Dataset](#conversation-dataset)
+
+4. Is your dataset in an "instruct" format, containing `{ instruction, response }`? If yes, check [Instruction Dataset](#instruction-dataset)
+
+If you went through the flow chart and did not find one that matches, it is recommended to preprocess your dataset into one of the above or create a thread on Github Discussion.
+
+::: {.callout-tip}
+You can mix and match within each approach or across approaches to train a model on a variety of datasets.
+:::
+
+### Pre-Tokenized Dataset
+
+We suggest this approach when you want to bring your own tokenized dataset.
+
+Axolotl expects the dataset to have three keys:
+- `input_ids`: from tokenizing formatted prompt
+- `attention_mask`: for masking padding. If you don't add padding, it would be equal to `len(input_ids) * [1]`
+- `labels`: this is the same as `input_ids`, however, if you want to mask certain tokens, you would set those indices to `-100`.
+
+::: {.callout-tip}
+Make sure to add BOS/EOS tokens to your prompt and mask it appropriately.
+:::
+
+A config for this would look like:
+
+```yaml
+datasets:
+  - path: A.jsonl
+    type:
+```
+
+::: {.callout-note}
+`type: ` is empty!
+:::
+
+Reference: [Pre-Tokenized Dataset Documentation](tokenized.qmd).
+
+### Template Free Dataset
+
+We reccomend this approach when you want granular control over the prompt formatting, special tokens, and masking, whilst letting Axolotl handle the tokenization. This is very useful if your dataset has unique prompts that differ across samples and where one single general template wouldn't suffice.
+
+In the example below, you could see that there is no proper structure. At the same time, it's very flexible as there are no constraints on how your prompt can look.
+
+```json
+{
+    "segments": [
+        {
+            "label": true,
+            "text": "<s>Hello\n"
+        },
+        {
+            "label": true,
+            "text": "hi there!. "
+        },
+        {
+            "label": false,
+            "text": "goodbye "
+        },
+        {
+            "label": true,
+            "text": "farewell</s>"
+        }
+    ]
+}
+```
+
+Each prompt must be have a key called `segments` which is a list of `{ text, label }`.
+
+```yaml
+datasets:
+  - path: A.jsonl
+    type: input_output
+```
+
+Reference: [Template Free Documentation](template_free.qmd).
+
+### Conversation Dataset
+
+`conversation` messages are a list of messages which usually contain a `role` and `content` key.
+
+::: {.callout-tip}
+Fun fact: Axolotl synonymously refers to "chat" messages as `conversation` messages due to how FastChat initially used this term to build a widely used [fastchat conversation](https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py) method for formatting chat messages prior to the creation of `chat_templates`.
+:::
+
+#### What are `chat_templates`?
+
+The current most popular and convenient method for inference is to use `chat_templates` for formatting prompts. Axolotl supports using `chat_templates` for training to ensure that the model performs in the same environment as in inference.
+
+Here's a quick rundown on `chat_template`: A `chat_template` is a Jinja2 template which formats a list of messages into a prompt.
+
+An example of a prompt formatted into a popular template called ChatML can be seen below:
+
+Single prompt (pretty-printed):
+```json
+{
+    "messages": [
+        {
+            "role": "user",
+            "content": "Hi"
+        },
+        {
+            "role": "assistant",
+            "content": "How can I help you?"
+        },
+        {
+            "role": "user",
+            "content": "Can you add 3+5?"
+        },
+        {
+            "role": "assistant",
+            "content": "The answer is 8."
+        }
+    ]
+}
+```
+
+The ChatML template is as follows:
+```jinja2
+{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% for message in messages %}{{'<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant\n' }}{% endif %}
+```
+
+The above prompt formatted into this template will result in:
+
+```
+<|im_start|>user
+Hi<|im_end|>
+<|im_start|>assistant
+How can I help you?<|im_end|>
+<|im_start|>user
+Can you add 3+5?<|im_end|>
+<|im_start|>assistant
+The answer is 8.<|im_end|>
+```
+
+By using delimiters (`<|im_start|>` and `<|im_end|>`), a prompt separates different speakers which helps the model identify which portion belongs to whom.
+
+#### Common Conversation Dataset formats
+
+Older conversation datasets with the following format are colloquially called `sharegpt` datasets.
+
+```json
+{"conversations": [{"from": "...", "value": "..."}]}
+```
+
+Newer conversation datasets usually follow the OpenAI format.
+
+```json
+{"messages": [{"role": "...", "content": "..."}]}
+```
+
+Axolotl supports both as well as allowing customization of any kind of key.
+
+#### Chat Template Usage
+
+To properly use this method, it is important to identify three things:
+
+1. Which `chat_template` would you use?
+
+2. What are the keys in your dataset, and what are the possible roles? For example, in OpenAI format, the keys would be `messages`, `role`, and `content`, respectively, whereas the possible roles are `system`, `user`, and `assistant`.
+
+3. What do you want to mask? For instance, only assistant messages, only last message, or nothing.
+
+##### Choosing a `chat_template`
+
+There are a lot of `chat_templates` out there. Axolotl supports the common ones: [supported chat templates](https://github.com/axolotl-ai-cloud/axolotl/blob/860609392184cf62a7e0ca676658b170e059ce6c/src/axolotl/utils/chat_templates.py#L17). For example, to use ChatML, it would be `chat_template: chatml`.
+
+However, it is also possible to use the already configured template within the tokenizer by specifying `chat_template: tokenizer_default`. If you want a fallback (in case some tokenizer does not have it pre-configured), you can do `chat_template: tokenizer_default_fallback_chatml` to fallback to the ChatML template if a tokenizer template was not found.
+
+One last but powerful approach is to bring your own template. This can be set via:
+
+```yaml
+chat_template_jinja: # your template
+```
+
+##### Setting `chat_template` dataset keys
+
+We currently default to OpenAI format for dataset keys, so if that's your current dataset format, there's nothing to do here.
+
+If your dataset format is different, here are the keys you should check (with their defaults):
+
+```yaml
+datasets:
+    ...
+    field_messages: messages  # this should point to the key containing the list of conversations
+    message_property_mappings:  # this is a mapping from keys in your dataset to keys in chat_template
+      role: role
+      content: content
+```
+
+In some `chat_templates` (e.g. [Gemma](https://huggingface.co/google/gemma-2b-it/blob/main/tokenizer_config.json#L1507)), the roles are hardcoded to `user` and `assistant`. Consequently, you may find it necessary to map the roles in your dataset to these above. We currently have some defaults that should work for common datasets, but if you get a `KeyError`, it would be necessary to add mapping for your roles. Here is an example of how it would look like:
+
+```yaml
+datasets:
+    ...
+    roles:
+      assistant:
+        - gpt
+        - model
+      user:
+        - human
+```
+
+In the example above, all `gpt` and `model` values are converted to `assistant`. All `human` values are converted to `user.`
+
+##### Handling masking
+
+The common use case for `chat_template` is for chat messages, therefore, it is common to mask all non-assistant messages. Assistant messages refer to the bot messages that you want the model to learn on.
+
+To train on all `assistant` messages, you would set the following configs.
+
+```yaml
+datasets:
+    ...
+    roles_to_train: ["assistant"]
+    train_on_eos: "turn"
+```
+
+The `train_on_eos` config means that it would mask all EOS tokens for turns that aren't assistant-turns. The other options are: `all` and `last` to choose which EOS to train on.
+
+Perhaps, you want to train on `assistant` and `narrator` roles, you can simply add `narrator` to the list of `roles_to_train`. You would also need to add it to the mapping of `roles` above.
+
+```yaml
+datasets:
+    ...
+    roles_to_train: ["assistant", "narrator"]
+    roles:
+      assistant:
+        - gpt
+        - model
+      user:
+        - human
+      narrator: ["narrator"]
+```
+
+::: {.callout-tip}
+As chat_templates may use hardcoded EOS/EOT tokens that are different from the tokenizer's EOS, it is highly recommended to set them. For example, `ChatML` uses `<|im_end|>` to end turns.
+
+```yaml
+special_tokens:
+  eos_token: <|im_end|>
+```
+
+:::
+
+##### Applying `chat_template`
+
+Once all the above steps are completed, you could combine all these configs together to form a bespoke configuration for your custom dataset.
+
+```yaml
+datasets:
+  - path: A.jsonl
+    type: chat_template
+
+    # step 1
+    chat_template: chatml
+
+    # step 2
+    field_messages: messages
+    message_property_mappings:
+      role: role
+      content: content
+
+    roles:
+      assistant:
+        - gpt
+        - model
+        - assistant
+      user:
+        - human
+        - user
+
+    # step 3
+    roles_to_train: ["assistant"]
+    train_on_eos: "turn"
+
+special_tokens:
+  eos_token: <|im_end|>
+```
+
+If this config were to be applied to the sample dataset above, the output would look as such (which can be retrieved via `axolotl preprocess config.yaml --debug`):
+
+```
+<|im_start|>(-100, 128256) user(-100, 882)
+(-100, 198) Hi(-100, 13347) <|im_end|>(-100, 128257)
+(-100, 198) <|im_start|>(-100, 128256) assistant(-100, 78191)
+(-100, 198) How(4438, 4438)  can(649, 649)  I(358, 358)  help(1520, 1520)  you(499, 499) ?(30, 30) <|im_end|>(128257, 128257)
+(-100, 198) <|im_start|>(-100, 128256) user(-100, 882)
+(-100, 198) Can(-100, 6854)  you(-100, 499)  add(-100, 923)  (-100, 220) 3(-100, 18) +(-100, 10) 5(-100, 20) ?(-100, 30) <|im_end|>(-100, 128257)
+(-100, 198) <|im_start|>(-100, 128256) assistant(-100, 78191)
+(-100, 198) The(791, 791)  answer(4320, 4320)  is(374, 374)  (220, 220) 8(23, 23) .(13, 13) <|im_end|>(128257, 128257)
+(-100, 198)
+```
+
+The first number refers to the label, the second refers to the `token_id`. For example, `-100` labels appear on non-assistant portions, meaning that they are masked during. For assistant portions, the label is the same as the `token_id`.
+
+::: {.callout-note}
+
+If during `preprocess`, there are a lot of warnings of `Could not find content __ boundary`, please check the FAQ section for [chat_templates](../faq.qmd#chat-templates).
+
+:::
+
+#### Reference
+
+Please see docs [here](conversation.qmd).
+
+### Instruction Dataset
+
+Instruction datasets are used to train instruction-following models and comprise a prompt, containing an instruction, and a single response. In contrast to chat datasets which may be multi-turn, instruct datasets are typically single-turn.
+
+An example is of a common format called Alpaca:
+```json
+{"instruction": "...", "input": "...", "output": "..."}
+```
+
+Using those keys, a prompt can be built based on it.
+```
+Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
+
+### Instruction:
+{instruction}
+
+### Input:
+{input}
+
+### Response:
+{output}
+```
+
+This can be configured as such:
+```yaml
+datasets:
+  - path: A.jsonl
+    type: alpaca
+```
+
+Axolotl supports many kinds of instruction dataset. All of them can be found here (https://axolotl-ai-cloud.github.io/axolotl/docs/dataset-formats/inst_tune.html) with their respective type and sample row format.
+
+
+Reference: [Instruction Dataset Documentation](inst_tune.qmd).
+
+#### Custom Instruct Prompt Format
+
+Due to the myriad possibilities of instruction formats, Axolotl allows customizing your own instruction format without having to dive into the code directly.
+
+In the example below, a sample row is used to output in `mistral_v1` format.
+```json
+{"input": "...", "output": "..."}
+```
+
+```yaml
+datasets:
+  - path: repo
+    type:
+      system_prompt: ""
+
+      field_system:
+      field_instruction: input
+      field_input:
+      field_output: output
+
+      # multi-line example with input
+      format: |-
+        [INST] {instruction} {input} [/INST]
+
+      # single-line example without input
+      no_input_format: "[INST] {instruction} [/INST]"
+```
+
+The config sets that the `field_instruction` is actually named `input`, and the `field_input` is empty as we don't have an `input` in this sample. Generally, `instruction` can be thought as the question to the model, and `input` as the additional information with `output` being the response. It is not necessary to have an `input` nor `system`. In the end, the most important part is to understand what format you want it to look like and how you can customize this to your use case.
+
+Reference: [Custom Instruct Prompt Format Documentation](inst_tune.qmd#how-to-add-custom-prompt-format).
+
+## Reinforcement Learning from Human Feedback (RLHF)
+
+As there are multiple RLHF methods with their own dataset requirements. Please see [RLHF documentation](../rlhf.qmd) for more detail.

docs/dataset-formats/pretraining.qmd

@@ -19,8 +19,14 @@ For pretraining, there is no prompt template or roles.  The only required field
 Axolotl usually loads the entire dataset into memory. This will be challenging for large datasets. Use the following config to enable streaming:
 
 ```{.yaml filename="config.yaml"}
-pretraining_dataset: # hf path only
-...
+pretraining_dataset:
+  - name:
+    path:
+    split:
+    text_column: # column in dataset with the data, usually `text`
+    type: pretrain
+    trust_remote_code:
+    skip: # number of rows of data to skip over from the beginning
 ```
 
 :::

docs/dataset-formats/stepwise_supervised.qmd

@@ -0,0 +1,26 @@
+---
+title: Stepwise Supervised Format
+description: Format for datasets with stepwise completions and labels
+order: 3
+---
+
+## Stepwise Supervised
+
+The stepwise supervised format is designed for chain-of-thought (COT) reasoning
+datasets where each example contains multiple completion steps and a preference label
+for each step.
+
+### Example
+
+Here's a simple example of a stepwise supervised dataset entry:
+
+```json
+{
+  "prompt": "Which number is larger, 9.8 or 9.11?",
+  "completions": [
+    "The fractional part of 9.8 is 0.8, while the fractional part of 9.11 is 0.11.",
+    "Since 0.11 is greater than 0.8, the number 9.11 is larger than 9.8."
+  ],
+  "labels": [true, false]
+}
+```

docs/dataset-formats/template_free.qmd

@@ -1,7 +1,239 @@
 ---
 title: Template-Free
 description: Construct prompts without a template.
+toc: true
+toc-depth: 3
 order: 4
 ---
 
-See [these docs](../input_output.qmd).
+## Background {#sec-background}
+
+### Masking Inputs {#masking-inputs}
+
+One of the most popular features of
+[axolotl](https://github.com/axolotl-ai-cloud/axolotl) is
+setting the following configuration value:
+
+
+```yaml
+train_on_inputs: false
+```
+
+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.
+
+### You may not want prompt templates {#sec-you-may-not-want-prompt-templates}
+
+However, there are many situations where you don't want to use one of
+these formats or templates. This is because they can:
+
+-   Add unnecessary boilerplate to your prompts.
+-   Create artifacts like special delimiters `<|im_start|>` that can
+    quickly become footguns if you don't include them correctly at
+    inference time.
+-   Enforce a *chat* interface when you do not want one. Sometimes you
+    just want to fine-tune a model to a very specific task and do NOT
+    want multi-turn conversations, roles, etc.
+-   Limit you to only certain roles that the template allows.
+
+### The `input_output` format {#sec-the-inputoutput-format}
+
+You can construct your prompts without a template by using the
+`input_output` format, by setting `type: input_output` in your
+configuration file like this:
+
+**config.yml**
+
+```yaml
+train_on_inputs: false # Mask segments of your data
+datasets:
+  - path: output.jsonl
+    type: input_output  # use template free prompt construction
+```
+
+Unlike `type: completion`, which is also template-free,
+`type: input_output` allows you to mask segments of your text. More
+details on how this works are described below.
+
+## Usage {#sec-usage}
+
+This is how you can use the `input_output` format:
+
+### 1. Prepare Data {#sec-1-prepare-data}
+
+To use the `input_output` format, collect your data in the following
+format into a jsonl file (below is the first row from the file
+`output`.jsonl` pretty printed):
+
+```bash
+$ head -n1 output.jsonl | python -m json.tool
+```
+
+:::{.cell-output .cell-output-stdout}
+    {
+        "segments": [
+            {
+                "label": true,
+                "text": "<s>Hello\n"
+            },
+            {
+                "label": true,
+                "text": "hi there!. "
+            },
+            {
+                "label": false,
+                "text": "goodbye "
+            },
+            {
+                "label": true,
+                "text": "farewell</s>"
+            }
+        ]
+    }
+:::
+
+Set `label:false` when you want to mask a segment of text so that the
+model isn't trained on it. Some things to keep in mind:
+
+> [!IMPORTANT]
+> 1.  **EOS, BOS, spaces, newlines etc. are entirely up to you. Axolotl
+    concatenates all the segments as-is.** The tokenizer doesn't add
+    anything additional. Notice how I added spaces, newlines, `<s>`
+    (BOS), and `</s>` (EOS) myself.
+> 2.  Make sure you check the materialized output to validate that the
+    prompt is getting assembled how you like.
+
+### 2. Use `type: input_output` {#sec-2-use-type-inputoutput}
+
+Let's materialize data with our `output.jsonl` file by setting
+`type: input_output` in our axolotl config:
+
+```yaml
+# training_config.yaml
+base_model: mistralai/Mistral-7B-v0.1
+data_seed: 49
+seed: 49
+
+datasets:
+  - path: output.jsonl
+    type: input_output
+val_set_size: 0.1
+
+sequence_len: 896
+sample_packing: false
+
+micro_batch_size: 2
+gradient_accumulation_steps: 3
+eval_batch_size: 2
+num_epochs: 1
+learning_rate: 0.0002
+
+train_on_inputs: false
+special_tokens:
+  bos_token: "<s>"
+  eos_token: "</s>"
+  unk_token: "<unk>"
+```
+
+You can use the following command to materialize your data. The
+`--debug` flag will print the tokens, along with the labels so you can
+verify that the correct items are being ignored:
+
+```bash
+axolotl preprocess training_config.yaml --debug
+
+...
+[2024-03-05 23:36:46,969] [INFO] [axolotl.check_example_labels:35] [PID:607731] [RANK:0] <s>(1, 1) Hello(22557, 22557)
+(13, 13) hi(12014, 12014) there(736, 736) !(28808, 28808) .(28723, 28723) (28705, 28705) good(-100, 1179) bye(-100, 17664) (-100, 28705) fare(19111, 19111) well(5458, 5458) </s>(2, 2)
+
+```
+
+The format is `decoded_token`(`label`, `token_id`), for example,
+`<s>(1, 1)` means that the token is `<s>`, the label is `1` and the
+token_id is `1`. When the label is `-100` then that token is ignored for
+training.
+
+### 3. Check the prompts {#sec-3-check-the-prompts}
+
+Here is another way to check the materialized output:
+
+```python
+from transformers import AutoTokenizer
+from datasets import load_from_disk
+import yaml
+
+directory = !ls last_run_prepared/
+with open('training_config.yaml', 'r') as f:
+    cfg = yaml.safe_load(f)
+model_id = cfg['base_model']
+tok = AutoTokenizer.from_pretrained(model_id)
+ds = load_from_disk(f'last_run_prepared/{directory[0]}/')
+```
+
+```python
+>>> row = ds[0]
+>>> print(tok.decode(row['input_ids']))
+<s> Hello
+    hi there!.  goodbye  farewell</s>
+```
+
+We can check that the right tokens are ignored by comparing the labels
+to each token:
+
+```python
+import pandas as pd
+pd.DataFrame([{'token': tok.decode(i), 'label': l, 'id':i} for i,l in
+              zip(row['input_ids'], row['labels'])])
+```
+
+| token | label | id    |
+|-------|-------|-------|
+| 0     | \<s\> | 1     |
+| 1     | Hello | 22557 |
+| 2     | \\n   | 13    |
+| 3     | hi    | 12014 |
+| 4     | there | 736   |
+| 5     | !     | 28808 |
+| 6     | .     | 28723 |
+| 7     |       | 28705 |
+| 8     | good  | -100  |
+| 9     | bye   | -100  |
+| 10    |       | -100  |
+| 11    | fare  | 19111 |
+| 12    | well  | 5458  |
+| 13    | \</s\>| 2     |
+
+
+
+If we look at the input data, the above table seems correct! (The jsonl
+version is repeated below for reference):
+
+
+```bash
+$ head -n1 output.jsonl | python -m json.tool
+```
+
+:::{.cell-output .cell-output-stdout}
+    {
+        "segments": [
+            {
+                "label": true,
+                "text": "<s>Hello\n"
+            },
+            {
+                "label": true,
+                "text": "hi there!. "
+            },
+            {
+                "label": false,
+                "text": "goodbye "
+            },
+            {
+                "label": true,
+                "text": "farewell</s>"
+            }
+        ]
+    }
+:::

docs/dataset_preprocessing.qmd

@@ -3,8 +3,11 @@ title: Dataset Preprocessing
 description: How datasets are processed
 ---
 
+## Overview
+
 Dataset pre-processing is the step where Axolotl takes each dataset you've configured alongside
-the (dataset format)[../dataset-formats/] and prompt strategies to:
+the [dataset format](docs/dataset-formats) and prompt strategies to:
+
  - parse the dataset based on the *dataset format*
  - transform the dataset to how you would interact with the model based on the *prompt strategy*
  - tokenize the dataset based on the configured model & tokenizer
@@ -12,10 +15,12 @@ the (dataset format)[../dataset-formats/] and prompt strategies to:
 
 The processing of the datasets can happen one of two ways:
 
-1. Before kicking off training by calling `python -m axolotl.cli.preprocess /path/to/your.yaml --debug`
+1. Before kicking off training by calling `axolotl preprocess config.yaml --debug`
 2. When training is started
 
-What are the benefits of pre-processing? When training interactively or for sweeps
+### What are the benefits of pre-processing?
+
+When training interactively or for sweeps
 (e.g. you are restarting the trainer often), processing the datasets can oftentimes be frustratingly
 slow. Pre-processing will cache the tokenized/formatted datasets according to a hash of dependent
 training parameters so that it will intelligently pull from its cache when possible.
@@ -28,8 +33,12 @@ default path of `./last_run_prepared/`, but will ignore anything already cached
 setting `dataset_prepared_path: ./last_run_prepared`, the trainer will use whatever pre-processed
 data is in the cache.
 
-What are the edge cases? Let's say you are writing a custom prompt strategy or using a user-defined
+### What are the edge cases?
+
+Let's say you are writing a custom prompt strategy or using a user-defined
 prompt template. Because the trainer cannot readily detect these changes, we cannot change the
-calculated hash value for the pre-processed dataset. If you have `dataset_prepared_path: ...` set
+calculated hash value for the pre-processed dataset.
+
+If you have `dataset_prepared_path: ...` set
 and change your prompt templating logic, it may not pick up the changes you made and you will be
 training over the old prompt.

docs/debugging.qmd

@@ -31,11 +31,13 @@ While debugging it's helpful to simplify your test scenario as much as possible.
     - Set `CUDA_VISIBLE_DEVICES` to a single GPU, ex: `export CUDA_VISIBLE_DEVICES=0`.
     - Set `dataset_processes: 1` in your axolotl config or run the training command with `--dataset_processes=1`.
 2. **Use a small dataset**: Construct or use a small dataset from HF Hub. When using a small dataset, you will often have to make sure `sample_packing: False` and `eval_sample_packing: False` to avoid errors.  If you are in a pinch and don't have time to construct a small dataset but want to use from the HF Hub, you can shard the data (this will still tokenize the entire dataset, but will only use a fraction of the data for training.  For example, to shard the dataset into 20 pieces, add the following to your axolotl config):
+
     ```yaml
-    dataset:
+    datasets:
         ...
         shards: 20
     ```
+
 3. **Use a small model**: A good example of a small model is [TinyLlama/TinyLlama-1.1B-Chat-v1.0](https://huggingface.co/TinyLlama/TinyLlama-1.1B-Chat-v1.0).
 4. **Minimize iteration time**: Make sure the training loop finishes as fast as possible, with these settings.
     - `micro_batch_size: 1`
@@ -71,7 +73,7 @@ Make sure you have an [editable install](https://setuptools.pypa.io/en/latest/us
 
 ```bash
 pip3 install packaging
-pip3 install -e '.[flash-attn,deepspeed]'
+pip3 install --no-build-isolation -e '.[flash-attn,deepspeed]'
 ```
 
 #### Remote Hosts
@@ -85,7 +87,7 @@ The easiest way to get started is to modify the [.vscode/launch.json](../.vscode
 
 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
+```json
 // .vscode/launch.json
 {
     "version": "0.2.0",
@@ -132,7 +134,7 @@ For example, to mimic the command `cd devtools && CUDA_VISIBLE_DEVICES=0 acceler
 
 Below is the [./vscode/tasks.json](../.vscode/tasks.json) file that defines the `cleanup-for-dataprep` task.  This task is run before each debugging session when you use the above configuration.  Note how there are two tasks that delete the two folders mentioned above.  The third task `cleanup-for-dataprep` is a composite task that combines the two tasks.  A composite task is necessary because VSCode does not allow you to specify multiple tasks in the `preLaunchTask` argument of the `launch.json` file.
 
-```jsonc
+```json
 // .vscode/tasks.json
 // this file is used by launch.json
 {
@@ -212,7 +214,7 @@ You will now be in the container.  Next, perform an editable install of Axolotl:
 
 ```bash
 pip3 install packaging
-pip3 install -e '.[flash-attn,deepspeed]'
+pip3 install --no-build-isolation -e '.[flash-attn,deepspeed]'
 ```
 
 ### Attach To Container

docs/faq.qmd

@@ -3,6 +3,7 @@ title: FAQ
 description: Frequently asked questions
 ---
 
+### General
 
 **Q: The trainer stopped and hasn't progressed in several minutes.**
 
@@ -19,3 +20,33 @@ description: Frequently asked questions
 **Q: AttributeError: 'DummyOptim' object has no attribute 'step'**
 
 > A: You may be using deepspeed with single gpu. Please don't set `deepspeed:` in yaml or cli.
+
+**Q: The codes is stuck on saving preprocessed datasets.**
+
+> A: This is usually an issue with the GPU. This can be resolved through setting the os environment variable `CUDA_VISIBLE_DEVICES=0`. If you are on runpod, this is usually a pod issue. Starting a new pod should take care of it.
+
+### Chat templates
+
+**Q: `jinja2.exceptions.UndefinedError: 'dict object' has no attribute 'content' / 'role' / ____`**
+
+> A: This means that the property mapping for the stated attribute does not exist when building `chat_template` prompt. For example, if `no attribute 'content'`, please check you have added the correct mapping for `content` under `message_property_mappings`.
+
+**Q: `Empty template generated for turn ___`**
+
+> A: The `content` is empty for that turn.
+
+**Q: `Could not find content start/end boundary for turn __`**
+
+> A: The specific turn's start/end could not be detected. Please ensure you have set the `eos_token` following your `chat_template`. Otherwise, this could be a `chat_template` which doesn't use proper boundaries for each turn (like system). On the rare occurrence, make sure your content is not `[[dummy_message]]`. Please let us know about this.
+
+**Q: `Content end boundary is before start boundary for turn ___`**
+
+> A: This is an edge case which should not occur. Please create an Issue if this happens.
+
+**Q: `Content end boundary is the same as start boundary for turn ___. This is likely an empty turn.`**
+
+> A: This is likely an empty turn.
+
+**Q: The EOS/EOT token is incorrectly being masked or not being masked.**
+
+> A: This is because of the mismatch between `tokenizer.eos_token` and EOS/EOT token in template. Please make sure to set `eos_token` under `special_tokens` to the same EOS/EOT token as in template.

docs/getting-started.qmd

@@ -0,0 +1,155 @@
+---
+title: "Quickstart"
+format:
+  html:
+    toc: true
+    toc-depth: 3
+    number-sections: true
+execute:
+  enabled: false
+---
+
+This guide will walk you through your first model fine-tuning project with Axolotl.
+
+## Quick Example {#sec-quick-example}
+
+Let's start by fine-tuning a small language model using LoRA. This example uses a 1B parameter model to ensure it runs on most GPUs.
+Assuming `axolotl` is installed (if not, see our [Installation Guide](installation.qmd))
+
+1. Download example configs:
+```bash
+axolotl fetch examples
+```
+
+2. Run the training:
+```bash
+axolotl train examples/llama-3/lora-1b.yml
+```
+
+That's it! Let's understand what just happened.
+
+## Understanding the Process {#sec-understanding}
+
+### The Configuration File {#sec-config}
+
+The YAML configuration file controls everything about your training. Here's what (part of) our example config looks like:
+
+```yaml
+base_model: NousResearch/Llama-3.2-1B
+# hub_model_id: username/custom_model_name
+
+datasets:
+  - path: teknium/GPT4-LLM-Cleaned
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.1
+output_dir: ./outputs/lora-out
+
+adapter: lora
+lora_model_dir:
+```
+
+See our [Config options](config.qmd) for more details.
+
+### Training {#sec-training}
+
+When you run `axolotl train`, Axolotl:
+
+1. Downloads the base model
+2. (If specified) applies LoRA adapter layers
+3. Loads and processes the dataset
+4. Runs the training loop
+5. Saves the trained model and / or LoRA weights
+
+## Your First Custom Training {#sec-custom}
+
+Let's modify the example for your own data:
+
+1. Create a new config file `my_training.yml`:
+
+```yaml
+base_model: NousResearch/Nous-Hermes-llama-1b-v1
+adapter: lora
+
+# Training settings
+micro_batch_size: 2
+num_epochs: 3
+learning_rate: 0.0003
+
+# Your dataset
+datasets:
+  - path: my_data.jsonl        # Your local data file
+    type: alpaca               # Or other format
+```
+
+This specific config is for LoRA fine-tuning a model with instruction tuning data using
+the `alpaca` dataset format, which has the following format:
+
+```json
+{
+    "instruction": "Write a description of alpacas.",
+    "input": "",
+    "output": "Alpacas are domesticated South American camelids..."
+}
+```
+
+Please see our [Dataset Formats](dataset-formats) for more dataset formats and how to
+format them.
+
+2. Prepare your JSONL data in the specified format (in this case, the expected `alpaca
+format):
+
+```json
+{"instruction": "Classify this text", "input": "I love this!", "output": "positive"}
+{"instruction": "Classify this text", "input": "Not good at all", "output": "negative"}
+```
+
+Please consult the supported [Dataset Formats](dataset-formats/) for more details.
+
+3. Run the training:
+
+```bash
+axolotl train my_training.yml
+```
+
+## Common Tasks {#sec-common-tasks}
+
+### Testing Your Model {#sec-testing}
+
+After training, test your model:
+
+```bash
+axolotl inference my_training.yml --lora-model-dir="./outputs/lora-out"
+```
+
+### Preprocessing Data {#sec-preprocessing}
+
+For large datasets, preprocess first:
+
+```bash
+axolotl preprocess my_training.yml
+```
+
+### Using a UI {#sec-ui}
+
+Launch a Gradio interface:
+
+```bash
+axolotl inference my_training.yml --lora-model-dir="./outputs/lora-out" --gradio
+```
+
+## Next Steps {#sec-next-steps}
+
+Now that you have the basics, you might want to:
+
+- Try different model architectures
+- Experiment with hyperparameters
+- Use more advanced training methods
+- Scale up to larger models
+
+Check our other guides for details on these topics:
+
+- [Configuration Guide](config.qmd) - Full configuration options
+- [Dataset Formats](dataset-formats) - Working with different data formats
+- [Multi-GPU Training](multi-gpu.qmd)
+- [Multi-Node Training](multi-node.qmd)

docs/inference.qmd

@@ -0,0 +1,147 @@
+---
+title: "Inference"
+format:
+  html:
+    toc: true
+    toc-depth: 3
+    number-sections: true
+execute:
+  enabled: false
+---
+
+This guide covers how to use your trained models for inference, including model loading, interactive testing, and common troubleshooting steps.
+
+## Quick Start {#sec-quickstart}
+
+### Basic Inference {#sec-basic}
+
+::: {.panel-tabset}
+
+## LoRA Models
+
+```{.bash}
+axolotl inference your_config.yml --lora-model-dir="./lora-output-dir"
+```
+
+## Full Fine-tuned Models
+
+```{.bash}
+axolotl inference your_config.yml --base-model="./completed-model"
+```
+
+:::
+
+## Advanced Usage {#sec-advanced}
+
+### Gradio Interface {#sec-gradio}
+
+Launch an interactive web interface:
+
+```{.bash}
+axolotl inference your_config.yml --gradio
+```
+
+### File-based Prompts {#sec-file-prompts}
+
+Process prompts from a text file:
+
+```{.bash}
+cat /tmp/prompt.txt | axolotl inference your_config.yml \
+  --base-model="./completed-model" --prompter=None
+```
+
+### Memory Optimization {#sec-memory}
+
+For large models or limited memory:
+
+```{.bash}
+axolotl inference your_config.yml --load-in-8bit=True
+```
+
+## Merging LoRA Weights {#sec-merging}
+
+Merge LoRA adapters with the base model:
+
+```{.bash}
+axolotl merge-lora your_config.yml --lora-model-dir="./completed-model"
+```
+
+### Memory Management for Merging {#sec-memory-management}
+
+::: {.panel-tabset}
+
+## Configuration Options
+
+```{.yaml}
+gpu_memory_limit: 20GiB  # Adjust based on your GPU
+lora_on_cpu: true        # Process on CPU if needed
+```
+
+## Force CPU Merging
+
+```{.bash}
+CUDA_VISIBLE_DEVICES="" axolotl merge-lora ...
+```
+
+:::
+
+## Tokenization {#sec-tokenization}
+
+### Common Issues {#sec-tokenization-issues}
+
+::: {.callout-warning}
+Tokenization mismatches between training and inference are a common source of problems.
+:::
+
+To debug:
+
+1. Check training tokenization:
+```{.bash}
+axolotl preprocess your_config.yml --debug
+```
+
+2. Verify inference tokenization by decoding tokens before model input
+
+3. Compare token IDs between training and inference
+
+### Special Tokens {#sec-special-tokens}
+
+Configure special tokens in your YAML:
+
+```{.yaml}
+special_tokens:
+  bos_token: "<s>"
+  eos_token: "</s>"
+  unk_token: "<unk>"
+tokens:
+  - "<|im_start|>"
+  - "<|im_end|>"
+```
+
+## Troubleshooting {#sec-troubleshooting}
+
+### Common Problems {#sec-common-problems}
+
+::: {.panel-tabset}
+
+## Memory Issues
+
+- Use 8-bit loading
+- Reduce batch sizes
+- Try CPU offloading
+
+## Token Issues
+
+- Verify special tokens
+- Check tokenizer settings
+- Compare training and inference preprocessing
+
+## Performance Issues
+
+- Verify model loading
+- Check prompt formatting
+- Ensure temperature/sampling settings
+
+:::
+
+For more details, see our [debugging guide](debugging.qmd).

docs/input_output.qmd

@@ -3,263 +3,4 @@ title: Template-free prompt construction
 description: "Template-free prompt construction with the `input_output` format"
 ---
 
-<!-- TOC -->
-
-- [Background](#background)
-    - [Masking Inputs](#masking-inputs)
-    - [You may not want prompt templates](#you-may-not-want-prompt-templates)
-    - [The `input_output` format](#the-input_output-format)
-- [Usage](#usage)
-    - [1. Prepare Data](#1-prepare-data)
-    - [2. Use `type: input_output`](#2-use-type-input_output)
-    - [3. Check the prompts](#3-check-the-prompts)
-
-<!-- /TOC -->
-
-<a id="markdown-background" name="background"></a>
-
-## Background
-
-<a id="markdown-masking-inputs" name="masking-inputs"></a>
-
-### Masking Inputs
-
-One of the most popular features of
-[axolotl](https://github.com/axolotl-ai-cloud/axolotl) is
-setting the following configuration value:
-
-
-```yaml
-train_on_inputs: false
-```
-
-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.
-
-<a id="markdown-you-may-not-want-prompt-templates" name="you-may-not-want-prompt-templates"></a>
-
-### You may not want prompt templates
-
-However, there are many situations where you don't want to use one of
-these formats or templates. This is because they can:
-
--   Add unnecessary boilerplate to your prompts.
--   Create artifacts like special delimiters `<|im_start|>` that can
-    quickly become footguns if you don't include them correctly at
-    inference time.
--   Enforce a *chat* interface when you do not want one. Sometimes you
-    just want to fine-tune a model to a very specific task and do NOT
-    want multi-turn conversations, roles, etc.
--   Limit you to only certain roles that the template allows.
-
-<a id="markdown-the-inputoutput-format" name="the-inputoutput-format"></a>
-
-### The `input_output` format
-
-You can construct your prompts without a template by using the
-`input_output` format, by setting `type: input_output` in your
-configuration file like this:
-
-**config.yml**
-
-```yaml
-train_on_inputs: false # Mask segments of your data
-datasets:
-  - path: output.jsonl
-    type: input_output  # use template free prompt construction
-```
-
-Unlike `type: completion`, which is also template-free,
-`type: input_output` allows you to mask segments of your text. More
-details on how this works are described below.
-
-<a id="markdown-usage" name="usage"></a>
-
-## Usage
-
-This is how you can use the `input_output` format:
-
-<a id="markdown-1-prepare-data" name="1-prepare-data"></a>
-
-### 1. Prepare Data
-
-To use the `input_output` format, collect your data in the following
-format into a jsonl file (below is the first row from the file
-`output`.jsonl` pretty printed):
-
-```bash
-$ head -n1 output.jsonl | python -m json.tool
-```
-
-:::{.cell-output .cell-output-stdout}
-    {
-        "segments": [
-            {
-                "label": true,
-                "text": "<s>Hello\n"
-            },
-            {
-                "label": true,
-                "text": "hi there!. "
-            },
-            {
-                "label": false,
-                "text": "goodbye "
-            },
-            {
-                "label": true,
-                "text": "farewell</s>"
-            }
-        ]
-    }
-:::
-
-Set `label:false` when you want to mask a segment of text so that the
-model isn't trained on it. Some things to keep in mind:
-
-> [!IMPORTANT]
-> 1.  **EOS, BOS, spaces, newlines etc. are entirely up to you. Axolotl
-    concatenates all the segments as-is.** The tokenizer doesn't add
-    anything additional. Notice how I added spaces, newlines, `<s>`
-    (BOS), and `</s>` (EOS) myself.
-> 2.  Make sure you check the materialized output to validate that the
-    prompt is getting assembled how you like.
-
-<a id="markdown-2-use-type-inputoutput" name="2-use-type-inputoutput"></a>
-
-### 2. Use `type: input_output`
-
-Let's materialize data with our `output.jsonl` file by setting
-`type: input_output` in our axolotl config:
-
-```yaml
-# training_config.yaml
-base_model: mistralai/Mistral-7B-v0.1
-data_seed: 49
-seed: 49
-
-datasets:
-  - path: output.jsonl
-    type: input_output
-val_set_size: 0.1
-
-sequence_len: 896
-sample_packing: false
-
-micro_batch_size: 2
-gradient_accumulation_steps: 3
-eval_batch_size: 2
-num_epochs: 1
-learning_rate: 0.0002
-
-train_on_inputs: false
-special_tokens:
-  bos_token: "<s>"
-  eos_token: "</s>"
-  unk_token: "<unk>"
-```
-
-You can use the following command to materialize your data. The
-`--debug` flag will print the tokens, along with the labels so you can
-verify that the correct items are being ignored:
-
-```bash
-$ python -m axolotl.cli.preprocess training_config.yaml --debug
-
-...
-[2024-03-05 23:36:46,969] [INFO] [axolotl.check_example_labels:35] [PID:607731] [RANK:0] <s>(1, 1) Hello(22557, 22557)
-(13, 13) hi(12014, 12014) there(736, 736) !(28808, 28808) .(28723, 28723) (28705, 28705) good(-100, 1179) bye(-100, 17664) (-100, 28705) fare(19111, 19111) well(5458, 5458) </s>(2, 2)
-
-```
-
-The format is `decoded_token`(`label`, `token_id`), for example,
-`<s>(1, 1)` means that the token is `<s>`, the label is `1` and the
-token_id is `1`. When the label is `-100` then that token is ignored for
-training.
-
-<a id="markdown-3-check-the-prompts" name="3-check-the-prompts"></a>
-
-### 3. Check the prompts
-
-Here is another way to check the materialized output:
-
-```python
-from transformers import AutoTokenizer
-from datasets import load_from_disk
-import yaml
-
-directory = !ls last_run_prepared/
-with open('training_config.yaml', 'r') as f:
-    cfg = yaml.safe_load(f)
-model_id = cfg['base_model']
-tok = AutoTokenizer.from_pretrained(model_id)
-ds = load_from_disk(f'last_run_prepared/{directory[0]}/')
-```
-
-```python
->>> row = ds[0]
->>> print(tok.decode(row['input_ids']))
-<s> Hello
-    hi there!.  goodbye  farewell</s>
-```
-
-We can check that the right tokens are ignored by comparing the labels
-to each token:
-
-```python
-import pandas as pd
-pd.DataFrame([{'token': tok.decode(i), 'label': l, 'id':i} for i,l in
-              zip(row['input_ids'], row['labels'])])
-```
-
-| token | label | id    |
-|-------|-------|-------|
-| 0     | \<s\> | 1     |
-| 1     | Hello | 22557 |
-| 2     | \\n   | 13    |
-| 3     | hi    | 12014 |
-| 4     | there | 736   |
-| 5     | !     | 28808 |
-| 6     | .     | 28723 |
-| 7     |       | 28705 |
-| 8     | good  | -100  |
-| 9     | bye   | -100  |
-| 10    |       | -100  |
-| 11    | fare  | 19111 |
-| 12    | well  | 5458  |
-| 13    | \</s\>| 2     |
-
-
-
-If we look at the input data, the above table seems correct! (The jsonl
-version is repeated below for reference):
-
-
-```bash
-$ head -n1 output.jsonl | python -m json.tool
-```
-
-:::{.cell-output .cell-output-stdout}
-    {
-        "segments": [
-            {
-                "label": true,
-                "text": "<s>Hello\n"
-            },
-            {
-                "label": true,
-                "text": "hi there!. "
-            },
-            {
-                "label": false,
-                "text": "goodbye "
-            },
-            {
-                "label": true,
-                "text": "farewell</s>"
-            }
-        ]
-    }
-:::
+The documentation moved to [here](dataset-formats/template_free.qmd).

docs/installation.qmd

@@ -0,0 +1,118 @@
+---
+title: "Installation"
+format:
+  html:
+    toc: true
+    toc-depth: 3
+    number-sections: true
+execute:
+  enabled: false
+---
+
+This guide covers all the ways you can install and set up Axolotl for your environment.
+
+## Requirements {#sec-requirements}
+
+- NVIDIA GPU (Ampere architecture or newer for `bf16` and Flash Attention) or AMD GPU
+- Python ≥3.10
+- PyTorch ≥2.4.1
+
+## Installation Methods {#sec-installation-methods}
+
+### PyPI Installation (Recommended) {#sec-pypi}
+
+```{.bash}
+pip3 install --no-build-isolation axolotl[flash-attn,deepspeed]
+```
+
+We use `--no-build-isolation` in order to detect the installed PyTorch version (if
+installed) in order not to clobber it, and so that we set the correct version of
+dependencies that are specific to the PyTorch version or other installed
+co-dependencies.
+
+### Edge/Development Build {#sec-edge-build}
+
+For the latest features between releases:
+
+```{.bash}
+git clone https://github.com/axolotl-ai-cloud/axolotl.git
+cd axolotl
+pip3 install packaging ninja
+pip3 install --no-build-isolation -e '.[flash-attn,deepspeed]'
+```
+
+### Docker {#sec-docker}
+
+```{.bash}
+docker run --gpus '"all"' --rm -it axolotlai/axolotl:main-latest
+```
+
+For development with Docker:
+
+```{.bash}
+docker compose up -d
+```
+
+::: {.callout-tip}
+### Advanced Docker Configuration
+```{.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 \
+  axolotlai/axolotl:main-latest
+```
+:::
+
+## Cloud Environments {#sec-cloud}
+
+### Cloud GPU Providers {#sec-cloud-gpu}
+
+For providers supporting Docker:
+
+- Use `axolotlai/axolotl-cloud:main-latest`
+- Available on:
+  - [Latitude.sh](https://latitude.sh/blueprint/989e0e79-3bf6-41ea-a46b-1f246e309d5c)
+  - [JarvisLabs.ai](https://jarvislabs.ai/templates/axolotl)
+  - [RunPod](https://runpod.io/gsc?template=v2ickqhz9s&ref=6i7fkpdz)
+
+### Google Colab {#sec-colab}
+
+Use our [example notebook](../examples/colab-notebooks/colab-axolotl-example.ipynb).
+
+## Platform-Specific Instructions {#sec-platform-specific}
+
+### macOS {#sec-macos}
+
+```{.bash}
+pip3 install --no-build-isolation -e '.'
+```
+
+See @sec-troubleshooting for Mac-specific issues.
+
+### Windows {#sec-windows}
+
+::: {.callout-important}
+We recommend using WSL2 (Windows Subsystem for Linux) or Docker.
+:::
+
+## Environment Managers {#sec-env-managers}
+
+### Conda/Pip venv {#sec-conda}
+
+1. Install Python ≥3.10
+2. Install PyTorch: https://pytorch.org/get-started/locally/
+3. Install Axolotl:
+   ```{.bash}
+   pip3 install packaging
+   pip3 install --no-build-isolation -e '.[flash-attn,deepspeed]'
+   ```
+4. (Optional) Login to Hugging Face:
+   ```{.bash}
+   huggingface-cli login
+   ```
+
+## Troubleshooting {#sec-troubleshooting}
+
+If you encounter installation issues, see our [FAQ](faq.qmd) and [Debugging Guide](debugging.qmd).

docs/lora_optims.qmd

@@ -0,0 +1,127 @@
+---
+title: "LoRA Optimizations"
+description: "Custom autograd functions and Triton kernels in Axolotl for optimized LoRA fine-tuning"
+---
+
+Inspired by [Unsloth](https://github.com/unslothai/unsloth), we've implemented two
+optimizations for LoRA and QLoRA fine-tuning, supporting both single GPU and multi-GPU
+(in the DDP and DeepSpeed settings) training. These include (1) SwiGLU and GEGLU activation function
+Triton kernels, and (2) LoRA MLP and attention custom autograd functions. Our goal was
+to leverage operator fusion and tensor re-use in order to improve speed and reduce
+memory usage during the forward and backward passes of these calculations.
+
+We currently support several common model architectures, including (but not limited to):
+
+- `llama`
+- `mistral`
+- `qwen2`
+- `gemma`
+- `gemma2`
+
+<details>
+
+The set of models we support is currently limited by our attention patching strategy,
+which assumes (and replaces) specific code blocks for query / key / value and output
+projections:
+
+```python
+ORIGINAL_QKV_CODE = """
+    query_states = self.q_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+    key_states = self.k_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+    value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+""".lstrip(
+    "\n"
+)
+
+ORIGINAL_O_CODE = """
+    attn_output = self.o_proj(attn_output)
+""".lstrip(
+    "\n"
+)
+```
+
+Is replaced with:
+
+```python
+PATCHED_QKV_CODE = """
+    query_states, key_states, value_states = self.apply_qkv(hidden_states)
+    query_states = query_states.view(hidden_shape).transpose(1, 2)
+    key_states = key_states.view(hidden_shape).transpose(1, 2)
+    value_states = value_states.view(hidden_shape).transpose(1, 2)
+""".lstrip(
+    "\n"
+)
+
+PATCHED_O_CODE = """
+    attn_output = self.apply_o(attn_output)
+""".lstrip(
+    "\n"
+)
+```
+
+Where `apply_qkv` and `apply_o` are defined in the `axolotl.kernels.lora` module.
+
+We welcome testing of other model architectures and / or PRs to expand our patching
+logic to be compatible with more of them.
+
+</details>
+
+## Usage
+
+These optimizations can be enabled in your Axolotl config YAML file. The
+`lora_mlp_kernel` option enables the optimized MLP path, while `lora_qkv_kernel` and
+`lora_o_kernel` enable the fused query-key-value projection and optimized output
+projection, respectively.
+
+```yaml
+lora_mlp_kernel: true
+lora_qkv_kernel: true
+lora_o_kernel: true
+```
+
+## Requirements
+
+- One or more NVIDIA or AMD GPUs (in order to use the Triton kernels)
+    - Note: Set `TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL=1` to enable [memory-efficient attention on AMD GPUs](https://github.com/ROCm/aotriton/issues/16#issuecomment-2346675491)
+- Targeted LoRA adapters cannot use Dropout
+    - This may limit model expressivity / cause overfitting
+- Targeted LoRA adapters cannot have bias terms
+    - This may limit model expressivity
+
+Models with pre-existing LoRA adapters that use Dropout or have bias terms may need to
+be re-finetuned without these features in order to be useful.
+
+## Implementation details
+
+### Custom autograd functions
+
+The LoRA MLP autograd function optimizes the entire MLP computation path. It fuses the
+LoRA and base weight computations together and provides a single, efficient backward
+pass for the entire MLP block.
+
+For attention components, similar optimizations are provided through a function that
+handles the query, key, and value projections, and a function that handles the output
+projection. They are designed to work with the existing `transformers` attention
+implementation via some monkey-patching logic.
+
+### Triton kernels
+
+Two activation functions (SwiGLU and GeGLU) are implemented with Triton kernels for
+improved speed and memory performance. These kernels handle both the forward and
+backward passes.
+
+### Integration
+
+The custom autograd functions and Triton kernels are designed to work together. The
+autograd function manages the high-level computation flow and gradient tracking, while
+calling the Triton kernels for the activation function computation. During the backward
+pass, the kernel computes both the activation output and the required gradients, which
+the autograd function then uses to compute the final gradients for the entire
+computation path.
+
+## Future Work
+
+- Support for additional model architectures
+- Support for the FSDP setting
+- Support for dropout and bias
+- Additional operator fusions

docs/lr_groups.qmd

@@ -0,0 +1,29 @@
+---
+title: Learning Rate Groups
+description: "Setting different learning rates by module name"
+---
+
+## Background
+
+Inspired by LoRA+, Axolotl allows practitioners to specify separate learning rates for each module or groups of
+modules in a model.
+
+## Example
+
+```yaml
+lr_groups:
+  - name: o_proj
+    modules:
+      - self_attn.o_proj.weight
+    lr: 1e-6
+  - name: q_proj
+    modules:
+      - model.layers.2.self_attn.q_proj.weight
+    lr: 1e-5
+
+learning_rate: 2e-5
+```
+
+In this example, we have a default learning rate of 2e-5 across the entire model, but we have a separate learning rate
+of 1e-6 for all the self attention `o_proj` modules across all layers, and a learning are of 1e-5 to the 3rd layer's
+self attention `q_proj` module.

docs/mac.qmd

@@ -19,4 +19,5 @@ Current support:
 - [ ] DeepSpeed
 
 Untested:
+
 - FSDP

docs/multi-gpu.qmd

@@ -0,0 +1,104 @@
+---
+title: "Multi-GPU"
+format:
+  html:
+    toc: true
+    toc-depth: 3
+    number-sections: true
+    code-tools: true
+execute:
+  enabled: false
+---
+
+This guide covers advanced training configurations for multi-GPU setups using Axolotl.
+
+## Overview {#sec-overview}
+
+Axolotl supports several methods for multi-GPU training:
+
+- DeepSpeed (recommended)
+- FSDP (Fully Sharded Data Parallel)
+- FSDP + QLoRA
+
+## DeepSpeed {#sec-deepspeed}
+
+DeepSpeed is the recommended approach for multi-GPU training due to its stability and performance. It provides various optimization levels through ZeRO stages.
+
+### Configuration {#sec-deepspeed-config}
+
+Add to your YAML config:
+
+```{.yaml}
+deepspeed: deepspeed_configs/zero1.json
+```
+
+### Usage {#sec-deepspeed-usage}
+
+```{.bash}
+# Passing arg via config
+axolotl train config.yml
+
+# Passing arg via cli
+axolotl train config.yml --deepspeed deepspeed_configs/zero1.json
+```
+
+### ZeRO Stages {#sec-zero-stages}
+
+We provide default configurations for:
+
+- ZeRO Stage 1 (`zero1.json`)
+- ZeRO Stage 2 (`zero2.json`)
+- ZeRO Stage 3 (`zero3.json`)
+
+Choose based on your memory requirements and performance needs.
+
+## FSDP {#sec-fsdp}
+
+### Basic FSDP Configuration {#sec-fsdp-config}
+
+```{.yaml}
+fsdp:
+  - full_shard
+  - auto_wrap
+fsdp_config:
+  fsdp_offload_params: true
+  fsdp_state_dict_type: FULL_STATE_DICT
+  fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer
+```
+
+### FSDP + QLoRA {#sec-fsdp-qlora}
+
+For combining FSDP with QLoRA, see our [dedicated guide](fsdp_qlora.qmd).
+
+## Performance Optimization {#sec-performance}
+
+### Liger Kernel Integration {#sec-liger}
+
+Please see [docs](custom_integrations.qmd#liger) for more info.
+
+## Troubleshooting {#sec-troubleshooting}
+
+### NCCL Issues {#sec-nccl}
+
+For NCCL-related problems, see our [NCCL troubleshooting guide](nccl.qmd).
+
+### Common Problems {#sec-common-problems}
+
+::: {.panel-tabset}
+
+## Memory Issues
+
+- Reduce `micro_batch_size`
+- Reduce `eval_batch_size`
+- Adjust `gradient_accumulation_steps`
+- Consider using a higher ZeRO stage
+
+## Training Instability
+
+- Start with DeepSpeed ZeRO-2
+- Monitor loss values
+- Check learning rates
+
+:::
+
+For more detailed troubleshooting, see our [debugging guide](debugging.qmd).

docs/multi-node.qmd

@@ -3,6 +3,18 @@ title: Multi Node
 description: How to use Axolotl on multiple machines
 ---
 
+The below are three ways to train multi-node in Axolotl.
+
+::: {.callout-important}
+Each machine needs a copy of Axolotl, we suggest using the same commit to ensure compatibility.
+
+You will also need to have the same configuration file for your model on each machine.
+
+Make sure the main machine is reachable by other machines.
+:::
+
+## Accelerate
+
 You will need to create a configuration for accelerate, either by using `accelerate config` and follow the instructions or you can use one of the preset below:
 
 ~/.cache/huggingface/accelerate/default_config.yaml
@@ -26,7 +38,7 @@ tpu_use_sudo: false
 use_cpu: false
 ```
 
-Configure your model to use FSDP with for example:
+Configure your model to use FSDP in the Axolotl yaml. For example:
 ```yaml
 fsdp:
   - full_shard
@@ -37,12 +49,40 @@ fsdp_config:
   fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer
 ```
 
-## Machine configuration
-
-On each machine you need a copy of Axolotl, we suggest using the same commit to ensure compatibility.
-
-You will also need to have the same configuration file for your model on each machine.
-
-On the main machine only, make sure the port you set as `main_process_port` is open in TCP and reachable by other machines.
-
 All you have to do now is launch using accelerate as you would usually do on each machine and voila, the processes will start once you have launched accelerate on every machine.
+
+## Raytrain
+
+Please see ray train doc [here](ray-integration.qmd).
+
+## Torchrun
+
+If you are using Infiniband, we recommend torchrun to utilize the full bandwidth.
+
+Set the following env (change buffersize/socketname depending on your system):
+
+```bash
+export NCCL_IB_DISABLE=0
+export NCCL_SOCKET_IFNAME="eth0,en,eth,em,bond"
+export NCCL_BUFFSIZE=2097152
+```
+
+Run the following on each node:
+
+```bash
+torchrun --nnodes $num_nodes --nproc_per_node $gpu_per_node --rdzv_id $rdzv_id --rdzv_backend c10d --rdzv_endpoint "$head_node_ip:$head_node_port" -m axolotl.cli.train config.yaml
+```
+
+Please make sure to substitute the placeholder variables.
+
+- `num_nodes`: Number of nodes (containing GPUs)
+- `gpu_per_node`: Number of gpus per node
+- `head_node_ip`: IP of the head node (make sure other machines can connect to this)
+- `head_node_port`: Port of the head node (make sure other machines can connect to this. Default 29400)
+- `rdzv_id`: A unique job ID that is used by the job across nodes.
+
+::: {.callout-note}
+You need to call `axolotl.cli.train` instead of `axolotl train` as the latter calls accelerate under the hood
+:::
+
+More info on the available configs can be found on the Pytorch docs [here](https://pytorch.org/docs/stable/elastic/run.html)

docs/nccl.qmd

@@ -13,13 +13,13 @@ Often, this timeout will happen after 30 minutes (the default setting) and is ac
 
 Forcing cross-GPU communication via [NVLink](https://en.wikipedia.org/wiki/NVLink) may help without increasing timeouts. To verify that your configuration is leveraging NVLink run the following command:
 
-```shell
+```bash
 nvidia-smi nvlink --status
 ```
 
 To force NCCL to use NVLink, simply set this in the environment:
 
-```shell
+```bash
 export NCCL_P2P_LEVEL=NVL
 ```
 
@@ -33,13 +33,13 @@ If NVLink is not available in your environment there are other options for ``NCC
 
 To validate that acceptable data transfer speeds exist for your training job, running [NCCL Tests](https://github.com/NVIDIA/nccl-tests/blob/master/README.md) can help pinpoint bottlenecks, for example:
 
-```shell
+```bash
 ./build/all_reduce_perf -b 8 -e 128M -f 2 -g 3
 ```
 
 It can be useful when debugging NCCL communication timeouts to activate additional logging in both PyTorch and NCCL:
 
-```shell
+```bash
 export NCCL_DEBUG=INFO
 export NCCL_DEBUG_SUBSYS=ALL
 export TORCH_DISTRIBUTED_DEBUG=INFO

docs/ray-integration.qmd

@@ -0,0 +1,91 @@
+---
+title: Ray Train
+description: How to use Axolotl with Ray Train
+---
+
+Axolotl supports using Ray as an alternative to `accelerate` for orchestrating training. This is especially useful for multi-node training since you only have to setup code and dependencies in a single node and launch training as if you were using a single node.
+
+With the `--use-ray` CLI flag, Axolotl will use Ray Train's [`TorchTrainer`](https://docs.ray.io/en/latest/train/api/doc/ray.train.torch.TorchTrainer.html#ray.train.torch.TorchTrainer) to run training.
+
+## Ray cluster setup
+
+A prerequisite using the Ray Train integration is to setup a Ray cluster on your desired node(s). For a detailed guide on how you can get started with ray clusters, check the official Ray docs [here](https://docs.ray.io/en/latest/cluster/getting-started.html).
+
+Every Ray cluster has one _head_ node and a set of worker nodes. The head node is just like any other worker node, but it also runs certain special processes related to scheduling and orchestration. Ray-enabled scripts are run on the head node and depending on the resources (number of CPUs, GPUs, etc) they request, will be scheduled to run certain tasks on the worker nodes. For more on key concepts behind a Ray cluster, you can refer this [doc](https://docs.ray.io/en/latest/cluster/key-concepts.html#cluster-key-concepts).
+
+## Sanity check
+
+To run a sanity check on whether your ray cluster is setup properly, execute the following on the head node:
+
+```bash
+ray status
+```
+
+The output should have a summary of your Ray cluster - list of all the nodes in your cluster, the number of CPUs and GPUs in your cluster, etc. For example, if you have a cluster with 1 CPU-only head node and 2 4xL40S worker nodes, the output can look like this:
+
+
+```
+Node status
+---------------------------------------------------------------
+Active:
+ 1 head
+Idle:
+ 2 4xL40S:48CPU-384GB
+Pending:
+ (no pending nodes)
+Recent failures:
+ (no failures)
+
+Resources
+---------------------------------------------------------------
+Usage:
+ 0.0/96.0 CPU
+ 0.0/8.0 GPU
+ 0B/800.00GiB memory
+ 0B/229.57GiB object_store_memory
+
+Demands:
+ (no resource demands)
+```
+
+You should also be able to see the same on the [Ray dashboard](https://docs.ray.io/en/latest/ray-observability/getting-started.html).
+
+
+## Configuring training with Ray Train
+
+You can find an example configuration at `configs/llama-3/lora-1b-ray.yaml`.
+
+The key parameters to note here are:
+
+```yaml
+use_ray: true
+ray_num_workers: 4
+# optional
+resources_per_worker:
+    GPU: 1
+```
+
+- `use_ray`: This is the flag that enables the Ray Train integration. You can either use the corresponding `--use-ray` flag in the CLI or set `use_ray` in the config file.
+- `ray_num_workers`: This is the number of workers/GPUs to use for training.
+- `resources_per_worker`: This is the Ray [resource request](https://docs.ray.io/en/latest/ray-core/scheduling/resources.html) for each worker. This can be used to request a specific GPU type or a custom resource for each worker. For example, if your ray cluster has GPUs of different types, and you only want to use NVIDIA L40S GPUs, you can do
+
+```yaml
+resources_per_worker:
+    accelerator_type:L40S: 0.001
+```
+
+## Launching training
+
+You can simply run the following command on the head node:
+
+```bash
+axolotl train examples/llama-3/lora-1b-ray.yml --use-ray
+```
+
+This will launch training on the head node and workers will be scheduled automatically by Ray Train to run on the appropriate head or worker nodes.
+
+You can also monitor training progress on the Ray dashboard.
+
+Coming back to the example on a Ray cluster with 1 head node and 2 4xL40S worker nodes, let's say you want to make use of all 8 GPUs. You would be able to just set `ray_num_workers: 8` and run the previous command. The Cluster tab will show the following:
+
+![Ray dashboard](./images/ray-cluster-dashboard.png)

docs/reward_modelling.qmd

@@ -0,0 +1,47 @@
+---
+title: "Reward Modelling"
+description: "Reward models are used to guide models towards behaviors which is preferred by humans, by training over large datasets annotated with human preferences. "
+---
+
+### Overview
+
+Reward modelling is a technique used to train models to predict the reward or value of a given input. This is particularly useful in reinforcement learning scenarios where the model needs to evaluate the quality of its actions or predictions.
+We support the reward modelling techniques supported by `trl`.
+
+### (Outcome) Reward Models
+
+Outcome reward models are trained using data which contains preference annotations for an entire interaction between the user and model (e.g. rather than per-turn or per-step).
+
+```yaml
+base_model: google/gemma-2-2b
+model_type: AutoModelForSequenceClassification
+num_labels: 1
+tokenizer_type: AutoTokenizer
+
+reward_model: true
+chat_template: gemma
+datasets:
+  - path: argilla/distilabel-intel-orca-dpo-pairs
+    type: bradley_terry.chat_template
+
+val_set_size: 0.1
+eval_steps: 100
+```
+
+### Process Reward Models (PRM)
+
+Process reward models are trained using data which contains preference annotations for each step in a series of interactions. Typically, PRMs are trained to provide reward signals over each step of a reasoning trace and are used for downstream reinforcement learning.
+```yaml
+base_model: Qwen/Qwen2.5-3B
+model_type: AutoModelForTokenClassification
+num_labels: 2
+
+process_reward_model: true
+datasets:
+  - path: trl-lib/math_shepherd
+    type: stepwise_supervised
+    split: train
+
+val_set_size: 0.1
+eval_steps: 100
+```

docs/rlhf.qmd

@@ -1,26 +1,39 @@
 ---
 title: "RLHF (Beta)"
 description: "Reinforcement Learning from Human Feedback is a method whereby a language model is optimized from data using human feedback."
+back-to-top-navigation: true
+toc: true
+toc-depth: 4
 ---
 
-### Overview
+## Overview
 
 Reinforcement Learning from Human Feedback is a method whereby a language model is optimized from data using human
 feedback. Various methods include, but not limited to:
 
+- [Direct Preference Optimization (DPO)](#dpo)
+- [Identity Preference Optimization (IPO)](#ipo)
+- [Kahneman-Tversky Optimization (KTO)](#kto)
+- [Odds Ratio Preference Optimization (ORPO)](#orpo)
 - Proximal Policy Optimization (PPO) (not yet supported in axolotl)
-- Direct Preference Optimization (DPO)
-- Identity Preference Optimization (IPO)
 
 
-### RLHF using Axolotl
+## RLHF using Axolotl
 
->[!IMPORTANT]
->This is a BETA feature and many features are not fully implemented. You are encouraged to open new PRs to improve the integration and functionality.
+::: {.callout-important}
+This is a BETA feature and many features are not fully implemented. You are encouraged to open new PRs to improve the integration and functionality.
+:::
 
-The various RL training methods are implemented in trl and wrapped via axolotl. Below are various examples with how you can use various preference datasets to train models that use ChatML
+We rely on the [TRL](https://github.com/huggingface/trl) library for implementations of various RL training methods, which we wrap around to expose in axolotl. Each method has their own supported ways of loading datasets and prompt formats.
+
+::: {.callout-tip}
+You can find what each method supports by going into `src/axolotl/prompt_strategies/{method}` where `{method}` is one of our supported methods. The `type: ` can be retrieved from `{method}.{function_name}`.
+:::
+
+### DPO
+
+Example config:
 
-#### DPO
 ```yaml
 rl: dpo
 datasets:
@@ -29,15 +42,268 @@ datasets:
     type: chatml.intel
   - path: argilla/ultrafeedback-binarized-preferences
     split: train
-    type: chatml.argilla
+    type: chatml
 ```
 
-#### IPO
+DPO supports the following types with the following dataset format:
+
+#### chatml.argilla
+
+```json
+{
+    "system": "...", // optional
+    "instruction": "...",
+    "chosen_response": "...",
+    "rejected_response": "..."
+}
+```
+
+#### chatml.argilla_chat
+
+```json
+{
+    "chosen": [
+        {"role": "user", "content": "..."},
+        {"role": "assistant", "content": "..."}
+    ],
+    "rejected": [
+        {"role": "user", "content": "..."},
+        {"role": "assistant", "content": "..."}
+    ]
+}
+```
+
+#### chatml.icr
+
+```json
+{
+    "system": "...", // optional
+    "input": "...",
+    "chosen": "...",
+    "rejected": "..."
+}
+```
+
+#### chatml.intel
+
+```json
+{
+    "system": "...", // optional
+    "question": "...",
+    "chosen": "...",
+    "rejected": "..."
+}
+```
+
+#### chatml.prompt_pairs
+
+```json
+{
+    "system": "...", // optional
+    "prompt": "...",
+    "chosen": "...",
+    "rejected": "..."
+}
+```
+
+#### chatml.ultra
+
+```json
+{
+    "system": "...", // optional
+    "prompt": "...",
+    "chosen": [
+        {"role": "user", "content": "..."},
+        {"role": "assistant", "content": "..."}
+    ],
+    "rejected": [
+        {"role": "user", "content": "..."},
+        {"role": "assistant", "content": "..."}
+    ]
+}
+```
+
+#### llama3.argilla
+
+```json
+{
+    "system": "...", // optional
+    "instruction": "...",
+    "chosen_response": "...",
+    "rejected_response": "..."
+}
+```
+
+#### llama3.argilla_chat
+
+```json
+{
+    "chosen": [
+        {"role": "user", "content": "..."},
+        {"role": "assistant", "content": "..."}
+    ],
+    "rejected": [
+        {"role": "user", "content": "..."},
+        {"role": "assistant", "content": "..."}
+    ]
+}
+```
+
+#### llama3.icr
+
+```json
+{
+    "system": "...", // optional
+    "input": "...",
+    "chosen": "...",
+    "rejected": "..."
+}
+```
+
+#### llama3.intel
+
+```json
+{
+    "system": "...", // optional
+    "question": "...",
+    "chosen": "...",
+    "rejected": "..."
+}
+```
+
+#### llama3.prompt_pairs
+
+```json
+{
+    "system": "...", // optional
+    "prompt": "...",
+    "chosen": "...",
+    "rejected": "..."
+}
+```
+
+#### llama3.ultra
+
+```json
+{
+    "system": "...", // optional
+    "prompt": "...",
+    "chosen": [
+        {"role": "user", "content": "..."},
+        {"role": "assistant", "content": "..."}
+    ],
+    "rejected": [
+        {"role": "user", "content": "..."},
+        {"role": "assistant", "content": "..."}
+    ]
+}
+```
+
+#### zephyr.nectar
+
+```json
+{
+    "prompt": "...",
+    "answers": [
+        {
+            "answer": "...",
+            "rank": 1
+        },
+        {
+            "answer": "...",
+            "rank": 2
+        }
+        // ... more answers with ranks
+    ]
+}
+```
+
+#### chat_template.default
+
+```yaml
+rl: dpo
+datasets:
+  - path: ...
+    split: train
+    type: chat_template.default
+    field_messages: "messages"
+    field_chosen: "chosen"
+    field_rejected: "rejected"
+    message_property_mappings:
+      role: role
+      content: content
+    roles:
+      user: ["user"]
+      assistant: ["assistant"]
+      system: ["system"]
+```
+
+Sample input format:
+
+```json
+{
+    "messages": [
+        {
+            "role": "system",
+            "content": "..."
+        },
+        {
+            "role": "user",
+            "content": "..."
+        },
+        // ... more messages
+    ],
+    "chosen": {
+        "role": "assistant",
+        "content": "..."
+    },
+    "rejected": {
+        "role": "assistant",
+        "content": "..."
+    }
+}
+```
+
+#### user_defined.default
+
+For custom behaviors,
+
+```yaml
+rl: dpo
+datasets:
+  - path: ...
+    split: train
+    type: user_defined.default
+
+    field_prompt: "prompt"
+    field_system: "system"
+    field_chosen: "chosen"
+    field_rejected: "rejected"
+    prompt_format: "{prompt}"
+    chosen_format: "{chosen}"
+    rejected_format: "{rejected}"
+```
+
+The input format is a simple JSON input with customizable fields based on the above config.
+
+```json
+{
+    "system": "...",  // optional
+    "prompt": "...",
+    "chosen": "...",
+    "rejected": "..."
+}
+```
+
+### IPO
+
+As IPO is just DPO with a different loss function, all supported options for DPO works here.
+
 ```yaml
 rl: ipo
 ```
 
-#### ORPO
+### ORPO
 
 Paper: https://arxiv.org/abs/2403.07691
 
@@ -52,7 +318,226 @@ datasets:
     type: chat_template.argilla
 ```
 
-#### Using local dataset files
+ORPO supports the following types with the following dataset format:
+
+#### chat_template.argilla
+
+```json
+{
+    "system": "...",  // optional
+    "prompt": "...",  // if available, will be taken as user message for single-turn instead of from list below
+
+    // chosen/rejected should be same till last content and only even-number of alternating user/assistant turns
+    "chosen": [
+        {"role": "user", "content": "..."},
+        {"role": "assistant", "content": "..."}
+    ],
+    "rejected": [
+        {"role": "user", "content": "..."},
+        {"role": "assistant", "content": "..."}
+    ]
+}
+```
+
+### KTO
+
+```yaml
+rl: kto
+rl_beta: 0.5
+kto_desirable_weight: 0.2
+
+remove_unused_columns: false
+
+datasets:
+  - path: argilla/ultrafeedback-binarized-preferences-cleaned-kto
+    type: llama3.ultra
+    split: train
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: true
+```
+
+KTO supports the following types with the following dataset format:
+
+#### chatml.argilla
+
+```json
+{
+    "system": "...", // optional
+    "instruction": "...",
+    "completion": "..."
+}
+```
+
+#### chatml.argilla_chat
+
+```json
+{
+    "chosen": [
+        {"role": "user", "content": "..."}
+    ],
+    "completion": [
+        {"role": "assistant", "content": "..."}
+    ]
+}
+```
+
+#### chatml.intel
+
+```json
+{
+    "system": "...", // optional
+    "question": "...",
+    "completion": "..."
+}
+```
+
+#### chatml.prompt_pairs
+
+```json
+{
+    "system": "...", // optional
+    "prompt": "...",
+    "completion": "..."
+}
+```
+
+#### chatml.ultra
+
+```json
+{
+    "system": "...", // optional
+    "prompt": "...",
+    "completion": "..."
+}
+```
+
+#### llama3.argilla
+
+```json
+{
+    "system": "...", // optional
+    "instruction": "...",
+    "completion": "..."
+}
+```
+
+#### llama3.argilla_chat
+
+```json
+{
+    "completion": [
+        {"role": "user", "content": "..."},
+        {"role": "assistant", "content": "..."}
+    ]
+}
+```
+
+#### llama3.intel
+
+```json
+{
+    "system": "...", // optional
+    "question": "...",
+    "completion": "..."
+}
+```
+
+#### llama3.prompt_pairs
+
+```json
+{
+    "system": "...", // optional
+    "prompt": "...",
+    "completion": "..."
+}
+```
+
+#### llama3.ultra
+
+```json
+{
+    "system": "...", // optional
+    "prompt": "...",
+    "completion": "..."
+}
+```
+
+#### user_defined.default
+
+For custom behaviors,
+
+```yaml
+rl: kto
+datasets:
+  - path: ...
+    split: train
+    type: user_defined.default
+
+    field_prompt: "prompt"
+    field_system: "system"
+    field_completion: "completion"
+    field_label: "label"
+    prompt_format: "{prompt}"
+    completion_format: "{completion}"
+```
+
+The input format is a simple JSON input with customizable fields based on the above config.
+
+```json
+{
+    "system": "...",  // optional
+    "prompt": "...",
+    "completion": "...",
+    "label": "..."
+}
+```
+
+### GRPO
+
+GRPO uses custom reward functions and transformations. Please have them ready locally.
+
+For ex, to load OpenAI's GSM8K and use a random reward for completions:
+
+```python
+# rewards.py
+import random
+
+def rand_reward_func(completions, **kwargs) -> list[float]:
+    return [random.uniform(0, 1) for _ in completions]
+
+def oai_gsm8k_transform(cfg, *args, **kwargs):
+    def transform_fn(example, tokenizer=None):
+        label = example["answer"].split("####")[-1].strip().replace(",", "")
+        return {
+            "prompt": [{"role": "user", "content": example["question"]},],
+            "answer": label,
+        }
+    return transform_fn, {"remove_columns": ["question"]}
+```
+
+```yaml
+rl: grpo
+
+trl:
+    beta: 0.001
+    max_completion_length: 256
+    use_vllm: True
+    vllm_device: auto
+    vllm_gpu_memory_utilization: 0.15
+    num_generations: 4
+    reward_funcs: ["rewards.rand_reward_func"]    # format: '{file_name}.{fn_name}'
+datasets:
+  - path: openai/gsm8k
+    name: main
+    type: rewards.oai_gsm8k_transform  # format: '{file_name}.{fn_name}'
+```
+
+To see other examples of custom reward functions, please see [TRL GRPO Docs](https://github.com/huggingface/trl/blob/main/docs/source/grpo_trainer.md#using-a-custom-reward-function).
+
+### Using local dataset files
+
 ```yaml
 datasets:
   - ds_type: json
@@ -62,9 +547,9 @@ datasets:
     type: chatml.intel
 ```
 
-#### Trl autounwrap for peft
+### TRL auto-unwrapping for PEFT
 
-Trl supports autounwrapping peft models, so that a ref model does not need to be additionally loaded, leading to less VRAM needed. This is on by default. To turn it off, pass the following config.
+TRL supports auto-unwrapping PEFT models for RL training paradigms which rely on a reference model. This significantly reduces memory pressure as an additional refreference model does not need to be loaded, and reference model log-probabilities can be obtained by disabling PEFT adapters. This is enabled by default. To turn it off, pass the following config:
 
 ```yaml
 # load ref model when adapter training.

docs/torchao.qmd

@@ -3,6 +3,12 @@ title: "PyTorch ao"
 description: "Custom data types and layouts for training and inference"
 ---
 
+To use experimental optimizers (`AdamWFp8`, `AdamW4bit`, `AdamW8bit`) from Pytorch Ao, please install the package as shown below.
+
+::: {.callout-tip}
+Some experimental optimizers are already present in regular Pytorch, so please re-check if you actually need this package!
+:::
+
 ### Installation
 
 Stable Release from the PyTorch index

docs/unsloth.qmd

@@ -8,6 +8,12 @@ description: "Hyper-optimized QLoRA finetuning for single GPUs"
 Unsloth provides hand-written optimized kernels for LLM finetuning that slightly improve speed and VRAM over
 standard industry baselines.
 
+::: {.callout-important}
+Due to breaking changes in transformers `v4.48.0`, users will need to downgrade to `<=v4.47.1` to use this patch.
+
+This will later be deprecated in favor of [LoRA Optimizations](lora_optims.qmd).
+:::
+
 
 ### Installation
 
@@ -17,7 +23,7 @@ The following will install the correct unsloth and extras from source.
 python scripts/unsloth_install.py | sh
 ```
 
-### Using unsloth w Axolotl
+### Usage
 
 Axolotl exposes a few configuration options to try out unsloth and get most of the performance gains.
 

examples/cerebras/btlm-ft.yml

@@ -1,6 +1,10 @@
 base_model: cerebras/btlm-3b-8k-base
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: GPT2Tokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 tokenizer_use_fast: true
 tokenizer_legacy: true
@@ -42,7 +46,7 @@ output_dir: ./outputs/btlm-out
 gradient_accumulation_steps: 1
 micro_batch_size: 1
 num_epochs: 1
-optimizer: adamw_torch
+optimizer: adamw_torch_fused
 adam_beta2: 0.95
 adam_eps: 0.000000001
 max_grad_norm: 1.0

examples/cerebras/qlora.yml

@@ -1,4 +1,7 @@
 base_model: cerebras/Cerebras-GPT-1.3B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 load_in_8bit: false
 load_in_4bit: true
 strict: false

examples/cloud/modal.yaml

@@ -0,0 +1,28 @@
+project_name:
+volumes:
+  - name: axolotl-data
+    mount: /workspace/data
+  - name: axolotl-artifacts
+    mount: /workspace/artifacts
+
+# environment variables from local to set as secrets
+secrets:
+  - HF_TOKEN
+  - WANDB_API_KEY
+
+# Which branch of axolotl to use remotely
+branch:
+
+# additional custom commands when building the image
+dockerfile_commands:
+
+gpu: h100
+gpu_count: 1
+
+# Train specific configurations
+memory: 128
+timeout: 86400
+
+# Preprocess specific configurations
+memory_preprocess: 32
+timeout_preprocess: 14400

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

@@ -1,6 +1,9 @@
 base_model: codellama/CodeLlama-13b-hf
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: CodeLlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: true
 load_in_4bit: false

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

@@ -1,6 +1,9 @@
 base_model: codellama/CodeLlama-13b-hf
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: CodeLlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true

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

@@ -1,6 +1,9 @@
 base_model: codellama/CodeLlama-34b-hf
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: CodeLlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: true
 load_in_4bit: false

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

@@ -1,6 +1,9 @@
 base_model: codellama/CodeLlama-34b-hf
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: CodeLlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true

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

@@ -1,6 +1,9 @@
 base_model: codellama/CodeLlama-7b-hf
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: CodeLlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: true
 load_in_4bit: false

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

@@ -1,6 +1,9 @@
 base_model: codellama/CodeLlama-7b-hf
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: CodeLlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true

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

@@ -24,7 +24,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "!pip install axolotl[deepspeed]"
+    "!pip install --no-build-isolation axolotl[deepspeed]"
    ]
   },
   {

examples/dbrx/16bit-lora.yaml

@@ -1,4 +1,7 @@
 base_model: LnL-AI/dbrx-base-converted-v2
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: false

examples/dbrx/8bit-lora.yaml

@@ -1,4 +1,7 @@
 base_model: LnL-AI/dbrx-base-converted-v2
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: true

examples/dbrx/fft-ds-zero3.yaml

@@ -1,4 +1,7 @@
 base_model: LnL-AI/dbrx-base-converted-v2
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: false

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

@@ -1,4 +1,6 @@
 base_model: deepseek-ai/DeepSeek-V2-Lite
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 trust_remote_code: true
 
 load_in_8bit: false
@@ -25,7 +27,7 @@ wandb_log_model:
 gradient_accumulation_steps: 8
 micro_batch_size: 1
 num_epochs: 1
-optimizer: adamw_torch
+optimizer: adamw_torch_fused
 lr_scheduler: cosine
 learning_rate: 2e-5
 

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

@@ -1,4 +1,7 @@
 base_model: axolotl-quants/DeepSeek-V2.5-bnb-nf4-bf16
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: false
@@ -18,8 +21,9 @@ datasets:
     type: chat_template
     split: train[:20%]
     field_messages: conversations
-    message_field_role: from
-    message_field_content: value
+    message_property_mappings:
+      role: from
+      content: value
 
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.0
@@ -44,7 +48,7 @@ peft_use_rslora: true
 gradient_accumulation_steps: 1
 micro_batch_size: 8
 num_epochs: 1
-optimizer: adamw_torch
+optimizer: adamw_torch_fused
 lr_scheduler: cosine
 learning_rate: 2e-5
 

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

@@ -1,7 +1,12 @@
 base_model: tiiuae/falcon-7b
-trust_remote_code: true
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+# required by falcon custom model code: https://huggingface.co/tiiuae/falcon-7b/tree/main
+trust_remote_code: true
 
 load_in_8bit: true
 load_in_4bit: false

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

@@ -1,10 +1,15 @@
 # 1b: tiiuae/falcon-rw-1b
 # 40b: tiiuae/falcon-40b
 base_model: tiiuae/falcon-7b
-# required by falcon custom model code: https://huggingface.co/tiiuae/falcon-7b/tree/main
-trust_remote_code: true
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+# required by falcon custom model code: https://huggingface.co/tiiuae/falcon-7b/tree/main
+trust_remote_code: true
+
 
 load_in_8bit: false
 # enable 4bit for QLoRA

examples/falcon/config-7b.yml

@@ -1,7 +1,12 @@
 base_model: tiiuae/falcon-7b
-trust_remote_code: true
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+# required by falcon custom model code: https://huggingface.co/tiiuae/falcon-7b/tree/main
+trust_remote_code: true
 
 load_in_8bit: false
 load_in_4bit: false

examples/gemma/qlora.yml

@@ -1,7 +1,10 @@
 # use google/gemma-7b if you have access
 base_model: mhenrichsen/gemma-7b
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true

examples/gemma2/qlora.yml

@@ -1,6 +1,9 @@
 base_model: google/gemma-2-9b
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true
@@ -13,8 +16,9 @@ datasets:
     type: chat_template
     drop_system_message: true
     field_messages: conversations
-    message_field_role: from
-    message_field_content: value
+    message_property_mappings:
+      role: from
+      content: value
 
 val_set_size: 0.0
 output_dir: ./outputs/out

examples/gemma2/reward-model.yaml

@@ -1,6 +1,10 @@
 base_model: google/gemma-2-2b
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForSequenceClassification
+num_labels: 1
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: false

examples/gptj/qlora.yml

@@ -1,4 +1,7 @@
 base_model: EleutherAI/gpt-j-6b
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 load_in_8bit: false
 load_in_4bit: true
 strict: false

examples/jamba/qlora.yaml

@@ -1,4 +1,7 @@
 base_model: ai21labs/Jamba-v0.1
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: false

examples/jamba/qlora_deepspeed.yaml

@@ -1,4 +1,6 @@
 base_model: ai21labs/Jamba-v0.1
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 trust_remote_code: true
 
 load_in_8bit: false

examples/jamba/qlora_fsdp_large.yaml

@@ -1,5 +1,8 @@
 base_model: ai21labs/AI21-Jamba-1.5-Large
+# optionally might have model_type or tokenizer_type
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_4bit: true
 strict: false
@@ -10,8 +13,9 @@ datasets:
     type: chat_template
     drop_system_message: true
     field_messages: conversations
-    message_field_role: from
-    message_field_content: value
+    message_property_mappings:
+      role: from
+      content: value
 
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.0
@@ -31,7 +35,7 @@ lora_target_linear: false
 gradient_accumulation_steps: 4
 micro_batch_size: 1
 num_epochs: 2
-optimizer: adamw_torch
+optimizer: adamw_torch_fused
 lr_scheduler: cosine
 learning_rate: 0.00001
 

examples/jeopardy-bot/config.yml

@@ -1,6 +1,10 @@
 base_model: huggyllama/llama-7b
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 load_in_8bit: false
 datasets:
   - path: openaccess-ai-collective/jeopardy

examples/llama-2/fft_optimized.yml

@@ -1,6 +1,9 @@
 base_model: NousResearch/Llama-2-7b-hf
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: false

examples/llama-2/gptq-lora.yml

@@ -1,8 +1,13 @@
 base_model: TheBloke/Llama-2-7B-GPTQ
-gptq: true
-gptq_disable_exllama: true
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+gptq: true
+gptq_disable_exllama: true
+
 tokenizer_use_fast: true
 tokenizer_legacy: true
 load_in_8bit: false
@@ -37,7 +42,7 @@ output_dir: ./outputs/model-out
 gradient_accumulation_steps: 1
 micro_batch_size: 1
 num_epochs: 4
-optimizer: adamw_torch
+optimizer: adamw_torch_fused
 adam_beta2: 0.95
 adam_eps: 0.00001
 max_grad_norm: 1.0

examples/llama-2/lisa.yml

@@ -1,6 +1,9 @@
 base_model: NousResearch/Llama-2-7b-hf
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: false

examples/llama-2/loftq.yml

@@ -1,6 +1,9 @@
 base_model: NousResearch/Llama-2-7b-hf
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: false

examples/llama-2/lora.yml

@@ -1,6 +1,9 @@
 base_model: NousResearch/Llama-2-7b-hf
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: true
 load_in_4bit: false

examples/llama-2/qlora-fsdp.yml

@@ -1,6 +1,9 @@
 base_model: NousResearch/Llama-2-7b-hf
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true
@@ -36,7 +39,7 @@ wandb_log_model:
 gradient_accumulation_steps: 4
 micro_batch_size: 4
 num_epochs: 4
-optimizer: adamw_torch
+optimizer: adamw_torch_fused
 lr_scheduler: cosine
 learning_rate: 0.00001
 

examples/llama-2/qlora.yml

@@ -1,6 +1,9 @@
 base_model: NousResearch/Llama-2-7b-hf
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true

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

@@ -1,5 +1,9 @@
 base_model: alpindale/Llama-3.2-11B-Vision-Instruct
+# optionally might have model_type or tokenizer_type or processor_type
 processor_type: AutoProcessor
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 strict: false
 
 # these 3 lines are needed for now to handle vision chat templates w images

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

@@ -1,4 +1,6 @@
 base_model: NousResearch/Meta-Llama-3.1-8B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 plugins:
   - axolotl.integrations.liger.LigerPlugin
@@ -15,8 +17,9 @@ datasets:
     type: chat_template
     split: train[:20%]
     field_messages: conversations
-    message_field_role: from
-    message_field_content: value
+    message_property_mappings:
+      role: from
+      content: value
 
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.02
@@ -35,7 +38,7 @@ wandb_log_model:
 gradient_accumulation_steps: 4
 micro_batch_size: 2
 num_epochs: 1
-optimizer: adamw_torch
+optimizer: adamw_torch_fused
 lr_scheduler: cosine
 learning_rate: 2e-5
 

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

@@ -1,4 +1,6 @@
 base_model: NousResearch/Meta-Llama-3.1-8B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: false

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

@@ -1,6 +1,9 @@
 base_model: meta-llama/Meta-Llama-3-8B-Instruct
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: true
 load_in_4bit: false
@@ -14,8 +17,9 @@ datasets:
     field_messages: conversation
     field_chosen: chosen
     field_rejected: rejected
-    message_field_role: role
-    message_field_content: content
+    message_property_mappings:
+      role: role
+      content: content
     roles:
       system:
         - system

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

@@ -1,6 +1,9 @@
 base_model: NousResearch/Meta-Llama-3-8B-Instruct
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: true
 load_in_4bit: false
@@ -11,8 +14,9 @@ datasets:
   - path: fozziethebeat/alpaca_messages_2k_test
     type: chat_template
     field_messages: messages
-    message_field_role: role
-    message_field_content: content
+    message_property_mappings:
+      role: role
+      content: content
     roles:
       user:
         - user

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

@@ -1,6 +1,9 @@
 base_model: meta-llama/Llama-3.2-1B
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: true
 load_in_4bit: false
@@ -14,8 +17,9 @@ datasets:
     field_messages: conversation
     field_chosen: chosen
     field_rejected: rejected
-    message_field_role: role
-    message_field_content: content
+    message_property_mappings:
+      role: role
+      content: content
     roles:
       system:
         - system
@@ -28,8 +32,9 @@ datasets:
     field_messages: conversation
     field_chosen: chosen
     field_rejected: rejected
-    message_field_role: role
-    message_field_content: content
+    message_property_mappings:
+      role: role
+      content: content
     roles:
       system:
         - system

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

@@ -1,6 +1,9 @@
 base_model: meta-llama/Llama-3.2-1B
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: true
 load_in_4bit: false

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

@@ -0,0 +1,82 @@
+base_model: NousResearch/Llama-3.2-1B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: teknium/GPT4-LLM-Cleaned
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.1
+output_dir: ./outputs/lora-out
+
+adapter: lora
+lora_model_dir:
+
+sequence_len: 2048
+sample_packing: true
+pad_to_sequence_len: true
+
+lora_r: 16
+lora_alpha: 32
+# Currently, we don't support dropout with our custom Triton kernels
+# lora_dropout: 0.05
+lora_fan_in_fan_out:
+lora_target_modules:
+  - gate_proj
+  - down_proj
+  - up_proj
+  - q_proj
+  - v_proj
+  - k_proj
+  - o_proj
+
+# These options enable our custom Triton kernels / autograd
+# functions for MLP and attention calculations
+lora_mlp_kernel: true
+lora_qkv_kernel: true
+lora_o_kernel: true
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 2
+micro_batch_size: 2
+num_epochs: 1
+optimizer: adamw_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+loss_watchdog_threshold: 5.0
+loss_watchdog_patience: 3
+
+warmup_steps: 10
+evals_per_epoch: 4
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:
+  pad_token: "<|end_of_text|>"

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

@@ -0,0 +1,79 @@
+base_model: NousResearch/Llama-3.2-1B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: teknium/GPT4-LLM-Cleaned
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.1
+output_dir: ./outputs/lora-out
+
+adapter: lora
+lora_model_dir:
+
+sequence_len: 2048
+sample_packing: true
+eval_sample_packing: true
+pad_to_sequence_len: true
+
+lora_r: 16
+lora_alpha: 32
+lora_dropout: 0.05
+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: 2
+micro_batch_size: 2
+num_epochs: 1
+optimizer: adamw_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+loss_watchdog_threshold: 5.0
+loss_watchdog_patience: 3
+
+warmup_steps: 10
+evals_per_epoch: 4
+saves_per_epoch: 1
+debug:
+deepspeed: deepspeed_configs/zero3.json
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:
+  pad_token: "<|end_of_text|>"
+
+use_ray: true
+ray_num_workers: 4

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

@@ -0,0 +1,76 @@
+base_model: NousResearch/Llama-3.2-1B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: teknium/GPT4-LLM-Cleaned
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.1
+output_dir: ./outputs/lora-out
+
+adapter: lora
+lora_model_dir:
+
+sequence_len: 2048
+sample_packing: true
+eval_sample_packing: true
+pad_to_sequence_len: true
+
+lora_r: 16
+lora_alpha: 32
+lora_dropout: 0.05
+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: 2
+micro_batch_size: 2
+num_epochs: 1
+optimizer: adamw_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+loss_watchdog_threshold: 5.0
+loss_watchdog_patience: 3
+
+warmup_steps: 10
+evals_per_epoch: 4
+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

@@ -1,6 +1,9 @@
 base_model: NousResearch/Meta-Llama-3-8B
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: true
 load_in_4bit: false

examples/llama-3/qlora-1b-kto.yaml

@@ -0,0 +1,77 @@
+base_model: meta-llama/Llama-3.2-1B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+rl: kto
+rl_beta: 0.5
+kto_desirable_weight: 0.2
+
+datasets:
+  - path: argilla/ultrafeedback-binarized-preferences-cleaned-kto
+    type: llama3.ultra
+    split: train
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.0
+output_dir: ./outputs/qlora-out
+
+remove_unused_columns: false
+
+adapter: qlora
+lora_model_dir:
+
+sequence_len: 2048
+sample_packing: false  # not supported with kto
+eval_sample_packing: false
+pad_to_sequence_len: false
+
+lora_r: 32
+lora_alpha: 64
+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: 1
+micro_batch_size: 2
+num_epochs: 1
+optimizer: adamw_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: true
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 20
+evals_per_epoch: 4
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:
+  pad_token: "<|end_of_text|>"

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

@@ -1,4 +1,6 @@
-base_model: meta-llama/Llama-3.2-1B
+base_model: NousResearch/Llama-3.2-1B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true
@@ -22,7 +24,6 @@ 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

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

@@ -1,5 +1,8 @@
 base_model: hugging-quants/Meta-Llama-3.1-405B-BNB-NF4-BF16
+# optionally might have model_type or tokenizer_type
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_4bit: true
 strict: false
@@ -27,7 +30,7 @@ lora_target_linear: true
 gradient_accumulation_steps: 4
 micro_batch_size: 1
 num_epochs: 2
-optimizer: adamw_torch
+optimizer: adamw_torch_fused
 lr_scheduler: cosine
 learning_rate: 0.00001
 

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

@@ -1,6 +1,9 @@
 base_model: casperhansen/llama-3-70b-fp16
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: AutoTokenizer  # PreTrainedTokenizerFast
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true
@@ -36,7 +39,7 @@ wandb_log_model:
 gradient_accumulation_steps: 4
 micro_batch_size: 1
 num_epochs: 4
-optimizer: adamw_torch
+optimizer: adamw_torch_fused
 lr_scheduler: cosine
 learning_rate: 0.00001