basic llama multipack

Allow in message objects the additional key `weight`, which can be set
to 0 (or 1) to cause that message to be masked out (or left unmasked)
for training (similar to [1]). This is helpful for training the model to be robust and
capable of error recovery upon a bad assistant message.
A missing `weight` key defaults to weight 1, to guarantee downward compatibility.

[1]: https://github.com/mistralai/mistral-finetune

.github/workflows/base.yml

@@ -16,17 +16,27 @@ jobs:
             cuda_version: 11.8.0
             python_version: "3.10"
             pytorch: 2.1.2
-            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 9.0+PTX"
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
           - cuda: "121"
             cuda_version: 12.1.0
             python_version: "3.10"
             pytorch: 2.1.2
-            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 9.0+PTX"
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
           - cuda: "121"
             cuda_version: 12.1.0
             python_version: "3.11"
             pytorch: 2.1.2
-            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 9.0+PTX"
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+          - cuda: "121"
+            cuda_version: 12.1.0
+            python_version: "3.11"
+            pytorch: 2.2.2
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
+          - cuda: "121"
+            cuda_version: 12.1.0
+            python_version: "3.11"
+            pytorch: 2.3.0
+            torch_cuda_arch_list: "7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX"
     steps:
       - name: Checkout
         uses: actions/checkout@v3

.github/workflows/lint.yml

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

.github/workflows/main.yml

@@ -28,7 +28,12 @@ jobs:
           - cuda: 121
             cuda_version: 12.1.0
             python_version: "3.11"
-            pytorch: 2.1.2
+            pytorch: 2.2.2
+            axolotl_extras:
+          - cuda: 121
+            cuda_version: 12.1.0
+            python_version: "3.11"
+            pytorch: 2.3.0
             axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:
@@ -63,7 +68,7 @@ jobs:
             ${{ (matrix.is_latest) && format('{0}-latest', steps.metadata.outputs.tags) || '' }}
           labels: ${{ steps.metadata.outputs.labels }}
 
-  build-axolotl-runpod:
+  build-axolotl-cloud:
     needs: build-axolotl
     if: ${{ ! contains(github.event.commits[0].message, '[skip docker]]') && github.repository_owner == 'OpenAccess-AI-Collective' }}
     # this job needs to be run on self-hosted GPU runners...
@@ -84,7 +89,12 @@ jobs:
           - cuda: 121
             cuda_version: 12.1.0
             python_version: "3.11"
-            pytorch: 2.1.2
+            pytorch: 2.2.2
+            axolotl_extras:
+          - cuda: 121
+            cuda_version: 12.1.0
+            python_version: "3.11"
+            pytorch: 2.3.0
             axolotl_extras:
     runs-on: axolotl-gpu-runner
     steps:
@@ -113,7 +123,47 @@ jobs:
           push: ${{ github.event_name != 'pull_request' }}
           tags: |
              ${{ steps.metadata.outputs.tags }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
-             winglian/axolotl-runpod:main-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
              ${{ (matrix.is_latest) && format('{0}-latest', steps.metadata.outputs.tags) || '' }}
-             ${{ (matrix.is_latest) && format('{0}-latest', 'winglian/axolotl-runpod:main') || '' }}
+          labels: ${{ steps.metadata.outputs.labels }}
+
+  build-axolotl-cloud-no-tmux:
+    needs: build-axolotl
+    if: ${{ ! contains(github.event.commits[0].message, '[skip docker]]') && github.repository_owner == 'OpenAccess-AI-Collective' }}
+    # this job needs to be run on self-hosted GPU runners...
+    strategy:
+      matrix:
+        include:
+          - cuda: 121
+            cuda_version: 12.1.0
+            python_version: "3.11"
+            pytorch: 2.3.0
+            axolotl_extras:
+    runs-on: axolotl-gpu-runner
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Docker metadata
+        id: metadata
+        uses: docker/metadata-action@v5
+        with:
+          images: winglian/axolotl-cloud-term
+      - name: Login to Docker Hub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v2
+      - name: Build
+        uses: docker/build-push-action@v5
+        with:
+          context: .
+          build-args: |
+            BASE_TAG=${{ github.ref_name }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
+            CUDA=${{ matrix.cuda }}
+          file: ./docker/Dockerfile-cloud-no-tmux
+          push: ${{ github.event_name != 'pull_request' }}
+          tags: |
+             ${{ steps.metadata.outputs.tags }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
+             ${{ (matrix.is_latest) && format('{0}-latest', steps.metadata.outputs.tags) || '' }}
           labels: ${{ steps.metadata.outputs.labels }}

.github/workflows/nightlies.yml

@@ -0,0 +1,128 @@
+name: docker-nightlies
+
+on:
+  workflow_dispatch:
+  schedule:
+    - cron: '0 0 * * *'  # Runs at 00:00 UTC every day
+
+jobs:
+  build-axolotl:
+    if: ${{ ! contains(github.event.commits[0].message, '[skip docker]]') && github.repository_owner == 'OpenAccess-AI-Collective' }}
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - cuda: 118
+            cuda_version: 11.8.0
+            python_version: "3.10"
+            pytorch: 2.1.2
+            axolotl_extras:
+            axolotl_args: "--extra-index-url https://download.pytorch.org/whl/cu118"
+            is_latest: true
+          - cuda: 121
+            cuda_version: 12.1.0
+            python_version: "3.10"
+            pytorch: 2.1.2
+            axolotl_extras:
+          - cuda: 121
+            cuda_version: 12.1.0
+            python_version: "3.11"
+            pytorch: 2.2.2
+            axolotl_extras:
+          - cuda: 121
+            cuda_version: 12.1.0
+            python_version: "3.11"
+            pytorch: 2.3.0
+            axolotl_extras:
+    runs-on: axolotl-gpu-runner
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Docker metadata
+        id: metadata
+        uses: docker/metadata-action@v5
+        with:
+          images: winglian/axolotl
+          tags: |
+            type=raw,value={{ branch }}-{{ date 'YYYYMMDD' }}
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+      - name: Login to Docker Hub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+      # guidance for testing before pushing: https://docs.docker.com/build/ci/github-actions/test-before-push/
+      - name: Build and export to Docker
+        uses: docker/build-push-action@v5
+        with:
+          context: .
+          build-args: |
+            BASE_TAG=${{ github.ref_name }}-base-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}
+            CUDA=${{ matrix.cuda }}
+            PYTORCH_VERSION=${{ matrix.pytorch }}
+            AXOLOTL_ARGS=${{ matrix.axolotl_args }}
+          file: ./docker/Dockerfile
+          push: ${{ github.event_name != 'pull_request' }}
+          tags: |
+            ${{ steps.metadata.outputs.tags }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
+          labels: ${{ steps.metadata.outputs.labels }}
+
+  build-axolotl-cloud:
+    needs: build-axolotl
+    if: ${{ ! contains(github.event.commits[0].message, '[skip docker]]') && github.repository_owner == 'OpenAccess-AI-Collective' }}
+    # this job needs to be run on self-hosted GPU runners...
+    strategy:
+      matrix:
+        include:
+          - cuda: 118
+            cuda_version: 11.8.0
+            python_version: "3.10"
+            pytorch: 2.1.2
+            axolotl_extras:
+            is_latest: true
+          - cuda: 121
+            cuda_version: 12.1.0
+            python_version: "3.10"
+            pytorch: 2.1.2
+            axolotl_extras:
+          - cuda: 121
+            cuda_version: 12.1.0
+            python_version: "3.11"
+            pytorch: 2.2.2
+            axolotl_extras:
+          - cuda: 121
+            cuda_version: 12.1.0
+            python_version: "3.11"
+            pytorch: 2.3.0
+            axolotl_extras:
+    runs-on: axolotl-gpu-runner
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Docker metadata
+        id: metadata
+        uses: docker/metadata-action@v5
+        with:
+          images: winglian/axolotl-cloud
+          tags: |
+            type=raw,value={{ branch }}-{{ date 'YYYYMMDD' }}
+      - name: Login to Docker Hub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ secrets.DOCKERHUB_USERNAME }}
+          password: ${{ secrets.DOCKERHUB_TOKEN }}
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v2
+      - name: Build
+        uses: docker/build-push-action@v5
+        with:
+          context: .
+          build-args: |
+            BASE_TAG=${{ github.ref_name }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
+            CUDA=${{ matrix.cuda }}
+          file: ./docker/Dockerfile-cloud
+          push: ${{ github.event_name != 'pull_request' }}
+          tags: |
+             ${{ steps.metadata.outputs.tags }}-py${{ matrix.python_version }}-cu${{ matrix.cuda }}-${{ matrix.pytorch }}${{ matrix.axolotl_extras != '' && '-' || '' }}${{ matrix.axolotl_extras }}
+          labels: ${{ steps.metadata.outputs.labels }}

.github/workflows/pypi.yml

@@ -25,7 +25,7 @@ jobs:
 
       - name: Install dependencies
         run: |
-          pip3 install wheel
+          pip3 install wheel packaging
           pip3 install -e .
           pip3 install -r requirements-tests.txt
 

.github/workflows/tests.yml

@@ -48,6 +48,8 @@ jobs:
 
       - name: Install dependencies
         run: |
+          pip3 install --upgrade pip
+          pip3 install --upgrade packaging
           pip3 install -U -e .
           pip3 install -r requirements-tests.txt
 
@@ -77,6 +79,16 @@ jobs:
             python_version: "3.10"
             pytorch: 2.1.2
             num_gpus: 1
+          - cuda: 121
+            cuda_version: 12.1.0
+            python_version: "3.11"
+            pytorch: 2.2.2
+            num_gpus: 1
+          - cuda: 121
+            cuda_version: 12.1.0
+            python_version: "3.11"
+            pytorch: 2.3.0
+            num_gpus: 1
     steps:
       - name: Checkout
         uses: actions/checkout@v4

.gitignore

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

README.md

@@ -34,17 +34,18 @@ Features:
   - [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)
-  - [How to Add Custom Prompts](#how-to-add-custom-prompts)
-  - [How to Use Custom Pretokenized Dataset](#how-to-use-your-custom-pretokenized-dataset)
 - [Config](#config)
   - [Train](#train)
   - [Inference](#inference-playground)
   - [Merge LORA to Base](#merge-lora-to-base)
   - [Special Tokens](#special-tokens)
+  - [All Config Options](#all-config-options)
 - Advanced Topics
   - [Multipack](./docs/multipack.qmd)<svg width="24" height="24" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg"><path d="M17 13.5v6H5v-12h6m3-3h6v6m0-6-9 9" class="icon_svg-stroke" stroke="#666" stroke-width="1.5" fill="none" fill-rule="evenodd" stroke-linecap="round" stroke-linejoin="round"></path></svg>
   - [RLHF & DPO](./docs/rlhf.qmd)<svg width="24" height="24" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg"><path d="M17 13.5v6H5v-12h6m3-3h6v6m0-6-9 9" class="icon_svg-stroke" stroke="#666" stroke-width="1.5" fill="none" fill-rule="evenodd" stroke-linecap="round" stroke-linejoin="round"></path></svg>
+  - [Dataset Pre-Processing](./docs/dataset_preprocessing.qmd)<svg width="24" height="24" viewBox="0 0 24 24" xmlns="http://www.w3.org/2000/svg"><path d="M17 13.5v6H5v-12h6m3-3h6v6m0-6-9 9" class="icon_svg-stroke" stroke="#666" stroke-width="1.5" fill="none" fill-rule="evenodd" stroke-linecap="round" stroke-linejoin="round"></path></svg>
 - [Common Errors](#common-errors-)
   - [Tokenization Mismatch b/w Training & Inference](#tokenization-mismatch-bw-inference--training)
 - [Debugging Axolotl](#debugging-axolotl)
@@ -82,6 +83,7 @@ Features:
 | llama       | ✅         | ✅    | ✅     | ✅             | ✅                 | ✅          | ✅            |
 | Mistral     | ✅         | ✅    | ✅     | ✅             | ✅                 | ✅          | ✅            |
 | Mixtral-MoE | ✅         | ✅    | ✅     | ❓             | ❓                 | ❓          | ❓            |
+| Mixtral8X22 | ✅         | ✅    | ✅     | ❓             | ❓                 | ❓          | ❓            |
 | Pythia      | ✅         | ✅    | ✅     | ❌             | ❌                 | ❌          | ❓            |
 | cerebras    | ✅         | ✅    | ✅     | ❌             | ❌                 | ❌          | ❓            |
 | btlm        | ✅         | ✅    | ✅     | ❌             | ❌                 | ❌          | ❓            |
@@ -108,7 +110,7 @@ Get started with Axolotl in just a few steps! This quickstart guide will walk yo
 git clone https://github.com/OpenAccess-AI-Collective/axolotl
 cd axolotl
 
-pip3 install packaging
+pip3 install packaging ninja
 pip3 install -e '.[flash-attn,deepspeed]'
 ```
 
@@ -122,11 +124,11 @@ accelerate launch -m axolotl.cli.train examples/openllama-3b/lora.yml
 
 # inference
 accelerate launch -m axolotl.cli.inference examples/openllama-3b/lora.yml \
-    --lora_model_dir="./lora-out"
+    --lora_model_dir="./outputs/lora-out"
 
 # gradio
 accelerate launch -m axolotl.cli.inference examples/openllama-3b/lora.yml \
-    --lora_model_dir="./lora-out" --gradio
+    --lora_model_dir="./outputs/lora-out" --gradio
 
 # remote yaml files - the yaml config can be hosted on a public URL
 # Note: the yaml config must directly link to the **raw** yaml
@@ -222,23 +224,17 @@ For cloud GPU providers that support docker images, use [`winglian/axolotl-cloud
   python get-pip.py
   ```
 
-  3. Install torch
-  ```bash
-  pip3 install -U torch --index-url https://download.pytorch.org/whl/cu118
-  ```
+  3. Install Pytorch https://pytorch.org/get-started/locally/
 
-  4. Axolotl
-  ```bash
-  git clone https://github.com/OpenAccess-AI-Collective/axolotl
-  cd axolotl
+  4. Follow instructions on quickstart.
 
-  pip3 install packaging
-  pip3 install -e '.[flash-attn,deepspeed]'
+  5. Run
+  ```bash
   pip3 install protobuf==3.20.3
   pip3 install -U --ignore-installed requests Pillow psutil scipy
   ```
 
-  5. Set path
+  6. Set path
   ```bash
   export LD_LIBRARY_PATH=/usr/lib/x86_64-linux-gnu:$LD_LIBRARY_PATH
   ```
@@ -297,188 +293,47 @@ HF_TOKEN=xx sky launch axolotl.yaml --env HF_TOKEN
 HF_TOKEN=xx BUCKET=<unique-name> sky spot launch axolotl-spot.yaml --env HF_TOKEN --env BUCKET
 ```
 
+#### Launching on public clouds via dstack
+To launch on GPU instance (both on-demand and spot instances) on public clouds (GCP, AWS, Azure, Lambda Labs, TensorDock, Vast.ai, and CUDO), you can use [dstack](https://dstack.ai/).
+
+Write a job description in YAML as below:
+
+```yaml
+# dstack.yaml
+type: task
+
+image: winglian/axolotl-cloud:main-20240429-py3.11-cu121-2.2.2
+
+env:
+  - HUGGING_FACE_HUB_TOKEN
+  - WANDB_API_KEY
+
+commands:
+  - accelerate launch -m axolotl.cli.train config.yaml
+
+ports:
+  - 6006
+
+resources:
+  gpu:
+    memory: 24GB..
+    count: 2
+```
+
+then, simply run the job with `dstack run` command. Append `--spot` option if you want spot instance. `dstack run` command will show you the instance with cheapest price across multi cloud services:
+
+```bash
+pip install dstack
+HUGGING_FACE_HUB_TOKEN=xxx WANDB_API_KEY=xxx dstack run . -f dstack.yaml # --spot
+```
+
+For further and fine-grained use cases, please refer to the official [dstack documents](https://dstack.ai/docs/) and the detailed description of [axolotl example](https://github.com/dstackai/dstack/tree/master/examples/fine-tuning/axolotl) on the official repository.
+
 ### Dataset
 
-Axolotl supports a variety of dataset formats. Below are some of the formats you can use.
-Have dataset(s) in one of the following format (JSONL recommended):
+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.
 
-#### Pretraining
-
-- `completion`: raw corpus
-  ```json
-  {"text": "..."}
-  ```
-
-Note: Axolotl usually loads the entire dataset into memory. This will be challenging for large datasets. Use the following config to enable streaming:
-
-```yaml
-pretraining_dataset: # hf path only
-```
-
-#### Supervised finetuning
-
-##### Instruction
-
-- `alpaca`: instruction; input(optional)
-  ```json
-  {"instruction": "...", "input": "...", "output": "..."}
-  ```
-
-<details>
-
-<summary>See other formats</summary>
-
-- `jeopardy`: question and answer
-  ```json
-  {"question": "...", "category": "...", "answer": "..."}
-  ```
-- `oasst`: instruction
-  ```json
-  {"INSTRUCTION": "...", "RESPONSE": "..."}
-  ```
-- `gpteacher`: instruction; input(optional)
-  ```json
-  {"instruction": "...", "input": "...", "response": "..."}
-  ```
-- `reflection`: instruction with reflect; input(optional)
-  ```json
-  {"instruction": "...", "input": "...", "output": "...", "reflection": "...", "corrected": "..."}
-  ```
-- `explainchoice`: question, choices, (solution OR explanation)
-  ```json
-  {"question": "...", "choices": ["..."], "solution": "...", "explanation": "..."}
-  ```
-- `concisechoice`: question, choices, (solution OR explanation)
-  ```json
-  {"question": "...", "choices": ["..."], "solution": "...", "explanation": "..."}
-  ```
-- `summarizetldr`: article and summary
-  ```json
-  {"article": "...", "summary": "..."}
-  ```
-- `alpaca_chat`: basic instruct for alpaca chat
-  ```json
-  {"instruction": "...", "input": "...", "response": "..."}
-  ```
-- `alpaca_chat.load_qa`: question and answer for alpaca chat
-  ```json
-  {"question": "...", "answer": "..."}
-  ```
-- `alpaca_chat.load_concise`: question and answer for alpaca chat, for concise answers
-  ```json
-  {"instruction": "...", "input": "...", "response": "..."}
-  ```
-- `alpaca_chat.load_camel_ai`: question and answer for alpaca chat, for load_camel_ai
-  ```json
-  {"message_1": "...", "message_2": "..."}
-  ```
-- `alpaca_w_system.load_open_orca`: support for open orca datasets with included system prompts, instruct
-  ```json
-  {"system_prompt": "...", "question": "...", "response": "..."}
-  ```
-- `context_qa`: in context question answering from an article
-  ```json
-  {"article": "...", "question": "...", "answer": "..."}
-  ```
-- `context_qa.load_v2`: in context question answering (alternate)
-  ```json
-  {"context": "...", "question": "...", "answer": "..."}
-  ```
-- `context_qa.load_404`: in context question answering from an article, with default response for no answer from context
-  ```json
-  {"article": "...", "unanswerable_question": "..."}
-  ```
-- `creative_acr.load_answer`: instruction and revision
-  ```json
-  {"instruction": "...", "revision": "..."}
-  ```
-- `creative_acr.load_critique`: critique
-  ```json
-  {"scores": "...", "critiques": "...", "instruction": "...", "answer": "..."}
-  ```
-- `creative_acr.load_revise`: critique and revise
-  ```json
-  {"scores": "...", "critiques": "...", "instruction": "...", "answer": "...", "revision": "..."}
-  ```
-- `metharme`: instruction, adds additional eos tokens
-  ```json
-  {"prompt": "...", "generation": "..."}
-  ```
-
-</details>
-
-##### Template-Free
-
-- `input_output`: template-free prompt construction
-  ```json
-   {"segments": [{"label": true|false, "text": "..."}]}
-  ```
-
-This is a special format that allows you to construct prompts without using templates. This is for advanced users who want more freedom with prompt construction.  See [these docs](docs/input_output.qmd) for more details.
-
-##### Conversation
-
-- `sharegpt`: conversations where `from` is `human`/`gpt`. (optional: first row with role `system` to override default system prompt)
-  ```json
-  {"conversations": [{"from": "...", "value": "..."}]}
-  ```
-
-<details>
-
-<summary>See other formats</summary>
-
-- `pygmalion`: pygmalion
-  ```json
-  {"conversations": [{"role": "...", "value": "..."}]}
-  ```
-- `sharegpt.load_role`: conversations where `role` is used instead of `from`
-  ```json
-  {"conversations": [{"role": "...", "value": "..."}]}
-  ```
-- `sharegpt.load_guanaco`: conversations where `from` is `prompter`/`assistant` instead of default sharegpt
-  ```json
-  {"conversations": [{"from": "...", "value": "..."}]}
-  ```
-- `sharegpt_jokes`: creates a chat where bot is asked to tell a joke, then explain why the joke is funny
-  ```json
-  {"conversations": [{"title": "...", "text": "...", "explanation": "..."}]}
-  ```
-
-</details>
-
-Note: `type: sharegpt` opens a special config `conversation:` that enables conversions to many Conversation types. See dataset section under [all yaml options](#all-yaml-options).
-
-#### How to add custom prompts
-
-For a dataset that is preprocessed for instruction purposes:
-
-```json
-{"input": "...", "output": "..."}
-```
-
-You can use this example in your YAML config:
-
-```yaml
-datasets:
-  - path: repo
-    type:
-      system_prompt: ""
-      field_system: system
-      field_instruction: input
-      field_output: output
-      format: "[INST] {instruction} [/INST]"
-      no_input_format: "[INST] {instruction} [/INST]"
-```
-See full config options under [all yaml options](#all-yaml-options).
-
-#### How to use your custom pretokenized dataset
-
-- Do not pass a `type:`
-- Columns in Dataset must be exactly `input_ids`, `attention_mask`, `labels`
-
-```yaml
-- path: ...
-```
+See [these docs](https://openaccess-ai-collective.github.io/axolotl/docs/dataset-formats/) for more information on how to use different dataset formats.
 
 ### Config
 
@@ -563,512 +418,9 @@ See [examples](examples) for quick start. It is recommended to duplicate and mod
     - v_proj
   ```
 
-<details id="all-yaml-options">
+#### All Config Options
 
-<summary>All yaml options (click to expand)</summary>
-
-```yaml
-# This is the huggingface model that contains *.pt, *.safetensors, or *.bin files
-# This can also be a relative path to a model on disk
-base_model: ./llama-7b-hf
-# You can specify an ignore pattern if the model repo contains more than 1 model type (*.pt, etc)
-base_model_ignore_patterns:
-# If the base_model repo on hf hub doesn't include configuration .json files,
-# You can set that here, or leave this empty to default to base_model
-base_model_config: ./llama-7b-hf
-# You can specify to choose a specific model revision from huggingface hub
-revision_of_model:
-# Optional tokenizer configuration path in case you want to use a different tokenizer
-# than the one defined in the base model
-tokenizer_config:
-# If you want to specify the type of model to load, AutoModelForCausalLM is a good choice too
-model_type: AutoModelForCausalLM
-# Corresponding tokenizer for the model AutoTokenizer is a good choice
-tokenizer_type: AutoTokenizer
-# Trust remote code for untrusted source
-trust_remote_code:
-# use_fast option for tokenizer loading from_pretrained, default to True
-tokenizer_use_fast:
-# Whether to use the legacy tokenizer setting, defaults to True
-tokenizer_legacy:
-# Resize the model embeddings when new tokens are added to multiples of 32
-# This is reported to improve training speed on some models
-resize_token_embeddings_to_32x:
-
-# (Internal use only)
-# Used to identify which the model is based on
-is_falcon_derived_model:
-is_llama_derived_model:
-is_qwen_derived_model:
-# Please note that if you set this to true, `padding_side` will be set to "left" by default
-is_mistral_derived_model:
-
-# optional overrides to the base model configuration
-overrides_of_model_config:
-  # RoPE Scaling https://github.com/huggingface/transformers/pull/24653
-  rope_scaling:
-    type: # linear | dynamic
-    factor: # float
-
-# optional overrides to the bnb 4bit quantization configuration
-# https://huggingface.co/docs/transformers/main/main_classes/quantization#transformers.BitsAndBytesConfig
-bnb_config_kwargs:
-  # These are default values
-  llm_int8_has_fp16_weight: false
-  bnb_4bit_quant_type: nf4
-  bnb_4bit_use_double_quant: true
-
-
-# Whether you are training a 4-bit GPTQ quantized model
-gptq: true
-
-# This will attempt to quantize the model down to 8 bits and use adam 8 bit optimizer
-load_in_8bit: true
-# Use bitsandbytes 4 bit
-load_in_4bit:
-
-# Use CUDA bf16
-bf16: true # bool or 'full' for `bf16_full_eval`. require >=ampere
-# Use CUDA fp16
-fp16: true
-# Use CUDA tf32
-tf32: true # require >=ampere
-
-# No AMP (automatic mixed precision)
-bfloat16: true # require >=ampere
-float16: true
-
-# Limit the memory for all available GPUs to this amount (if an integer, expressed in gigabytes); default: unset
-gpu_memory_limit: 20GiB
-# Do the LoRA/PEFT loading on CPU -- this is required if the base model is so large it takes up most or all of the available GPU VRAM, e.g. during a model and LoRA merge
-lora_on_cpu: true
-
-# A list of one or more datasets to finetune the model with
-datasets:
-  # HuggingFace dataset repo | s3://,gs:// path | "json" for local dataset, make sure to fill data_files
-  - path: vicgalle/alpaca-gpt4
-  # The type of prompt to use for training. [alpaca, sharegpt, gpteacher, oasst, reflection]
-    type: alpaca # format | format:<prompt_style> (chat/instruct) | <prompt_strategies>.load_<load_fn>
-    ds_type: # Optional[str] (json|arrow|parquet|text|csv) defines the datatype when path is a file
-    data_files: # Optional[str] path to source data files
-    shards: # Optional[int] number of shards to split data into
-    name: # Optional[str] name of dataset configuration to load
-    train_on_split: train # Optional[str] name of dataset split to load from
-
-    # Optional[str] fastchat conversation type, only used with type: sharegpt
-    conversation: # Options (see Conversation 'name'): https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py
-    field_human: # Optional[str]. Human key to use for conversation.
-    field_model: # Optional[str]. Assistant key to use for conversation.
-    # Add additional keys from your dataset as input or output roles
-    roles:
-      input: # Optional[List[str]]. These will be masked based on train_on_input
-      output: # Optional[List[str]].
-
-  # Custom user instruction prompt
-  - path: repo
-    type:
-      # The below are defaults. only set what's needed if you use a different column name.
-      system_prompt: ""
-      system_format: "{system}"
-      field_system: system
-      field_instruction: instruction
-      field_input: input
-      field_output: output
-
-      # Customizable to be single line or multi-line
-      # Use {instruction}/{input} as key to be replaced
-      # 'format' can include {input}
-      format: |-
-        User: {instruction} {input}
-        Assistant:
-      # 'no_input_format' cannot include {input}
-      no_input_format: "{instruction} "
-
-      # For `completion` datsets only, uses the provided field instead of `text` column
-      field:
-
-# If false, the datasets will not be shuffled and will keep their original order in `datasets`.
-# The same applies to the `test_datasets` option and the `pretraining_dataset` option. Default is true.
-shuffle_merged_datasets: true
-
-# A list of one or more datasets to eval the model with.
-# You can use either test_datasets, or val_set_size, but not both.
-test_datasets:
-  - path: /workspace/data/eval.jsonl
-    ds_type: json
-    # You need to specify a split. For "json" datasets the default split is called "train".
-    split: train
-    type: completion
-    data_files:
-      - /workspace/data/eval.jsonl
-
-# use RL training: 'dpo', 'ipo', 'kto_pair'
-rl:
-
-# Saves the desired chat template to the tokenizer_config.json for easier inferencing
-# Currently supports chatml and inst (mistral/mixtral)
-chat_template: chatml
-# Changes the default system message
-default_system_message: You are a helpful assistant. Please give a long and detailed answer. # Currently only supports chatml.
-# Axolotl attempts to save the dataset as an arrow after packing the data together so
-# subsequent training attempts load faster, relative path
-dataset_prepared_path: data/last_run_prepared
-# Push prepared dataset to hub
-push_dataset_to_hub: # repo path
-# The maximum number of processes to use while preprocessing your input dataset. This defaults to `os.cpu_count()`
-# if not set.
-dataset_processes: # defaults to os.cpu_count() if not set
-# Keep dataset in memory while preprocessing
-# Only needed if cached dataset is taking too much storage
-dataset_keep_in_memory:
-# push checkpoints to hub
-hub_model_id: # private repo path to push finetuned model
-# how to push checkpoints to hub
-# https://huggingface.co/docs/transformers/v4.31.0/en/main_classes/trainer#transformers.TrainingArguments.hub_strategy
-hub_strategy:
-# Whether to use hf `use_auth_token` for loading datasets. Useful for fetching private datasets
-# Required to be true when used in combination with `push_dataset_to_hub`
-hf_use_auth_token: # boolean
-# How much of the dataset to set aside as evaluation. 1 = 100%, 0.50 = 50%, etc. 0 for no eval.
-val_set_size: 0.04
-# Num shards for whole dataset
-dataset_shard_num:
-# Index of shard to use for whole dataset
-dataset_shard_idx:
-
-# The maximum length of an input to train with, this should typically be less than 2048
-# as most models have a token/context limit of 2048
-sequence_len: 2048
-# Pad inputs so each step uses constant sized buffers
-# This will reduce memory fragmentation and may prevent OOMs, by re-using memory more efficiently
-pad_to_sequence_len:
-# Use efficient multi-packing with block diagonal attention and per sequence position_ids. Recommend set to 'true'
-sample_packing:
-# Set to 'false' if getting errors during eval with sample_packing on.
-eval_sample_packing:
-# You can set these packing optimizations AFTER starting a training at least once.
-# The trainer will provide recommended values for these values.
-sample_packing_eff_est:
-total_num_tokens:
-
-# Passed through to transformers when loading the model when launched without accelerate
-# Use `sequential` when training w/ model parallelism to limit memory
-device_map:
-# Defines the max memory usage per gpu on the system. Passed through to transformers when loading the model.
-max_memory:
-
-# If you want to use 'lora' or 'qlora' or leave blank to train all parameters in original model
-adapter: lora
-# If you already have a lora model trained that you want to load, put that here.
-# This means after training, if you want to test the model, you should set this to the value of `output_dir`.
-# Note that if you merge an adapter to the base model, a new subdirectory `merged` will be created under the `output_dir`.
-lora_model_dir:
-
-# LoRA hyperparameters
-# For more details about the following options, see:
-# https://www.anyscale.com/blog/fine-tuning-llms-lora-or-full-parameter-an-in-depth-analysis-with-llama-2
-lora_r: 8
-lora_alpha: 16
-lora_dropout: 0.05
-lora_target_modules:
-  - q_proj
-  - v_proj
-#  - k_proj
-#  - o_proj
-#  - gate_proj
-#  - down_proj
-#  - up_proj
-lora_target_linear: # If true, will target all linear modules
-peft_layers_to_transform: # The layer indices to transform, otherwise, apply to all layers
-
-# If you added new tokens to the tokenizer, you may need to save some LoRA modules because they need to know the new tokens.
-# For LLaMA and Mistral, you need to save `embed_tokens` and `lm_head`. It may vary for other models.
-# `embed_tokens` converts tokens to embeddings, and `lm_head` converts embeddings to token probabilities.
-# https://github.com/huggingface/peft/issues/334#issuecomment-1561727994
-lora_modules_to_save:
-#  - embed_tokens
-#  - lm_head
-
-lora_fan_in_fan_out: false
-
-peft:
-  # Configuration options for loftq initialization for LoRA
-  # https://huggingface.co/docs/peft/developer_guides/quantization#loftq-initialization
-  loftq_config:
-    loftq_bits:  # typically 4 bits
-
-# ReLoRA configuration
-# Must use either 'lora' or 'qlora' adapter, and does not support fsdp or deepspeed
-relora_steps: # Number of steps per ReLoRA restart
-relora_warmup_steps: # Number of per-restart warmup steps
-relora_anneal_steps: # Number of anneal steps for each relora cycle
-relora_prune_ratio: # threshold for optimizer magnitude when pruning
-relora_cpu_offload: # True to perform lora weight merges on cpu during restarts, for modest gpu memory savings
-
-# wandb configuration if you're using it
-# Make sure your `WANDB_API_KEY` environment variable is set (recommended) or you login to wandb with `wandb login`.
-wandb_mode: # "offline" to save run metadata locally and not sync to the server, "disabled" to turn off wandb
-wandb_project: # Your wandb project name
-wandb_entity: # A wandb Team name if using a Team
-wandb_watch:
-wandb_name: # Set the name of your wandb run
-wandb_run_id: # Set the ID of your wandb run
-wandb_log_model: # "checkpoint" to log model to wandb Artifacts every `save_steps` or "end" to log only at the end of training
-
-# mlflow configuration if you're using it
-mlflow_tracking_uri: # URI to mlflow
-mlflow_experiment_name: # Your experiment name
-hf_mlflow_log_artifacts:  # set to true to copy each saved checkpoint on each save to mlflow artifact registry
-
-# 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
-torch_compile_backend:  # Optional[str]
-
-# Training hyperparameters
-
-# If greater than 1, backpropagation will be skipped and the gradients will be accumulated for the given number of steps.
-gradient_accumulation_steps: 1
-# The number of samples to include in each batch. This is the number of samples sent to each GPU.
-micro_batch_size: 2
-eval_batch_size:
-num_epochs: 4
-warmup_steps: 100  # cannot use with warmup_ratio
-warmup_ratio: 0.05  # cannot use with warmup_steps
-learning_rate: 0.00003
-lr_quadratic_warmup:
-logging_steps:
-eval_steps: # Leave empty to eval at each epoch, integers for every N steps. decimal for fraction of total steps
-evals_per_epoch: # number of times per epoch to run evals, mutually exclusive with eval_steps
-save_strategy: # Set to `no` to skip checkpoint saves
-save_steps: # Leave empty to save at each epoch
-saves_per_epoch: # number of times per epoch to save a checkpoint, mutually exclusive with save_steps
-save_total_limit: # Checkpoints saved at a time
-# Maximum number of iterations to train for. It precedes num_epochs which means that
-# if both are set, num_epochs will not be guaranteed.
-# e.g., when 1 epoch is 1000 steps => `num_epochs: 2` and `max_steps: 100` will train for 100 steps
-max_steps:
-
-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]
-
-loss_watchdog_threshold: # High loss value, indicating the learning has broken down (a good estimate is ~2 times the loss at the start of training)
-loss_watchdog_patience: # Number of high-loss steps in a row before the trainer aborts (default: 3)
-
-# Save model as safetensors (require safetensors package)
-save_safetensors:
-
-# Whether to mask out or include the human's prompt from the training labels
-train_on_inputs: false
-# Group similarly sized data to minimize padding.
-# May be slower to start, as it must download and sort the entire dataset.
-# Note that training loss may have an oscillating pattern with this enabled.
-group_by_length: false
-
-# Whether to use gradient checkpointing https://huggingface.co/docs/transformers/v4.18.0/en/performance#gradient-checkpointing
-gradient_checkpointing: false
-# additional kwargs to pass to the trainer for gradient checkpointing
-# gradient_checkpointing_kwargs:
-#   use_reentrant: true
-
-# Stop training after this many evaluation losses have increased in a row
-# https://huggingface.co/transformers/v4.2.2/_modules/transformers/trainer_callback.html#EarlyStoppingCallback
-early_stopping_patience: 3
-
-# Specify a scheduler and kwargs to use with the optimizer
-lr_scheduler: # 'one_cycle' | 'log_sweep' | empty for cosine
-lr_scheduler_kwargs:
-cosine_min_lr_ratio: # decay lr to some percentage of the peak lr, e.g. cosine_min_lr_ratio=0.1 for 10% of peak lr
-cosine_constant_lr_ratio: # freeze lr at some percentage of the step, e.g. cosine_constant_lr_ratio=0.8 means start cosine_min_lr at 80% of training step (https://arxiv.org/pdf/2308.04014.pdf)
-
-# For one_cycle optim
-lr_div_factor: # Learning rate div factor
-
-# Specify optimizer
-# Valid values are driven by the Transformers OptimizerNames class, see:
-# https://github.com/huggingface/transformers/blob/95b374952dc27d8511541d6f5a4e22c9ec11fb24/src/transformers/training_args.py#L134
-#
-# Note that not all optimizers may be available in your environment, ex: 'adamw_anyprecision' is part of
-# torchdistx, 'adamw_bnb_8bit' is part of bnb.optim.Adam8bit, etc. When in doubt, it is recommended to start with the optimizer used
-# in the examples/ for your model and fine-tuning use case.
-#
-# Valid values for 'optimizer' include:
-# - adamw_hf
-# - adamw_torch
-# - adamw_torch_fused
-# - adamw_torch_xla
-# - adamw_apex_fused
-# - adafactor
-# - adamw_anyprecision
-# - sgd
-# - adagrad
-# - adamw_bnb_8bit
-# - lion_8bit
-# - lion_32bit
-# - paged_adamw_32bit
-# - paged_adamw_8bit
-# - paged_lion_32bit
-# - paged_lion_8bit
-# - galore_adamw
-# - galore_adamw_8bit
-# - galore_adafactor
-# - galore_adamw_layerwise
-# - galore_adamw_8bit_layerwise
-# - galore_adafactor_layerwise
-optimizer:
-# Dictionary of arguments to pass to the optimizer
-optim_args:
-# For Galore Optimizers the following optim_args are available
-# rank:  # type: int
-# update_proj_gap  # type: int
-# scale  # type: float
-# proj_type:  # type: str, default = std
-
-# The target modules to optimize, i.e. the module names that you would like to train, right now this is used only for GaLore algorithm
-optim_target_modules:
-# - self_attn  # for llama
-# - mlp
-
-# Specify weight decay
-weight_decay:
-# adamw hyperparams
-adam_beta1:
-adam_beta2:
-adam_epsilon:
-# Gradient clipping max norm
-max_grad_norm:
-
-# Augmentation techniques
-# NEFT https://arxiv.org/abs/2310.05914, set this to a number (paper default is 5) to add noise to embeddings
-# currently only supported on Llama and Mistral
-neftune_noise_alpha:
-
-# Whether to bettertransformers
-flash_optimum:
-# Whether to use xformers attention patch https://github.com/facebookresearch/xformers:
-xformers_attention:
-# Whether to use flash attention patch https://github.com/Dao-AILab/flash-attention:
-flash_attention:
-flash_attn_cross_entropy:  # Whether to use flash-attention cross entropy implementation - advanced use only
-flash_attn_rms_norm:  # Whether to use flash-attention rms norm implementation - advanced use only
-flash_attn_fuse_qkv: # Whether to fuse QKV into a single operation
-flash_attn_fuse_mlp: # Whether to fuse part of the MLP into a single operation
-# Whether to use scaled-dot-product attention
-# https://pytorch.org/docs/stable/generated/torch.nn.functional.scaled_dot_product_attention.html
-sdp_attention:
-# Shifted-sparse attention (only llama) - https://arxiv.org/pdf/2309.12307.pdf
-s2_attention:
-# Resume from a specific checkpoint dir
-resume_from_checkpoint:
-# If resume_from_checkpoint isn't set and you simply want it to start where it left off.
-# Be careful with this being turned on between different models.
-auto_resume_from_checkpoints: false
-
-# Don't mess with this, it's here for accelerate and torchrun
-local_rank:
-
-# Add or change special tokens.
-# If you add tokens here, you don't need to add them to the `tokens` list.
-special_tokens:
-  # bos_token: "<s>"
-  # eos_token: "</s>"
-  # unk_token: "<unk>"
-
-# Add extra tokens.
-tokens:
-
-# FSDP
-fsdp:
-fsdp_config:
-
-# Deepspeed config path. e.g., deepspeed_configs/zero3.json
-deepspeed:
-
-# Advanced DDP Arguments
-ddp_timeout:
-ddp_bucket_cap_mb:
-ddp_broadcast_buffers:
-
-# Path to torch distx for optim 'adamw_anyprecision'
-torchdistx_path:
-
-# Set to HF dataset for type: 'completion' for streaming instead of pre-tokenize
-pretraining_dataset:
-
-# Debug mode
-debug:
-
-# Seed
-seed:
-
-# Allow overwrite yml config using from cli
-strict:
-```
-
-</details>
-
-<details>
-<summary> Understanding of batch size and gradient accumulation steps </summary>
-<br/>
-Gradient accumulation means accumulating gradients over several mini-batches and updating the model weights afterward. When the samples in each batch are diverse, this technique doesn't significantly impact learning.
-
-This method allows for effective training with larger effective batch sizes without needing proportionally larger memory. Here's why:
-
-1. **Memory Consumption with Batch Size**: The primary reason increasing the batch size impacts memory is due to the storage requirements for intermediate activations. When you forward propagate a batch through a network, you have to store the activations at each layer for each sample in the batch, because these activations are used during backpropagation to compute gradients. Therefore, larger batches mean more activations, leading to greater GPU memory consumption.
-
-2. **Gradient Accumulation**: With gradient accumulation, you're effectively simulating a larger batch size by accumulating gradients over several smaller batches (or micro-batches). However, at any given time, you're only forward and backward propagating a micro-batch. This means you only store activations for the micro-batch, not the full accumulated batch. As a result, you can simulate the effect of a larger batch size without the memory cost of storing activations for a large batch.
-
-**Example 1:**
-Micro batch size: 3
-Gradient accumulation steps: 2
-Number of GPUs: 3
-Total batch size = 3 * 2 * 3 = 18
-
-```
-| GPU 1          | GPU 2          | GPU 3          |
-|----------------|----------------|----------------|
-| S1, S2, S3     | S4, S5, S6     | S7, S8, S9     |
-| e1, e2, e3     | e4, e5, e6     | e7, e8, e9     |
-|----------------|----------------|----------------|
-| → (accumulate) | → (accumulate) | → (accumulate) |
-|----------------|----------------|----------------|
-| S10, S11, S12  | S13, S14, S15  | S16, S17, S18  |
-| e10, e11, e12  | e13, e14, e15  | e16, e17, e18  |
-|----------------|----------------|----------------|
-| → (apply)      | → (apply)      | → (apply)      |
-
-Accumulated gradient for the weight w1 after the second iteration (considering all GPUs):
-Total gradient for w1 = e1 + e2 + e3 + e4 + e5 + e6 + e7 + e8 + e9 + e10 + e11 + e12 + e13 + e14 + e15 + e16 + e17 + e18
-
-Weight update for w1:
-w1_new = w1_old - learning rate x (Total gradient for w1 / 18)
-```
-
-**Example 2:**
-Micro batch size: 2
-Gradient accumulation steps: 1
-Number of GPUs: 3
-Total batch size = 2 * 1 * 3 = 6
-
-```
-| GPU 1     | GPU 2     | GPU 3     |
-|-----------|-----------|-----------|
-| S1, S2    | S3, S4    | S5, S6    |
-| e1, e2    | e3, e4    | e5, e6    |
-|-----------|-----------|-----------|
-| → (apply) | → (apply) | → (apply) |
-
-Accumulated gradient for the weight w1 (considering all GPUs):
-Total gradient for w1 = e1 + e2 + e3 + e4 + e5 + e6
-
-Weight update for w1:
-w1_new = w1_old - learning rate × (Total gradient for w1 / 6)
-```
-
-</details>
+See [these docs](docs/config.qmd) for all config options.
 
 ### Train
 
@@ -1112,7 +464,7 @@ deepspeed: deepspeed_configs/zero1.json
 ```
 
 ```shell
-accelerate launch -m axolotl.cli.train examples/llama-2/config.py --deepspeed deepspeed_configs/zero1.json
+accelerate launch -m axolotl.cli.train examples/llama-2/config.yml --deepspeed deepspeed_configs/zero1.json
 ```
 
 ##### FSDP
@@ -1257,7 +609,7 @@ If you decode a prompt constructed by axolotl, you might see spaces between toke
 3. Make sure the inference string from #2 looks **exactly** like the data you fine tuned on from #1, including spaces and new lines.  If they aren't the same, adjust your inference server accordingly.
 4. As an additional troubleshooting step, you can look at the token ids between 1 and 2 to make sure they are identical.
 
-Having misalignment between your prompts during training and inference can cause models to perform very poorly, so it is worth checking this.  See [this blog post](https://hamel.dev/notes/llm/05_tokenizer_gotchas.html) for a concrete example.
+Having misalignment between your prompts during training and inference can cause models to perform very poorly, so it is worth checking this.  See [this blog post](https://hamel.dev/notes/llm/finetuning/05_tokenizer_gotchas.html) for a concrete example.
 
 ## Debugging Axolotl
 
@@ -1299,14 +651,8 @@ Bugs? Please check the [open issues](https://github.com/OpenAccess-AI-Collective
 
 PRs are **greatly welcome**!
 
-Please run below to setup env
+Please run the quickstart instructions followed by the below to setup env:
 ```bash
-git clone https://github.com/OpenAccess-AI-Collective/axolotl
-cd axolotl
-
-pip3 install packaging
-pip3 install -e '.[flash-attn,deepspeed]'
-
 pip3 install -r requirements-dev.txt -r requirements-tests.txt
 pre-commit install
 

_quarto.yml

@@ -30,20 +30,20 @@ website:
           # TODO Edit folder structure after we have more docs.
             - docs/debugging.qmd
             - docs/multipack.qmd
-            - docs/fdsp_qlora.qmd
+            - docs/fsdp_qlora.qmd
             - docs/input_output.qmd
             - docs/rlhf.qmd
             - docs/nccl.qmd
             - docs/mac.qmd
             - docs/multi-node.qmd
+        - section: "Dataset Formats"
+          contents: docs/dataset-formats/*
         - section: "Reference"
           contents:
             - docs/config.qmd
         - docs/faq.qmd
 
 
-
-
 format:
   html:
     theme: materia

cicd/Dockerfile.jinja

@@ -22,6 +22,7 @@ RUN git fetch origin +$GITHUB_REF && \
     git checkout FETCH_HEAD
 
 # If AXOLOTL_EXTRAS is set, append it in brackets
+RUN pip install causal_conv1d
 RUN if [ "$AXOLOTL_EXTRAS" != "" ] ; then \
         pip install -e .[deepspeed,flash-attn,mamba-ssm,galore,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
     else \

cicd/cicd.sh

@@ -1,4 +1,5 @@
 #!/bin/bash
+set -e
 
 pytest --ignore=tests/e2e/ /workspace/axolotl/tests/
 pytest /workspace/axolotl/tests/e2e/patched/

deepspeed_configs/zero3_bf16_cpuoffload_all.json

@@ -0,0 +1,41 @@
+{
+  "zero_force_ds_cpu_optimizer": false,
+  "zero_allow_untested_optimizer": true,
+  "zero_optimization": {
+    "stage": 3,
+    "offload_optimizer": {
+      "device": "cpu",
+      "pin_memory": true
+    },
+    "offload_param": {
+      "device": "cpu",
+      "pin_memory": true
+    },
+    "overlap_comm": true,
+    "contiguous_gradients": true,
+    "sub_group_size": 0,
+    "reduce_bucket_size": "auto",
+    "stage3_prefetch_bucket_size": "auto",
+    "stage3_param_persistence_threshold": "auto",
+    "stage3_max_live_parameters": 0,
+    "stage3_max_reuse_distance": 0,
+    "stage3_gather_16bit_weights_on_model_save": true
+  },
+  "bf16": {
+    "enabled": true
+  },
+  "fp16": {
+    "enabled": "auto",
+    "auto_cast": false,
+    "loss_scale": 0,
+    "initial_scale_power": 32,
+    "loss_scale_window": 1000,
+    "hysteresis": 2,
+    "min_loss_scale": 1
+  },
+  "gradient_accumulation_steps": "auto",
+  "gradient_clipping": "auto",
+  "train_batch_size": "auto",
+  "train_micro_batch_size_per_gpu": "auto",
+  "wall_clock_breakdown": false
+}

deepspeed_configs/zero3_bf16_cpuoffload_params.json

@@ -0,0 +1,37 @@
+{
+  "zero_force_ds_cpu_optimizer": false,
+  "zero_allow_untested_optimizer": true,
+  "zero_optimization": {
+    "stage": 3,
+    "offload_param": {
+      "device": "cpu",
+      "pin_memory": true
+    },
+    "overlap_comm": true,
+    "contiguous_gradients": true,
+    "sub_group_size": 0,
+    "reduce_bucket_size": "auto",
+    "stage3_prefetch_bucket_size": "auto",
+    "stage3_param_persistence_threshold": "auto",
+    "stage3_max_live_parameters": 0,
+    "stage3_max_reuse_distance": 0,
+    "stage3_gather_16bit_weights_on_model_save": true
+  },
+  "bf16": {
+    "enabled": true
+  },
+  "fp16": {
+    "enabled": "auto",
+    "auto_cast": false,
+    "loss_scale": 0,
+    "initial_scale_power": 32,
+    "loss_scale_window": 1000,
+    "hysteresis": 2,
+    "min_loss_scale": 1
+  },
+  "gradient_accumulation_steps": "auto",
+  "gradient_clipping": "auto",
+  "train_batch_size": "auto",
+  "train_micro_batch_size_per_gpu": "auto",
+  "wall_clock_breakdown": false
+}

docker/Dockerfile

@@ -11,7 +11,7 @@ ARG PYTORCH_VERSION="2.1.2"
 ENV PYTORCH_VERSION=$PYTORCH_VERSION
 
 RUN apt-get update && \
-    apt-get install -y --allow-change-held-packages vim curl nano libnccl2 libnccl-dev
+    apt-get install -y --allow-change-held-packages vim curl nano libnccl2 libnccl-dev rsync s3fs
 
 WORKDIR /workspace
 
@@ -20,6 +20,7 @@ RUN git clone --depth=1 https://github.com/OpenAccess-AI-Collective/axolotl.git
 WORKDIR /workspace/axolotl
 
 # If AXOLOTL_EXTRAS is set, append it in brackets
+RUN pip install causal_conv1d
 RUN if [ "$AXOLOTL_EXTRAS" != "" ] ; then \
         pip install -e .[deepspeed,flash-attn,mamba-ssm,galore,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
     else \

docker/Dockerfile-cloud-no-tmux

@@ -0,0 +1,27 @@
+ARG BASE_TAG=main
+FROM winglian/axolotl:$BASE_TAG
+
+ENV HF_DATASETS_CACHE="/workspace/data/huggingface-cache/datasets"
+ENV HUGGINGFACE_HUB_CACHE="/workspace/data/huggingface-cache/hub"
+ENV TRANSFORMERS_CACHE="/workspace/data/huggingface-cache/hub"
+ENV HF_HOME="/workspace/data/huggingface-cache/hub"
+ENV HF_HUB_ENABLE_HF_TRANSFER="1"
+
+EXPOSE 8888
+EXPOSE 22
+
+COPY scripts/cloud-entrypoint-term.sh /root/cloud-entrypoint.sh
+COPY scripts/motd /etc/motd
+
+RUN pip install jupyterlab notebook ipywidgets && \
+    jupyter lab clean
+RUN apt install --yes --no-install-recommends openssh-server tmux sudo && \
+    pip3 install -U --no-cache-dir grpcio ray[default]==2.9.3 && \
+    mkdir -p ~/.ssh && \
+    chmod 700 ~/.ssh && \
+    printf "[ ! -z \"\$TERM\" -a -r /etc/motd ] && cat /etc/motd\n" >> ~/.bashrc && \
+    chmod +x /workspace/axolotl/scripts/cloud-entrypoint.sh && \
+    chmod +x /root/cloud-entrypoint.sh
+
+ENTRYPOINT ["/root/cloud-entrypoint.sh"]
+CMD ["sleep", "infinity"]

docs/batch_vs_grad.qmd

@@ -0,0 +1,59 @@
+---
+title: Batch size vs Gradient accumulation
+description: Understanding of batch size and gradient accumulation steps
+---
+
+Gradient accumulation means accumulating gradients over several mini-batches and updating the model weights afterward. When the samples in each batch are diverse, this technique doesn't significantly impact learning.
+
+This method allows for effective training with larger effective batch sizes without needing proportionally larger memory. Here's why:
+
+1. **Memory Consumption with Batch Size**: The primary reason increasing the batch size impacts memory is due to the storage requirements for intermediate activations. When you forward propagate a batch through a network, you have to store the activations at each layer for each sample in the batch, because these activations are used during backpropagation to compute gradients. Therefore, larger batches mean more activations, leading to greater GPU memory consumption.
+
+2. **Gradient Accumulation**: With gradient accumulation, you're effectively simulating a larger batch size by accumulating gradients over several smaller batches (or micro-batches). However, at any given time, you're only forward and backward propagating a micro-batch. This means you only store activations for the micro-batch, not the full accumulated batch. As a result, you can simulate the effect of a larger batch size without the memory cost of storing activations for a large batch.
+
+**Example 1:**
+Micro batch size: 3
+Gradient accumulation steps: 2
+Number of GPUs: 3
+Total batch size = 3 * 2 * 3 = 18
+
+```
+| GPU 1          | GPU 2          | GPU 3          |
+|----------------|----------------|----------------|
+| S1, S2, S3     | S4, S5, S6     | S7, S8, S9     |
+| e1, e2, e3     | e4, e5, e6     | e7, e8, e9     |
+|----------------|----------------|----------------|
+| → (accumulate) | → (accumulate) | → (accumulate) |
+|----------------|----------------|----------------|
+| S10, S11, S12  | S13, S14, S15  | S16, S17, S18  |
+| e10, e11, e12  | e13, e14, e15  | e16, e17, e18  |
+|----------------|----------------|----------------|
+| → (apply)      | → (apply)      | → (apply)      |
+
+Accumulated gradient for the weight w1 after the second iteration (considering all GPUs):
+Total gradient for w1 = e1 + e2 + e3 + e4 + e5 + e6 + e7 + e8 + e9 + e10 + e11 + e12 + e13 + e14 + e15 + e16 + e17 + e18
+
+Weight update for w1:
+w1_new = w1_old - learning rate x (Total gradient for w1 / 18)
+```
+
+**Example 2:**
+Micro batch size: 2
+Gradient accumulation steps: 1
+Number of GPUs: 3
+Total batch size = 2 * 1 * 3 = 6
+
+```
+| GPU 1     | GPU 2     | GPU 3     |
+|-----------|-----------|-----------|
+| S1, S2    | S3, S4    | S5, S6    |
+| e1, e2    | e3, e4    | e5, e6    |
+|-----------|-----------|-----------|
+| → (apply) | → (apply) | → (apply) |
+
+Accumulated gradient for the weight w1 (considering all GPUs):
+Total gradient for w1 = e1 + e2 + e3 + e4 + e5 + e6
+
+Weight update for w1:
+w1_new = w1_old - learning rate × (Total gradient for w1 / 6)
+```

docs/config.qmd

@@ -3,15 +3,456 @@ title: Config options
 description: A complete list of all configuration options.
 ---
 
-```{python}
-#|echo: false
-#|output: asis
-import re
-# Regex pattern to match the YAML block including its code fence
-pattern = r'<details[^>]*id="all-yaml-options"[^>]*>.*?<summary>All yaml options.*?```yaml(.*?)```.*?</details>'
+```yaml
+# This is the huggingface model that contains *.pt, *.safetensors, or *.bin files
+# This can also be a relative path to a model on disk
+base_model: ./llama-7b-hf
+# You can specify an ignore pattern if the model repo contains more than 1 model type (*.pt, etc)
+base_model_ignore_patterns:
+# If the base_model repo on hf hub doesn't include configuration .json files,
+# You can set that here, or leave this empty to default to base_model
+base_model_config: ./llama-7b-hf
+# You can specify to choose a specific model revision from huggingface hub
+revision_of_model:
+# Optional tokenizer configuration path in case you want to use a different tokenizer
+# than the one defined in the base model
+tokenizer_config:
+# If you want to specify the type of model to load, AutoModelForCausalLM is a good choice too
+model_type: AutoModelForCausalLM
+# Corresponding tokenizer for the model AutoTokenizer is a good choice
+tokenizer_type: AutoTokenizer
+# Trust remote code for untrusted source
+trust_remote_code:
+# use_fast option for tokenizer loading from_pretrained, default to True
+tokenizer_use_fast:
+# Whether to use the legacy tokenizer setting, defaults to True
+tokenizer_legacy:
+# Resize the model embeddings when new tokens are added to multiples of 32
+# This is reported to improve training speed on some models
+resize_token_embeddings_to_32x:
 
-with open('../README.md', 'r') as f:
-    doc = f.read()
-match = re.search(pattern, doc, re.DOTALL)
-print("```yaml", match.group(1).strip(), "```", sep="\n")
+# (Internal use only)
+# Used to identify which the model is based on
+is_falcon_derived_model:
+is_llama_derived_model:
+is_qwen_derived_model:
+# Please note that if you set this to true, `padding_side` will be set to "left" by default
+is_mistral_derived_model:
+
+# optional overrides to the base model configuration
+overrides_of_model_config:
+  # RoPE Scaling https://github.com/huggingface/transformers/pull/24653
+  rope_scaling:
+    type: # linear | dynamic
+    factor: # float
+
+# optional overrides to the bnb 4bit quantization configuration
+# https://huggingface.co/docs/transformers/main/main_classes/quantization#transformers.BitsAndBytesConfig
+bnb_config_kwargs:
+  # These are default values
+  llm_int8_has_fp16_weight: false
+  bnb_4bit_quant_type: nf4
+  bnb_4bit_use_double_quant: true
+
+
+# Whether you are training a 4-bit GPTQ quantized model
+gptq: true
+
+# This will attempt to quantize the model down to 8 bits and use adam 8 bit optimizer
+load_in_8bit: true
+# Use bitsandbytes 4 bit
+load_in_4bit:
+
+# Use CUDA bf16
+bf16: true # bool or 'full' for `bf16_full_eval`. require >=ampere
+# Use CUDA fp16
+fp16: true
+# Use CUDA tf32
+tf32: true # require >=ampere
+
+# No AMP (automatic mixed precision)
+bfloat16: true # require >=ampere
+float16: true
+
+# Limit the memory for all available GPUs to this amount (if an integer, expressed in gigabytes); default: unset
+gpu_memory_limit: 20GiB
+# Do the LoRA/PEFT loading on CPU -- this is required if the base model is so large it takes up most or all of the available GPU VRAM, e.g. during a model and LoRA merge
+lora_on_cpu: true
+
+# A list of one or more datasets to finetune the model with
+datasets:
+  # HuggingFace dataset repo | s3://,gs:// path | "json" for local dataset, make sure to fill data_files
+  - path: vicgalle/alpaca-gpt4
+  # The type of prompt to use for training. [alpaca, sharegpt, gpteacher, oasst, reflection]
+    type: alpaca # format | format:<prompt_style> (chat/instruct) | <prompt_strategies>.load_<load_fn>
+    ds_type: # Optional[str] (json|arrow|parquet|text|csv) defines the datatype when path is a file
+    data_files: # Optional[str] path to source data files
+    shards: # Optional[int] number of shards to split data into
+    name: # Optional[str] name of dataset configuration to load
+    train_on_split: train # Optional[str] name of dataset split to load from
+
+    # Optional[str] fastchat conversation type, only used with type: sharegpt
+    conversation: # Options (see Conversation 'name'): https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py
+    field_human: # Optional[str]. Human key to use for conversation.
+    field_model: # Optional[str]. Assistant key to use for conversation.
+    # Add additional keys from your dataset as input or output roles
+    roles:
+      input: # Optional[List[str]]. These will be masked based on train_on_input
+      output: # Optional[List[str]].
+
+  # Custom user instruction prompt
+  - path: repo
+    type:
+      # The below are defaults. only set what's needed if you use a different column name.
+      system_prompt: ""
+      system_format: "{system}"
+      field_system: system
+      field_instruction: instruction
+      field_input: input
+      field_output: output
+
+      # Customizable to be single line or multi-line
+      # Use {instruction}/{input} as key to be replaced
+      # 'format' can include {input}
+      format: |-
+        User: {instruction} {input}
+        Assistant:
+      # 'no_input_format' cannot include {input}
+      no_input_format: "{instruction} "
+
+      # For `completion` datsets only, uses the provided field instead of `text` column
+      field:
+
+# If false, the datasets will not be shuffled and will keep their original order in `datasets`.
+# The same applies to the `test_datasets` option and the `pretraining_dataset` option. Default is true.
+shuffle_merged_datasets: true
+
+# A list of one or more datasets to eval the model with.
+# You can use either test_datasets, or val_set_size, but not both.
+test_datasets:
+  - path: /workspace/data/eval.jsonl
+    ds_type: json
+    # You need to specify a split. For "json" datasets the default split is called "train".
+    split: train
+    type: completion
+    data_files:
+      - /workspace/data/eval.jsonl
+
+# use RL training: 'dpo', 'ipo', 'kto_pair'
+rl:
+
+# Saves the desired chat template to the tokenizer_config.json for easier inferencing
+# Currently supports chatml and inst (mistral/mixtral)
+chat_template: chatml
+# Changes the default system message
+default_system_message: You are a helpful assistant. Please give a long and detailed answer. # Currently only supports chatml.
+# Axolotl attempts to save the dataset as an arrow after packing the data together so
+# subsequent training attempts load faster, relative path
+dataset_prepared_path: data/last_run_prepared
+# Push prepared dataset to hub
+push_dataset_to_hub: # repo path
+# The maximum number of processes to use while preprocessing your input dataset. This defaults to `os.cpu_count()`
+# if not set.
+dataset_processes: # defaults to os.cpu_count() if not set
+# Keep dataset in memory while preprocessing
+# Only needed if cached dataset is taking too much storage
+dataset_keep_in_memory:
+# push checkpoints to hub
+hub_model_id: # private repo path to push finetuned model
+# how to push checkpoints to hub
+# https://huggingface.co/docs/transformers/v4.31.0/en/main_classes/trainer#transformers.TrainingArguments.hub_strategy
+hub_strategy:
+# Whether to use hf `use_auth_token` for loading datasets. Useful for fetching private datasets
+# Required to be true when used in combination with `push_dataset_to_hub`
+hf_use_auth_token: # boolean
+# How much of the dataset to set aside as evaluation. 1 = 100%, 0.50 = 50%, etc. 0 for no eval.
+val_set_size: 0.04
+# Num shards for whole dataset
+dataset_shard_num:
+# Index of shard to use for whole dataset
+dataset_shard_idx:
+
+# The maximum length of an input to train with, this should typically be less than 2048
+# as most models have a token/context limit of 2048
+sequence_len: 2048
+# Pad inputs so each step uses constant sized buffers
+# This will reduce memory fragmentation and may prevent OOMs, by re-using memory more efficiently
+pad_to_sequence_len:
+# Use efficient multi-packing with block diagonal attention and per sequence position_ids. Recommend set to 'true'
+sample_packing:
+# Set to 'false' if getting errors during eval with sample_packing on.
+eval_sample_packing:
+# You can set these packing optimizations AFTER starting a training at least once.
+# The trainer will provide recommended values for these values.
+sample_packing_eff_est:
+total_num_tokens:
+# Increasing the following values helps with packing, but usually only slightly (<%1.)
+# The number of samples packed at a time.
+sample_packing_group_size: 100000
+# The number of samples which can be packed into one sequence. Increase if using a large sequence_len with many short samples.
+sample_packing_bin_size: 200
+
+# Passed through to transformers when loading the model when launched without accelerate
+# Use `sequential` when training w/ model parallelism to limit memory
+device_map:
+# Defines the max memory usage per gpu on the system. Passed through to transformers when loading the model.
+max_memory:
+
+# If you want to use 'lora' or 'qlora' or leave blank to train all parameters in original model
+adapter: lora
+# If you already have a lora model trained that you want to load, put that here.
+# This means after training, if you want to test the model, you should set this to the value of `output_dir`.
+# Note that if you merge an adapter to the base model, a new subdirectory `merged` will be created under the `output_dir`.
+lora_model_dir:
+
+# LoRA hyperparameters
+# For more details about the following options, see:
+# https://www.anyscale.com/blog/fine-tuning-llms-lora-or-full-parameter-an-in-depth-analysis-with-llama-2
+lora_r: 8
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_modules:
+  - q_proj
+  - v_proj
+#  - k_proj
+#  - o_proj
+#  - gate_proj
+#  - down_proj
+#  - up_proj
+lora_target_linear: # If true, will target all linear modules
+peft_layers_to_transform: # The layer indices to transform, otherwise, apply to all layers
+
+# If you added new tokens to the tokenizer, you may need to save some LoRA modules because they need to know the new tokens.
+# For LLaMA and Mistral, you need to save `embed_tokens` and `lm_head`. It may vary for other models.
+# `embed_tokens` converts tokens to embeddings, and `lm_head` converts embeddings to token probabilities.
+# https://github.com/huggingface/peft/issues/334#issuecomment-1561727994
+lora_modules_to_save:
+#  - embed_tokens
+#  - lm_head
+
+lora_fan_in_fan_out: false
+
+# LoRA+ hyperparameters
+# For more details about the following options, see:
+# https://arxiv.org/abs/2402.12354  and `src/axolotl/core/train_builder.py`
+loraplus_lr_ratio: # loraplus learning rate ratio lr_B / lr_A. Recommended value is 2^4.
+loraplus_lr_embedding: #  loraplus learning rate for lora embedding layers. Default value is 1e-6.
+
+peft:
+  # Configuration options for loftq initialization for LoRA
+  # https://huggingface.co/docs/peft/developer_guides/quantization#loftq-initialization
+  loftq_config:
+    loftq_bits:  # typically 4 bits
+
+# ReLoRA configuration
+# Must use either 'lora' or 'qlora' adapter, and does not support fsdp or deepspeed
+relora_steps: # Number of steps per ReLoRA restart
+relora_warmup_steps: # Number of per-restart warmup steps
+relora_anneal_steps: # Number of anneal steps for each relora cycle
+relora_prune_ratio: # threshold for optimizer magnitude when pruning
+relora_cpu_offload: # True to perform lora weight merges on cpu during restarts, for modest gpu memory savings
+
+# wandb configuration if you're using it
+# Make sure your `WANDB_API_KEY` environment variable is set (recommended) or you login to wandb with `wandb login`.
+wandb_mode: # "offline" to save run metadata locally and not sync to the server, "disabled" to turn off wandb
+wandb_project: # Your wandb project name
+wandb_entity: # A wandb Team name if using a Team
+wandb_watch:
+wandb_name: # Set the name of your wandb run
+wandb_run_id: # Set the ID of your wandb run
+wandb_log_model: # "checkpoint" to log model to wandb Artifacts every `save_steps` or "end" to log only at the end of training
+
+# mlflow configuration if you're using it
+mlflow_tracking_uri: # URI to mlflow
+mlflow_experiment_name: # Your experiment name
+hf_mlflow_log_artifacts:  # set to true to copy each saved checkpoint on each save to mlflow artifact registry
+
+# 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
+torch_compile_backend:  # Optional[str]
+
+# Training hyperparameters
+
+# If greater than 1, backpropagation will be skipped and the gradients will be accumulated for the given number of steps.
+gradient_accumulation_steps: 1
+# The number of samples to include in each batch. This is the number of samples sent to each GPU.
+# Batch size per gpu = micro_batch_size * gradient_accumulation_steps
+micro_batch_size: 2
+eval_batch_size:
+num_epochs: 4
+warmup_steps: 100  # cannot use with warmup_ratio
+warmup_ratio: 0.05  # cannot use with warmup_steps
+learning_rate: 0.00003
+lr_quadratic_warmup:
+logging_steps:
+eval_steps: # Leave empty to eval at each epoch, integers for every N steps. decimal for fraction of total steps
+evals_per_epoch: # number of times per epoch to run evals, mutually exclusive with eval_steps
+save_strategy: # Set to `"no"` to skip checkpoint saves
+save_steps: # Leave empty to save at each epoch
+saves_per_epoch: # number of times per epoch to save a checkpoint, mutually exclusive with save_steps
+save_total_limit: # Checkpoints saved at a time
+# Maximum number of iterations to train for. It precedes num_epochs which means that
+# if both are set, num_epochs will not be guaranteed.
+# e.g., when 1 epoch is 1000 steps => `num_epochs: 2` and `max_steps: 100` will train for 100 steps
+max_steps:
+
+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]
+
+loss_watchdog_threshold: # High loss value, indicating the learning has broken down (a good estimate is ~2 times the loss at the start of training)
+loss_watchdog_patience: # Number of high-loss steps in a row before the trainer aborts (default: 3)
+
+# Save model as safetensors (require safetensors package)
+save_safetensors:
+
+# Whether to mask out or include the human's prompt from the training labels
+train_on_inputs: false
+# Group similarly sized data to minimize padding.
+# May be slower to start, as it must download and sort the entire dataset.
+# Note that training loss may have an oscillating pattern with this enabled.
+group_by_length: false
+
+# Whether to use gradient checkpointing https://huggingface.co/docs/transformers/v4.18.0/en/performance#gradient-checkpointing
+gradient_checkpointing: false
+# additional kwargs to pass to the trainer for gradient checkpointing
+# gradient_checkpointing_kwargs:
+#   use_reentrant: true
+
+# Stop training after this many evaluation losses have increased in a row
+# https://huggingface.co/transformers/v4.2.2/_modules/transformers/trainer_callback.html#EarlyStoppingCallback
+early_stopping_patience: 3
+
+# Specify a scheduler and kwargs to use with the optimizer
+lr_scheduler: # 'one_cycle' | 'log_sweep' | empty for cosine
+lr_scheduler_kwargs:
+cosine_min_lr_ratio: # decay lr to some percentage of the peak lr, e.g. cosine_min_lr_ratio=0.1 for 10% of peak lr
+cosine_constant_lr_ratio: # freeze lr at some percentage of the step, e.g. cosine_constant_lr_ratio=0.8 means start cosine_min_lr at 80% of training step (https://arxiv.org/pdf/2308.04014.pdf)
+
+# For one_cycle optim
+lr_div_factor: # Learning rate div factor
+
+# Specify optimizer
+# Valid values are driven by the Transformers OptimizerNames class, see:
+# https://github.com/huggingface/transformers/blob/95b374952dc27d8511541d6f5a4e22c9ec11fb24/src/transformers/training_args.py#L134
+#
+# Note that not all optimizers may be available in your environment, ex: 'adamw_anyprecision' is part of
+# torchdistx, 'adamw_bnb_8bit' is part of bnb.optim.Adam8bit, etc. When in doubt, it is recommended to start with the optimizer used
+# in the examples/ for your model and fine-tuning use case.
+#
+# Valid values for 'optimizer' include:
+# - adamw_hf
+# - adamw_torch
+# - adamw_torch_fused
+# - adamw_torch_xla
+# - adamw_apex_fused
+# - adafactor
+# - adamw_anyprecision
+# - sgd
+# - adagrad
+# - adamw_bnb_8bit
+# - lion_8bit
+# - lion_32bit
+# - paged_adamw_32bit
+# - paged_adamw_8bit
+# - paged_lion_32bit
+# - paged_lion_8bit
+# - galore_adamw
+# - galore_adamw_8bit
+# - galore_adafactor
+# - galore_adamw_layerwise
+# - galore_adamw_8bit_layerwise
+# - galore_adafactor_layerwise
+optimizer:
+# Dictionary of arguments to pass to the optimizer
+optim_args:
+# For Galore Optimizers the following optim_args are available
+# rank:  # type: int
+# update_proj_gap  # type: int
+# scale  # type: float
+# proj_type:  # type: str, default = std
+
+# The target modules to optimize, i.e. the module names that you would like to train, right now this is used only for GaLore algorithm
+optim_target_modules:
+# - self_attn  # for llama
+# - mlp
+
+# Specify weight decay
+weight_decay:
+# adamw hyperparams
+adam_beta1:
+adam_beta2:
+adam_epsilon:
+# Gradient clipping max norm
+max_grad_norm:
+
+# Augmentation techniques
+# NEFT https://arxiv.org/abs/2310.05914, set this to a number (paper default is 5) to add noise to embeddings
+# currently only supported on Llama and Mistral
+neftune_noise_alpha:
+
+# Whether to bettertransformers
+flash_optimum:
+# Whether to use xformers attention patch https://github.com/facebookresearch/xformers:
+xformers_attention:
+# Whether to use flash attention patch https://github.com/Dao-AILab/flash-attention:
+flash_attention:
+flash_attn_cross_entropy:  # Whether to use flash-attention cross entropy implementation - advanced use only
+flash_attn_rms_norm:  # Whether to use flash-attention rms norm implementation - advanced use only
+flash_attn_fuse_qkv: # Whether to fuse QKV into a single operation
+flash_attn_fuse_mlp: # Whether to fuse part of the MLP into a single operation
+# Whether to use scaled-dot-product attention
+# https://pytorch.org/docs/stable/generated/torch.nn.functional.scaled_dot_product_attention.html
+sdp_attention:
+# Shifted-sparse attention (only llama) - https://arxiv.org/pdf/2309.12307.pdf
+s2_attention:
+# Resume from a specific checkpoint dir
+resume_from_checkpoint:
+# If resume_from_checkpoint isn't set and you simply want it to start where it left off.
+# Be careful with this being turned on between different models.
+auto_resume_from_checkpoints: false
+
+# Don't mess with this, it's here for accelerate and torchrun
+local_rank:
+
+# Add or change special tokens.
+# If you add tokens here, you don't need to add them to the `tokens` list.
+special_tokens:
+  # bos_token: "<s>"
+  # eos_token: "</s>"
+  # unk_token: "<unk>"
+  # pad_token: "[PAD]"
+
+# Add extra tokens.
+tokens:
+
+# FSDP
+fsdp:
+fsdp_config:
+
+# Deepspeed config path. e.g., deepspeed_configs/zero3.json
+deepspeed:
+
+# Advanced DDP Arguments
+ddp_timeout:
+ddp_bucket_cap_mb:
+ddp_broadcast_buffers:
+
+# Path to torch distx for optim 'adamw_anyprecision'
+torchdistx_path:
+
+# Set to HF dataset for type: 'completion' for streaming instead of pre-tokenize
+pretraining_dataset:
+
+# Debug mode
+debug:
+
+# Seed
+seed:
+
+# Allow overwrite yml config using from cli
+strict:
 ```

docs/dataset-formats/conversation.qmd

@@ -0,0 +1,63 @@
+---
+title: Conversation
+description: Conversation format for supervised fine-tuning.
+order: 3
+---
+
+## sharegpt
+
+conversations where `from` is `human`/`gpt`. (optional: first row with role `system` to override default system prompt)
+
+```{.json filename="data.jsonl"}
+{"conversations": [{"from": "...", "value": "..."}]}
+```
+
+Note: `type: sharegpt` opens special configs:
+- `conversation`: enables conversions to many Conversation types. Refer to the 'name' [here](https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py) for options.
+- `roles`: allows you to specify the roles for input and output. This is useful for datasets with custom roles such as `tool` etc to support masking.
+- `field_human`: specify the key to use instead of `human` in the conversation.
+- `field_model`: specify the key to use instead of `gpt` in the conversation.
+
+```yaml
+datasets:
+    path: ...
+    type: sharegpt
+
+    conversation: # Options (see Conversation 'name'): https://github.com/lm-sys/FastChat/blob/main/fastchat/conversation.py
+    field_human: # Optional[str]. Human key to use for conversation.
+    field_model: # Optional[str]. Assistant key to use for conversation.
+    # Add additional keys from your dataset as input or output roles
+    roles:
+      input: # Optional[List[str]]. These will be masked based on train_on_input
+      output: # Optional[List[str]].
+```
+
+## pygmalion
+
+```{.json filename="data.jsonl"}
+{"conversations": [{"role": "...", "value": "..."}]}
+```
+
+## sharegpt.load_role
+
+conversations where `role` is used instead of `from`
+
+```{.json filename="data.jsonl"}
+{"conversations": [{"role": "...", "value": "..."}]}
+```
+
+## sharegpt.load_guanaco
+
+conversations where `from` is `prompter` `assistant` instead of default sharegpt
+
+```{.json filename="data.jsonl"}
+{"conversations": [{"from": "...", "value": "..."}]}
+```
+
+## sharegpt_jokes
+
+creates a chat where bot is asked to tell a joke, then explain why the joke is funny
+
+```{.json filename="data.jsonl"}
+{"conversations": [{"title": "...", "text": "...", "explanation": "..."}]}
+```

docs/dataset-formats/index.qmd

@@ -0,0 +1,14 @@
+---
+title: Dataset Formats
+description: Supported dataset formats.
+listing:
+  fields: [title, description]
+  type: table
+  sort-ui: false
+  filter-ui: false
+  max-description-length: 250
+---
+
+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:

docs/dataset-formats/inst_tune.qmd

@@ -0,0 +1,189 @@
+---
+title: Instruction Tuning
+description: Instruction tuning formats for supervised fine-tuning.
+order: 2
+---
+
+## alpaca
+
+instruction; input(optional)
+
+```{.json filename="data.jsonl"}
+{"instruction": "...", "input": "...", "output": "..."}
+```
+
+## jeopardy
+
+question and answer
+
+```{.json filename="data.jsonl"}
+{"question": "...", "category": "...", "answer": "..."}
+```
+
+## oasst
+
+instruction
+
+```{.json filename="data.jsonl"}
+{"INSTRUCTION": "...", "RESPONSE": "..."}
+```
+
+## gpteacher
+
+instruction; input(optional)
+
+```{.json filename="data.jsonl"}
+{"instruction": "...", "input": "...", "response": "..."}
+```
+
+## reflection
+
+instruction with reflect; input(optional)
+
+```{.json filename="data.jsonl"}
+{"instruction": "...", "input": "...", "output": "...", "reflection": "...", "corrected": "..."}
+```
+
+## explainchoice
+
+question, choices, (solution OR explanation)
+
+```{.json filename="data.jsonl"}
+{"question": "...", "choices": ["..."], "solution": "...", "explanation": "..."}
+```
+
+## concisechoice
+
+question, choices, (solution OR explanation)
+
+```{.json filename="data.jsonl"}
+{"question": "...", "choices": ["..."], "solution": "...", "explanation": "..."}
+```
+
+## summarizetldr
+
+article and summary
+
+```{.json filename="data.jsonl"}
+{"article": "...", "summary": "..."}
+```
+
+## alpaca_chat
+
+basic instruct for alpaca chat
+
+```{.json filename="data.jsonl"}
+{"instruction": "...", "input": "...", "response": "..."}
+```
+
+## alpaca_chat.load_qa
+
+question and answer for alpaca chat
+
+```{.json filename="data.jsonl"}
+{"question": "...", "answer": "..."}
+```
+
+## alpaca_chat.load_concise
+
+question and answer for alpaca chat, for concise answers
+
+```{.json filename="data.jsonl"}
+{"instruction": "...", "input": "...", "response": "..."}
+```
+
+## alpaca_chat.load_camel_ai
+
+question and answer for alpaca chat, for load_camel_ai
+
+```{.json filename="data.jsonl"}
+{"message_1": "...", "message_2": "..."}
+```
+
+## alpaca_w_system.load_open_orca
+
+support for open orca datasets with included system prompts, instruct
+
+```{.json filename="data.jsonl"}
+{"system_prompt": "...", "question": "...", "response": "..."}
+```
+
+## context_qa
+
+in context question answering from an article
+
+```{.json filename="data.jsonl"}
+{"article": "...", "question": "...", "answer": "..."}
+```
+
+## context_qa.load_v2
+
+in context question answering (alternate)
+
+```{.json filename="data.jsonl"}
+{"context": "...", "question": "...", "answer": "..."}
+```
+
+## context_qa.load_404
+
+in context question answering from an article, with default response for no answer from context
+
+```{.json filename="data.jsonl"}
+{"article": "...", "unanswerable_question": "..."}
+```
+
+## creative_acr.load_answer
+
+instruction and revision
+
+```{.json filename="data.jsonl"}
+{"instruction": "...", "revision": "..."}
+```
+
+## creative_acr.load_critique
+
+critique
+
+```{.json filename="data.jsonl"}
+{"scores": "...", "critiques": "...", "instruction": "...", "answer": "..."}
+```
+
+## creative_acr.load_revise
+
+critique and revise
+
+```{.json filename="data.jsonl"}
+{"scores": "...", "critiques": "...", "instruction": "...", "answer": "...", "revision": "..."}
+```
+
+## metharme
+
+instruction, adds additional eos tokens
+
+```{.json filename="data.jsonl"}
+{"prompt": "...", "generation": "..."}
+```
+
+## How to add custom prompt format
+
+For a dataset that is preprocessed for instruction purposes:
+
+```{.json filename="data.jsonl"}
+{"input": "...", "output": "..."}
+```
+
+You can use this example in your YAML config:
+
+```{.yaml filename="config.yaml"}
+datasets:
+  - path: repo
+    type:
+      system_prompt: ""
+      field_system: system
+      field_instruction: input
+      field_output: output
+      format: "[INST] {instruction} [/INST]"
+      no_input_format: "[INST] {instruction} [/INST]"
+```
+
+See full config options under [here](../config.qmd).

docs/dataset-formats/pretraining.qmd

@@ -0,0 +1,26 @@
+---
+title: Pre-training
+description: Data format for a pre-training completion task.
+order: 1
+---
+
+For pretraining, there is no prompt template or roles.  The only required field is `text`:
+
+```{.json filename="data.jsonl"}
+{"text": "first row"}
+{"text": "second row"}
+...
+```
+
+:::{.callout-note}
+
+### Streaming is recommended for large datasets
+
+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
+...
+```
+
+:::

docs/dataset-formats/template_free.qmd

@@ -0,0 +1,7 @@
+---
+title: Template-Free
+description: Construct prompts without a template.
+order: 4
+---
+
+See [these docs](../input_output.qmd).

docs/dataset-formats/tokenized.qmd

@@ -0,0 +1,12 @@
+---
+title: Custom Pre-Tokenized Dataset
+description: How to use a custom pre-tokenized dataset.
+order: 5
+---
+
+- Do not pass a `type:` in your axolotl config.
+- Columns in Dataset must be exactly `input_ids`, `attention_mask`, `labels`
+
+```{.yaml filename="config.yml"}
+- path: ...
+```

docs/dataset_preprocessing.qmd

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

docs/fsdp_qlora.qmd

@@ -1,5 +1,5 @@
 ---
-title: FDSP + QLoRA
+title: "FDSP + QLoRA"
 description: Use FSDP with QLoRA to fine-tune large LLMs on consumer GPUs.
 format:
   html:

docs/input_output.qmd

@@ -43,7 +43,7 @@ labels so that your model can focus on predicting the outputs only.
 ### You may not want prompt templates
 
 However, there are many situations where you don't want to use one of
-these formats or templates (I usually don't!). This is because they can:
+these formats or templates. This is because they can:
 
 -   Add unnecessary boilerplate to your prompts.
 -   Create artifacts like special delimiters `<|im_start|>` that can
@@ -91,8 +91,9 @@ format into a jsonl file (below is the first row from the file
 
 ```bash
 $ head -n1 output.jsonl | python -m json.tool
+```
 
-{.cell-output .cell-output-stdout}
+:::{.cell-output .cell-output-stdout}
     {
         "segments": [
             {
@@ -113,7 +114,7 @@ $ head -n1 output.jsonl | python -m json.tool
             }
         ]
     }
-```
+:::
 
 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:
@@ -238,8 +239,9 @@ version is repeated below for reference):
 
 ```bash
 $ head -n1 output.jsonl | python -m json.tool
+```
 
-{.cell-output .cell-output-stdout}
+:::{.cell-output .cell-output-stdout}
     {
         "segments": [
             {
@@ -260,4 +262,4 @@ $ head -n1 output.jsonl | python -m json.tool
             }
         ]
     }
-```
+:::

docs/rlhf.qmd

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

examples/cerebras/btlm-ft.yml

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

examples/cerebras/qlora.yml

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

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

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

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

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

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

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

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

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

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

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

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

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

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

@@ -1,216 +1,223 @@
 {
-  "cells": [
-    {
-      "cell_type": "markdown",
-      "metadata": {
-        "id": "AKjdG7tbTb-n"
-      },
-      "source": [
-        "# Example notebook for running Axolotl on google colab"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {
-        "id": "RcbNpOgWRcii"
-      },
-      "outputs": [],
-      "source": [
-        "import torch\n",
-        "# Check so there is a gpu available, a T4(free tier) is enough to run this notebook\n",
-        "assert (torch.cuda.is_available()==True)"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {
-        "id": "h3nLav8oTRA5"
-      },
-      "source": [
-        "## Install Axolotl and dependencies"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {
-        "colab": {
-          "base_uri": "https://localhost:8080/"
-        },
-        "id": "3c3yGAwnOIdi",
-        "outputId": "e3777b5a-40ef-424f-e181-62dfecd1dd01"
-      },
-      "outputs": [],
-      "source": [
-        "!pip install torch==\"2.1.2\"\n",
-        "!pip install -e git+https://github.com/OpenAccess-AI-Collective/axolotl#egg=axolotl\n",
-        "!pip install flash-attn==\"2.5.0\"\n",
-        "!pip install deepspeed==\"0.13.1\""
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {
-        "id": "BW2MFr7HTjub"
-      },
-      "source": [
-        "## Create an yaml config file"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {
-        "id": "9pkF2dSoQEUN"
-      },
-      "outputs": [],
-      "source": [
-        "import yaml\n",
-        "\n",
-        "# Your YAML string\n",
-        "yaml_string = \"\"\"\n",
-        "base_model: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T\n",
-        "model_type: LlamaForCausalLM\n",
-        "tokenizer_type: LlamaTokenizer\n",
-        "is_llama_derived_model: true\n",
-        "\n",
-        "load_in_8bit: false\n",
-        "load_in_4bit: true\n",
-        "strict: false\n",
-        "\n",
-        "datasets:\n",
-        "  - path: mhenrichsen/alpaca_2k_test\n",
-        "    type: alpaca\n",
-        "dataset_prepared_path:\n",
-        "val_set_size: 0.05\n",
-        "output_dir: ./qlora-out\n",
-        "\n",
-        "adapter: qlora\n",
-        "lora_model_dir:\n",
-        "\n",
-        "sequence_len: 1096\n",
-        "sample_packing: true\n",
-        "pad_to_sequence_len: true\n",
-        "\n",
-        "lora_r: 32\n",
-        "lora_alpha: 16\n",
-        "lora_dropout: 0.05\n",
-        "lora_target_modules:\n",
-        "lora_target_linear: true\n",
-        "lora_fan_in_fan_out:\n",
-        "\n",
-        "wandb_project:\n",
-        "wandb_entity:\n",
-        "wandb_watch:\n",
-        "wandb_name:\n",
-        "wandb_log_model:\n",
-        "\n",
-        "mlflow_experiment_name: colab-example\n",
-        "\n",
-        "gradient_accumulation_steps: 1\n",
-        "micro_batch_size: 1\n",
-        "num_epochs: 4\n",
-        "max_steps: 20\n",
-        "optimizer: paged_adamw_32bit\n",
-        "lr_scheduler: cosine\n",
-        "learning_rate: 0.0002\n",
-        "\n",
-        "train_on_inputs: false\n",
-        "group_by_length: false\n",
-        "bf16: false\n",
-        "fp16: true\n",
-        "tf32: false\n",
-        "\n",
-        "gradient_checkpointing: true\n",
-        "early_stopping_patience:\n",
-        "resume_from_checkpoint:\n",
-        "local_rank:\n",
-        "logging_steps: 1\n",
-        "xformers_attention:\n",
-        "flash_attention: false\n",
-        "\n",
-        "warmup_steps: 10\n",
-        "evals_per_epoch:\n",
-        "saves_per_epoch:\n",
-        "debug:\n",
-        "deepspeed:\n",
-        "weight_decay: 0.0\n",
-        "fsdp:\n",
-        "fsdp_config:\n",
-        "special_tokens:\n",
-        "\n",
-        "\"\"\"\n",
-        "\n",
-        "# Convert the YAML string to a Python dictionary\n",
-        "yaml_dict = yaml.safe_load(yaml_string)\n",
-        "\n",
-        "# Specify your file path\n",
-        "file_path = 'test_axolotl.yaml'\n",
-        "\n",
-        "# Write the YAML file\n",
-        "with open(file_path, 'w') as file:\n",
-        "    yaml.dump(yaml_dict, file)\n"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {
-        "id": "bidoj8YLTusD"
-      },
-      "source": [
-        "## Launch the training"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {
-        "colab": {
-          "base_uri": "https://localhost:8080/"
-        },
-        "id": "ydTI2Jk2RStU",
-        "outputId": "d6d0df17-4b53-439c-c802-22c0456d301b"
-      },
-      "outputs": [],
-      "source": [
-        "# Buy using the ! the comand will be executed as a bash command\n",
-        "!accelerate launch -m axolotl.cli.train /content/test_axolotl.yaml"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {},
-      "source": [
-        "## Play with inference"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {},
-      "outputs": [],
-      "source": [
-        "# Buy using the ! the comand will be executed as a bash command\n",
-        "!accelerate launch -m axolotl.cli.inference /content/test_axolotl.yaml \\\n",
-        "    --qlora_model_dir=\"./qlora-out\" --gradio"
-      ]
-    }
-  ],
-  "metadata": {
-    "accelerator": "GPU",
-    "colab": {
-      "gpuType": "T4",
-      "provenance": []
-    },
-    "kernelspec": {
-      "display_name": "Python 3",
-      "name": "python3"
-    },
-    "language_info": {
-      "name": "python"
-    }
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "id": "AKjdG7tbTb-n"
+   },
+   "source": [
+    "# Example notebook for running Axolotl on google colab"
+   ]
   },
-  "nbformat": 4,
-  "nbformat_minor": 0
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "id": "RcbNpOgWRcii"
+   },
+   "outputs": [],
+   "source": [
+    "import torch\n",
+    "# Check so there is a gpu available, a T4(free tier) is enough to run this notebook\n",
+    "assert (torch.cuda.is_available()==True)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "id": "h3nLav8oTRA5"
+   },
+   "source": [
+    "## Install Axolotl and dependencies"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "colab": {
+     "base_uri": "https://localhost:8080/"
+    },
+    "id": "3c3yGAwnOIdi",
+    "outputId": "e3777b5a-40ef-424f-e181-62dfecd1dd01"
+   },
+   "outputs": [],
+   "source": [
+    "!pip install torch==\"2.1.2\"\n",
+    "!pip install -e git+https://github.com/OpenAccess-AI-Collective/axolotl#egg=axolotl\n",
+    "!pip install flash-attn==\"2.5.0\"\n",
+    "!pip install deepspeed==\"0.13.1\"!pip install mlflow==\"2.13.0\""
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "id": "BW2MFr7HTjub"
+   },
+   "source": [
+    "## Create an yaml config file"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "id": "9pkF2dSoQEUN"
+   },
+   "outputs": [],
+   "source": [
+    "import yaml\n",
+    "\n",
+    "# Your YAML string\n",
+    "yaml_string = \"\"\"\n",
+    "base_model: TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T\n",
+    "model_type: LlamaForCausalLM\n",
+    "tokenizer_type: LlamaTokenizer\n",
+    "\n",
+    "load_in_8bit: false\n",
+    "load_in_4bit: true\n",
+    "strict: false\n",
+    "\n",
+    "datasets:\n",
+    "  - path: mhenrichsen/alpaca_2k_test\n",
+    "    type: alpaca\n",
+    "dataset_prepared_path:\n",
+    "val_set_size: 0.05\n",
+    "output_dir: ./outputs/qlora-out\n",
+    "\n",
+    "adapter: qlora\n",
+    "lora_model_dir:\n",
+    "\n",
+    "sequence_len: 4096\n",
+    "sample_packing: true\n",
+    "eval_sample_packing: false\n",
+    "pad_to_sequence_len: true\n",
+    "\n",
+    "lora_r: 32\n",
+    "lora_alpha: 16\n",
+    "lora_dropout: 0.05\n",
+    "lora_target_modules:\n",
+    "lora_target_linear: true\n",
+    "lora_fan_in_fan_out:\n",
+    "\n",
+    "wandb_project:\n",
+    "wandb_entity:\n",
+    "wandb_watch:\n",
+    "wandb_name:\n",
+    "wandb_log_model:\n",
+    "\n",
+    "gradient_accumulation_steps: 4\n",
+    "micro_batch_size: 2\n",
+    "num_epochs: 4\n",
+    "optimizer: paged_adamw_32bit\n",
+    "lr_scheduler: cosine\n",
+    "learning_rate: 0.0002\n",
+    "\n",
+    "train_on_inputs: false\n",
+    "group_by_length: false\n",
+    "bf16: auto\n",
+    "fp16:\n",
+    "tf32: false\n",
+    "\n",
+    "gradient_checkpointing: true\n",
+    "early_stopping_patience:\n",
+    "resume_from_checkpoint:\n",
+    "local_rank:\n",
+    "logging_steps: 1\n",
+    "xformers_attention:\n",
+    "flash_attention: true\n",
+    "\n",
+    "warmup_steps: 10\n",
+    "evals_per_epoch: 4\n",
+    "saves_per_epoch: 1\n",
+    "debug:\n",
+    "deepspeed:\n",
+    "weight_decay: 0.0\n",
+    "fsdp:\n",
+    "fsdp_config:\n",
+    "special_tokens:\n",
+    "\n",
+    "\"\"\"\n",
+    "\n",
+    "# Convert the YAML string to a Python dictionary\n",
+    "yaml_dict = yaml.safe_load(yaml_string)\n",
+    "\n",
+    "# Specify your file path\n",
+    "file_path = 'test_axolotl.yaml'\n",
+    "\n",
+    "# Write the YAML file\n",
+    "with open(file_path, 'w') as file:\n",
+    "    yaml.dump(yaml_dict, file)\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {
+    "id": "bidoj8YLTusD"
+   },
+   "source": [
+    "## Launch the training"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {
+    "colab": {
+     "base_uri": "https://localhost:8080/"
+    },
+    "id": "ydTI2Jk2RStU",
+    "outputId": "d6d0df17-4b53-439c-c802-22c0456d301b"
+   },
+   "outputs": [],
+   "source": [
+    "# Buy using the ! the comand will be executed as a bash command\n",
+    "!accelerate launch -m axolotl.cli.train /content/test_axolotl.yaml"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Play with inference"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Buy using the ! the comand will be executed as a bash command\n",
+    "!accelerate launch -m axolotl.cli.inference /content/test_axolotl.yaml \\\n",
+    "    --qlora_model_dir=\"./qlora-out\" --gradio"
+   ]
+  }
+ ],
+ "metadata": {
+  "accelerator": "GPU",
+  "colab": {
+   "gpuType": "T4",
+   "provenance": []
+  },
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.12.1"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
 }

examples/dbrx/16bit-lora.yaml

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

examples/dbrx/8bit-lora.yaml

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

examples/dbrx/README.md

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

examples/dbrx/fft-ds-zero3.yaml

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

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

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

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

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

examples/falcon/config-7b.yml

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

examples/gemma/qlora.yml

@@ -12,7 +12,7 @@ datasets:
   - path: mhenrichsen/alpaca_2k_test
     type: alpaca
 val_set_size: 0.1
-output_dir: ./out
+output_dir: ./outputs/out
 
 adapter: qlora
 lora_r: 32
@@ -21,7 +21,8 @@ lora_dropout: 0.05
 lora_target_linear: true
 
 sequence_len: 4096
-sample_packing: false
+sample_packing: true
+eval_sample_packing: false
 pad_to_sequence_len: true
 
 wandb_project:

examples/gptj/qlora.yml

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

examples/jamba/README.md

@@ -0,0 +1,10 @@
+# Jamba
+
+- ✅ qlora w/ deepspeed Zero-2 needs at least 2x GPUs and
+  - 35GiB VRAM per GPU w minimal context length
+  - 56GiB VRAM per GPU (w multipack enabled)
+- ✅ qlora w/ deepspeed Zero-3 needs at least 2x GPUs and 67GiB VRAM (wtf?)
+- ✅ qlora single-gpu, ~51GiB VRAM
+- ✅ multipack
+- ❓ FSDP
+- ❓ 8-bit LoRA

examples/jamba/qlora.yaml

@@ -0,0 +1,62 @@
+base_model: ai21labs/Jamba-v0.1
+trust_remote_code: true
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+datasets:
+  - path: mhenrichsen/alpaca_2k_test
+    type: alpaca
+dataset_prepared_path:
+val_set_size: 0.0
+output_dir: ./outputs/out
+
+sequence_len: 4096
+sample_packing: false
+pad_to_sequence_len: false
+eval_sample_packing: false
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+adapter: qlora
+lora_r: 8
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+
+low_cpu_mem_usage: true
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 2
+optimizer: paged_adamw_8bit
+lr_scheduler: cosine
+learning_rate: 0.00001
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 10
+evals_per_epoch:
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+special_tokens:

examples/jamba/qlora_deepspeed.yaml

@@ -0,0 +1,62 @@
+base_model: ai21labs/Jamba-v0.1
+trust_remote_code: true
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+datasets:
+  - path: mhenrichsen/alpaca_2k_test
+    type: alpaca
+dataset_prepared_path:
+val_set_size: 0.0
+output_dir: ./outputs/out
+
+sequence_len: 4096
+sample_packing: false
+pad_to_sequence_len: false
+eval_sample_packing: false
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+adapter: qlora
+lora_r: 8
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+
+low_cpu_mem_usage: true
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 2
+optimizer: paged_adamw_8bit
+lr_scheduler: cosine
+learning_rate: 0.00001
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 10
+evals_per_epoch:
+saves_per_epoch: 1
+debug:
+deepspeed: deepspeed_configs/zero2.json
+weight_decay: 0.0
+special_tokens:

examples/jeopardy-bot/config.yml

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

examples/llama-2/fft_optimized.yml

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

examples/llama-2/gptq-lora.yml

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

examples/llama-2/lisa.yml

@@ -0,0 +1,75 @@
+base_model: NousResearch/Llama-2-7b-hf
+model_type: LlamaForCausalLM
+tokenizer_type: LlamaTokenizer
+
+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: 0.05
+output_dir: ./outputs/lisa-out
+
+sequence_len: 4096
+sample_packing: true
+pad_to_sequence_len: true
+
+adapter:
+lora_model_dir:
+lora_r:
+lora_alpha:
+lora_dropout:
+lora_target_linear:
+lora_fan_in_fan_out:
+
+lisa_n_layers: 4
+lisa_step_interval: 20
+lisa_layers_attribute: model.layers
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 2
+micro_batch_size: 1
+num_epochs: 1
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 5e-5 # recommendation from lisa paper for 7b
+
+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
+flash_attn_cross_entropy: false
+flash_attn_rms_norm: true
+flash_attn_fuse_qkv: false
+flash_attn_fuse_mlp: true
+
+warmup_steps: 100
+evals_per_epoch: 4
+eval_table_size:
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.1
+fsdp:
+fsdp_config:
+special_tokens:
+  bos_token: "<s>"
+  eos_token: "</s>"
+  unk_token: "<unk>"

examples/llama-2/loftq.yml

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

examples/llama-2/lora.yml

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

examples/llama-2/qlora-fsdp.yml

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

examples/llama-2/qlora.yml

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

examples/llama-2/relora.yml

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

examples/llama-3/README.md

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

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

@@ -0,0 +1,58 @@
+base_model: meta-llama/Meta-Llama-3-8B
+model_type: LlamaForCausalLM
+tokenizer_type: AutoTokenizer
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: tatsu-lab/alpaca
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.05
+output_dir: ./outputs/out
+
+sequence_len: 8192
+sample_packing: true
+pad_to_sequence_len: true
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 8
+micro_batch_size: 1
+num_epochs: 1
+optimizer: paged_adamw_8bit
+lr_scheduler: cosine
+learning_rate: 2e-5
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+early_stopping_patience:
+resume_from_checkpoint:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 100
+evals_per_epoch: 2
+eval_table_size:
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:
+  pad_token: <|end_of_text|>

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

@@ -0,0 +1,76 @@
+base_model: meta-llama/Meta-Llama-3-8B-Instruct
+model_type: LlamaForCausalLM
+tokenizer_type: AutoTokenizer
+
+load_in_8bit: true
+load_in_4bit: false
+strict: false
+
+chat_template: llama3
+datasets:
+  - path: fozziethebeat/alpaca_messages_2k_test
+    type: chat_template
+    chat_template: llama3
+    field_messages: messages
+    message_field_role: role
+    message_field_content: content
+    roles:
+      user:
+        - user
+      assistant:
+        - assistant
+
+dataset_prepared_path:
+val_set_size: 0.05
+output_dir: ./outputs/lora-out
+
+sequence_len: 4096
+sample_packing: false
+pad_to_sequence_len: true
+
+adapter: lora
+lora_model_dir:
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 2
+num_epochs: 4
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+s2_attention:
+
+warmup_steps: 10
+evals_per_epoch: 4
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:

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

@@ -0,0 +1,70 @@
+base_model: meta-llama/Meta-Llama-3-8B
+model_type: LlamaForCausalLM
+tokenizer_type: AutoTokenizer
+
+load_in_8bit: true
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: mhenrichsen/alpaca_2k_test
+    type: alpaca
+dataset_prepared_path:
+val_set_size: 0.05
+output_dir: ./outputs/lora-out
+
+sequence_len: 4096
+sample_packing: true
+pad_to_sequence_len: true
+
+adapter: lora
+lora_model_dir:
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+lora_modules_to_save:
+  - embed_tokens
+  - lm_head
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 2
+num_epochs: 4
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+s2_attention:
+
+warmup_steps: 10
+evals_per_epoch: 4
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:
+   pad_token: <|end_of_text|>

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

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

examples/llama-3/qlora.yml

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

examples/mamba/config.yml

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

examples/mistral/bigstral-ds-zero3.yaml

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

examples/mistral/config.yml

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

examples/mistral/lora-mps.yml

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

examples/mistral/lora.yml

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

examples/mistral/mistral-qlora-fsdp.yml

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

examples/mistral/mistral-qlora-orpo.yml

@@ -0,0 +1,82 @@
+base_model: mistralai/Mistral-7B-v0.1
+model_type: MistralForCausalLM
+tokenizer_type: LlamaTokenizer
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+rl: orpo
+orpo_alpha: 0.1
+remove_unused_columns: false
+
+chat_template: chatml
+datasets:
+  - path: argilla/ultrafeedback-binarized-preferences-cleaned
+    type: chat_template.argilla
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.1
+output_dir: ./outputs/mistral-qlora-orpo-out
+
+adapter: qlora
+lora_model_dir:
+
+sequence_len: 4096
+sample_packing: false
+pad_to_sequence_len: true
+
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+lora_target_modules:
+  - gate_proj
+  - down_proj
+  - up_proj
+  - q_proj
+  - v_proj
+  - k_proj
+  - o_proj
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 2
+num_epochs: 1
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+loss_watchdog_threshold: 5.0
+loss_watchdog_patience: 3
+
+warmup_steps: 10
+evals_per_epoch: 4
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:

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

@@ -0,0 +1,81 @@
+base_model: mistral-community/Mixtral-8x22B-v0.1
+model_type: AutoModelForCausalLM
+tokenizer_type: LlamaTokenizer
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+datasets:
+  - path: tatsu-lab/alpaca
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.02
+output_dir: ./outputs/qlora-out
+
+model_config:
+  output_router_logits: true
+
+adapter: qlora
+lora_model_dir:
+
+sequence_len: 1024
+sample_packing: false
+pad_to_sequence_len: false
+
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 2
+num_epochs: 1
+optimizer: adamw_torch
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: true
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+loss_watchdog_threshold: 5.0
+loss_watchdog_patience: 3
+
+warmup_steps: 10
+evals_per_epoch: 4
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+weight_decay: 0.0
+fsdp:
+  - full_shard
+  - auto_wrap
+fsdp_config:
+  fsdp_limit_all_gathers: true
+  fsdp_sync_module_states: true
+  fsdp_offload_params: true
+  fsdp_use_orig_params: false
+  fsdp_cpu_ram_efficient_loading: true
+  fsdp_transformer_layer_cls_to_wrap: MixtralSparseMoeBlock
+  fsdp_state_dict_type: FULL_STATE_DICT
+  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
+special_tokens:

examples/mistral/mixtral-qlora-fsdp.yml

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

examples/mistral/mixtral.yml

@@ -12,7 +12,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.0
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 
 ## You can optionally freeze the entire model and unfreeze a subset of parameters
 unfrozen_parameters:

examples/mistral/mixtral_22.yml

@@ -0,0 +1,61 @@
+base_model: mistral-community/Mixtral-8x22B-v0.1
+model_type: AutoModelForCausalLM
+tokenizer_type: LlamaTokenizer
+trust_remote_code: true
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+unfrozen_parameters:
+  - ^lm_head.weight$
+  - ^model.embed_tokens.weight$
+  - model.layers.4[4-9]+.block_sparse_moe.gate
+  - model.layers.4[4-9]+.block_sparse_moe.experts
+  - model.layers.5[0-5]+.block_sparse_moe.gate
+  - model.layers.5[0-5]+.block_sparse_moe.experts
+
+model_config:
+  output_router_logits: true
+
+datasets:
+  - path: yahma/alpaca-cleaned
+    type: alpaca
+output_dir: ./outputs/out
+
+sequence_len: 8000
+sample_packing: true
+pad_to_sequence_len: true
+
+gradient_accumulation_steps: 1
+micro_batch_size: 1
+num_epochs: 3
+optimizer: adamw_bnb_8bit
+lr_scheduler: cosine
+learning_rate: 0.0001
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+save_total_limit: 1
+save_steps:
+debug:
+deepspeed: deepspeed_configs/zero3_bf16_cpuoffload_all.json
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:
+  eos_token: "<|im_end|>"
+tokens:
+  - "<|im_start|>"

examples/mistral/qlora.yml

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

examples/mpt-7b/config.yml

@@ -23,7 +23,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./mpt-alpaca-7b
+output_dir: ./outputs/mpt-alpaca-7b
 gradient_accumulation_steps: 1
 micro_batch_size: 1
 num_epochs: 4

examples/openllama-3b/config.yml

@@ -25,7 +25,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./openllama-out
+output_dir: ./outputs/openllama-out
 gradient_accumulation_steps: 1
 micro_batch_size: 1
 num_epochs: 4

examples/openllama-3b/lora.yml

@@ -31,7 +31,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 gradient_accumulation_steps: 1
 micro_batch_size: 2
 num_epochs: 4

examples/openllama-3b/qlora.yml

@@ -25,7 +25,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./qlora-out
+output_dir: ./outputs/qlora-out
 gradient_accumulation_steps: 1
 micro_batch_size: 2
 num_epochs: 4

examples/phi/phi-ft.yml

@@ -12,7 +12,7 @@ datasets:
 
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./phi-sft-out
+output_dir: ./outputs/phi-sft-out
 
 sequence_len: 2048
 sample_packing: true

examples/phi/phi-qlora.yml

@@ -12,7 +12,7 @@ datasets:
 
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./phi-sft-out
+output_dir: ./outputs/phi-sft-out
 
 sequence_len: 2048
 sample_packing: true

examples/phi/phi2-ft.yml

@@ -12,7 +12,7 @@ datasets:
 
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./phi-sft-out
+output_dir: ./outputs/phi-sft-out
 
 sequence_len: 2048
 sample_packing: true

examples/phi/phi3-ft-fsdp.yml

@@ -0,0 +1,83 @@
+base_model: microsoft/Phi-3-mini-4k-instruct
+model_type: AutoModelForCausalLM
+tokenizer_type: AutoTokenizer
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: mhenrichsen/alpaca_2k_test
+    type: alpaca
+
+dataset_prepared_path:
+val_set_size: 0
+output_dir: ./phi-sft-out
+
+sequence_len: 4096
+sample_packing: true
+pad_to_sequence_len: true
+trust_remote_code: true
+
+adapter:
+lora_model_dir:
+lora_r:
+lora_alpha:
+lora_dropout:
+lora_target_linear:
+lora_fan_in_fan_out:
+
+wandb_project: phi3
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 2
+micro_batch_size: 12
+num_epochs: 2
+optimizer: adamw_torch
+adam_beta2: 0.95
+adam_epsilon: 0.00001
+max_grad_norm: 1.0
+lr_scheduler: cosine
+learning_rate: 0.000003
+
+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: 100
+evals_per_epoch: 4
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.1
+fsdp:
+  - full_shard
+  - auto_wrap
+fsdp_config:
+  fsdp_limit_all_gathers: true
+  fsdp_sync_module_states: true
+  fsdp_offload_params: true
+  fsdp_use_orig_params: false
+  fsdp_cpu_ram_efficient_loading: true
+  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
+  fsdp_transformer_layer_cls_to_wrap: Phi3DecoderLayer
+  fsdp_state_dict_type: FULL_STATE_DICT
+  fsdp_sharding_strategy: FULL_SHARD
+resize_token_embeddings_to_32x: true
+special_tokens:
+  pad_token: "<|endoftext|>"

examples/phi/phi3-ft.yml

@@ -0,0 +1,64 @@
+base_model: microsoft/Phi-3-mini-4k-instruct
+trust_remote_code: true
+model_type: AutoModelForCausalLM
+tokenizer_type: AutoTokenizer
+chat_template: phi_3
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: garage-bAInd/Open-Platypus
+    type: alpaca:phi
+
+dataset_prepared_path:
+val_set_size: 0.01
+output_dir: ./out
+
+sequence_len: 4096
+sample_packing: true
+pad_to_sequence_len: true
+
+adapter: lora
+lora_model_dir:
+lora_r: 64
+lora_alpha: 32
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+
+gradient_accumulation_steps: 1
+micro_batch_size: 2
+num_epochs: 1
+optimizer: adamw_torch
+adam_beta2: 0.95
+adam_epsilon: 0.00001
+max_grad_norm: 1.0
+lr_scheduler: cosine
+learning_rate: 5.0e-6
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: True
+early_stopping_patience: 3
+logging_steps: 1
+flash_attention: true
+
+eval_steps: 1000
+save_steps: 5000
+eval_table_size: 2
+eval_batch_size: 2
+eval_sample_packing: false
+eval_max_new_tokens: 32
+eval_causal_lm_metrics: ["perplexity"]
+do_causal_lm_eval: true
+
+warmup_ratio: 0.2
+debug: true
+weight_decay: 0.1
+resize_token_embeddings_to_32x: true

examples/pythia-12b/config.yml

@@ -26,7 +26,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./pythia-12b
+output_dir: ./outputs/pythia-12b
 gradient_accumulation_steps: 1
 micro_batch_size: 1
 num_epochs: 5

examples/pythia/lora.yml

@@ -20,7 +20,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./lora-alpaca-pythia
+output_dir: ./outputs/lora-alpaca-pythia
 gradient_accumulation_steps: 1
 micro_batch_size: 4
 num_epochs: 4

examples/qwen/README.md

@@ -0,0 +1,10 @@
+# Qwen
+
+TODO
+
+# Qwen2 MoE
+
+✅ multipack
+✅ qwen2_moe 4-bit QLoRA
+✅ qwen2_moe 16-bit LoRA
+❓ qwen2_moe 8-bit LoRA

examples/qwen/lora.yml

@@ -13,7 +13,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 2048  # supports up to 8192
 sample_packing: false

examples/qwen/qlora.yml

@@ -13,7 +13,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 2048  # supports up to 8192
 sample_packing: false

examples/qwen/qwen2-moe-lora.yaml

@@ -0,0 +1,64 @@
+base_model: Qwen/Qwen1.5-MoE-A2.7B
+trust_remote_code: true
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: mhenrichsen/alpaca_2k_test
+    type: alpaca
+dataset_prepared_path:
+val_set_size: 0.05
+output_dir: ./outputs/out
+
+sequence_len: 1024  # supports up to 32k
+sample_packing: false
+pad_to_sequence_len: false
+
+adapter: lora
+lora_model_dir:
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 4
+optimizer: paged_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: false
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 10
+evals_per_epoch: 4
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:

examples/qwen/qwen2-moe-qlora.yaml

@@ -0,0 +1,64 @@
+base_model: Qwen/Qwen1.5-MoE-A2.7B
+trust_remote_code: true
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+datasets:
+  - path: mhenrichsen/alpaca_2k_test
+    type: alpaca
+dataset_prepared_path:
+val_set_size: 0.05
+output_dir: ./outputs/out
+
+sequence_len: 1024  # supports up to 32k
+sample_packing: false
+pad_to_sequence_len: false
+
+adapter: qlora
+lora_model_dir:
+lora_r: 32
+lora_alpha: 16
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 4
+micro_batch_size: 1
+num_epochs: 4
+optimizer: paged_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: false
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 10
+evals_per_epoch: 4
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:

examples/qwen2/qlora-fsdp.yaml

@@ -0,0 +1,75 @@
+base_model: Qwen/Qwen2-7B
+trust_remote_code: true
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+datasets:
+  - path: tatsu-lab/alpaca
+    type: alpaca
+dataset_prepared_path:
+val_set_size: 0.05
+output_dir: ./outputs/out
+
+sequence_len: 2048
+sample_packing: true
+eval_sample_packing: true
+pad_to_sequence_len: true
+
+adapter: qlora
+lora_model_dir:
+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: 4
+micro_batch_size: 1
+num_epochs: 4
+optimizer: adamw_torch
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: true
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: false
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 10
+evals_per_epoch: 4
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+  - full_shard
+  - auto_wrap
+fsdp_config:
+  fsdp_limit_all_gathers: true
+  fsdp_sync_module_states: true
+  fsdp_offload_params: true
+  fsdp_use_orig_params: false
+  fsdp_cpu_ram_efficient_loading: true
+  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
+  fsdp_transformer_layer_cls_to_wrap: Qwen2DecoderLayer
+  fsdp_state_dict_type: FULL_STATE_DICT
+special_tokens:

examples/redpajama/config-3b.yml

@@ -24,7 +24,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./redpajama-alpaca-3b
+output_dir: ./outputs/redpajama-alpaca-3b
 batch_size: 4
 micro_batch_size: 1
 num_epochs: 4

examples/replit-3b/config-lora.yml

@@ -23,7 +23,7 @@ wandb_entity:
 wandb_watch:
 wandb_name:
 wandb_log_model:
-output_dir: ./lora-replit
+output_dir: ./outputs/lora-replit
 batch_size: 8
 micro_batch_size: 1
 num_epochs: 4

examples/stablelm-2/1.6b/fft.yml

@@ -12,7 +12,7 @@ datasets:
     type: alpaca
 dataset_prepared_path: last_run_prepared
 val_set_size: 0.05
-output_dir: ./out
+output_dir: ./outputs/out
 
 sequence_len: 4096
 sample_packing: true

examples/stablelm-2/1.6b/lora.yml

@@ -12,7 +12,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0.05
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/starcoder2/qlora.yml

@@ -11,7 +11,7 @@ datasets:
 
 dataset_prepared_path:
 val_set_size: 0.2
-output_dir: ./qlora
+output_dir: ./outputs/qlora
 
 adapter: qlora
 lora_model_dir:

examples/tiny-llama/lora-mps.yml

@@ -11,7 +11,7 @@ datasets:
     type: alpaca
 dataset_prepared_path:
 val_set_size: 0
-output_dir: ./lora-out
+output_dir: ./outputs/lora-out
 
 sequence_len: 4096
 sample_packing: true

examples/tiny-llama/lora.yml

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

examples/tiny-llama/pretrain.yml

@@ -14,7 +14,7 @@ pretraining_dataset:
   type: pretrain
 dataset_prepared_path:
 val_set_size: 0.0
-output_dir: ./model-out
+output_dir: ./outputs/model-out
 
 sequence_len: 2048
 sample_packing: true