resume from optimizer checkpoint only

* fix: allow trainer builder to use custom jinja chat template

* chore: use get_chat_template_from_config

Co-authored-by: Chirag Jain <jain.chirag925@gmail.com>

* fix: swap imports

---------

Co-authored-by: Chirag Jain <jain.chirag925@gmail.com>

.github/workflows/pypi.yml

@@ -13,10 +13,13 @@ jobs:
     permissions:
       contents: write
     steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+
       - name: Create release
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: gh release create "$GITHUB_REF_NAME" # GITHUB_REF_NAME is the tag name in `on.push.tags` workflows
+        run: gh release create "$GITHUB_REF_NAME" --generate-notes
   pypi-publish:
     name: Upload release to PyPI
     runs-on: ubuntu-latest
@@ -38,7 +41,7 @@ jobs:
       - name: Install dependencies
         run: |
           pip3 install wheel packaging
-          pip3 install -e .
+          pip3 install --no-build-isolation -e .
           pip3 install -r requirements-dev.txt -r requirements-tests.txt
 
       - name: Extract tag name

.github/workflows/tests-nightly.yml

@@ -44,6 +44,11 @@ jobs:
           python-version: ${{ matrix.python_version }}
           cache: 'pip' # caching pip dependencies
 
+      - name: upgrade pip
+        run: |
+          pip3 install --upgrade pip
+          pip3 install --upgrade packaging setuptools wheel
+
       - name: Install PyTorch
         run: |
           pip3 install torch==${{ matrix.pytorch_version }} --index-url https://download.pytorch.org/whl/cpu
@@ -60,11 +65,15 @@ jobs:
         run: |
           pip3 install --upgrade pip
           pip3 install --upgrade packaging
-          pip3 install -U -e .
+          pip3 install --no-build-isolation -U -e .
           python scripts/unsloth_install.py | sh
           python scripts/cutcrossentropy_install.py | sh
           pip3 install -r requirements-dev.txt -r requirements-tests.txt
 
+      - name: Make sure PyTorch version wasn't clobbered
+        run: |
+          python -c "import torch; assert '${{ matrix.pytorch_version }}' in torch.__version__"
+
       - name: Ensure axolotl CLI was installed
         run: |
           axolotl --help

.github/workflows/tests.yml

@@ -78,19 +78,23 @@ jobs:
       - name: Install dependencies
         run: |
           pip3 show torch
-          pip3 install -U -e .
+          pip3 install --no-build-isolation -U -e .
           python scripts/unsloth_install.py | sh
           python scripts/cutcrossentropy_install.py | sh
           pip3 install -r requirements-dev.txt -r requirements-tests.txt
 
+      - name: Make sure PyTorch version wasn't clobbered
+        run: |
+          python -c "import torch; assert '${{ matrix.pytorch_version }}' in torch.__version__"
+
       - name: Ensure axolotl CLI was installed
         run: |
           axolotl --help
 
       - name: Run tests
         run: |
-          pytest -n8 --dist loadfile --ignore=tests/e2e/ --ignore=tests/patched/ tests/
-          pytest tests/patched/
+          pytest -v -n8 --dist loadfile --ignore=tests/e2e/ --ignore=tests/patched/ tests/
+          pytest -v tests/patched/
 
       - name: cleanup pip cache
         run: |
@@ -120,7 +124,7 @@ jobs:
       - name: upgrade pip
         run: |
           pip3 install --upgrade pip
-          pip3 install --upgrade packaging setuptools wheel
+          pip3 install --upgrade packaging setuptools setuptools_scm build wheel
 
       - name: Install PyTorch
         run: |
@@ -129,20 +133,24 @@ jobs:
       - name: Install dependencies
         run: |
           pip3 show torch
-          python3 setup.py sdist
-          pip3 install dist/axolotl*.tar.gz
+          python -m build --no-isolation --sdist
+          pip3 install --no-build-isolation dist/axolotl*.tar.gz
           python scripts/unsloth_install.py | sh
           python scripts/cutcrossentropy_install.py | sh
           pip3 install -r requirements-dev.txt -r requirements-tests.txt
 
+      - name: Make sure PyTorch version wasn't clobbered
+        run: |
+          python -c "import torch; assert '${{ matrix.pytorch_version }}' in torch.__version__"
+
       - name: Ensure axolotl CLI was installed
         run: |
           axolotl --help
 
       - name: Run tests
         run: |
-          pytest -n8 --dist loadfile --ignore=tests/e2e/ --ignore=tests/patched/ tests/
-          pytest tests/patched/
+          pytest -v -n8 --dist loadfile --ignore=tests/e2e/ --ignore=tests/patched/ tests/
+          pytest -v tests/patched/
 
       - name: cleanup pip cache
         run: |

.gitignore

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

MANIFEST.in

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

README.md

@@ -10,9 +10,13 @@
     <img src="https://img.shields.io/github/license/axolotl-ai-cloud/axolotl.svg?color=blue" alt="GitHub License">
     <img src="https://github.com/axolotl-ai-cloud/axolotl/actions/workflows/tests.yml/badge.svg" alt="tests">
     <a href="https://github.com/axolotl-ai-cloud/axolotl/releases"><img src="https://img.shields.io/github/release/axolotl-ai-cloud/axolotl.svg" alt="Releases"></a>
+    <br/>
+    <a href="https://github.com/axolotl-ai-cloud/axolotl/graphs/contributors"><img src="https://img.shields.io/github/contributors-anon/axolotl-ai-cloud/axolotl?color=yellow&style=flat-square" alt="contributors" style="height: 20px;"></a>
     <img src="https://img.shields.io/github/stars/axolotl-ai-cloud/axolotl" alt="GitHub Repo stars">
-</p>
-<p align="center">
+    <br/>
+    <a href="https://discord.com/invite/HhrNrHJPRb"><img src="https://img.shields.io/badge/discord-7289da.svg?style=flat-square&logo=discord" alt="discord" style="height: 20px;"></a>
+    <a href="https://twitter.com/axolotl_ai"><img src="https://img.shields.io/twitter/follow/axolotl_ai?style=social" alt="twitter" style="height: 20px;"></a>
+    <br/>
     <img src="https://github.com/axolotl-ai-cloud/axolotl/actions/workflows/tests-nightly.yml/badge.svg" alt="tests-nightly">
     <img src="https://github.com/axolotl-ai-cloud/axolotl/actions/workflows/multi-gpu-e2e.yml/badge.svg" alt="multigpu-semi-weekly tests">
 </p>
@@ -42,7 +46,8 @@ Features:
 - [Axolotl](#axolotl)
   - [Table of Contents](#table-of-contents)
   - [Quickstart ⚡](#quickstart-)
-    - [Usage](#usage)
+    - [Edge Builds](#edge-builds-)
+    - [Axolotl CLI Usage](#axolotl-cli-usage)
   - [Badge ❤🏷️](#badge-️)
   - [Contributing 🤝](#contributing-)
   - [Sponsors 🤝❤](#sponsors-)
@@ -107,58 +112,49 @@ Get started with Axolotl in just a few steps! This quickstart guide will walk yo
 **Requirements**: *Nvidia* GPU (Ampere architecture or newer for `bf16` and Flash Attention) or *AMD* GPU, Python >=3.10 and PyTorch >=2.3.1.
 
 ```bash
-git clone https://github.com/axolotl-ai-cloud/axolotl
+pip3 install --no-build-isolation axolotl[flash-attn,deepspeed]
+
+# download examples and optionally deepspeed configs to the local path
+axolotl fetch examples
+axolotl fetch deepspeed_configs  # OPTIONAL
+
+# finetune using lora
+axolotl train examples/llama-3/lora-1b.yml
+```
+
+### Edge Builds 🏎️
+
+If you're looking for the latest features and updates between releases, you'll need to install
+from source.
+
+```bash
+git clone https://github.com/axolotl-ai-cloud/axolotl.git
 cd axolotl
-
 pip3 install packaging ninja
-pip3 install -e '.[flash-attn,deepspeed]'
+pip3 install --no-build-isolation -e '.[flash-attn,deepspeed]'
 ```
 
-### Usage
-```bash
-# preprocess datasets - optional but recommended
-CUDA_VISIBLE_DEVICES="0" python -m axolotl.cli.preprocess examples/openllama-3b/lora.yml
-
-# finetune lora
-accelerate launch -m axolotl.cli.train examples/openllama-3b/lora.yml
-
-# inference
-accelerate launch -m axolotl.cli.inference examples/openllama-3b/lora.yml \
-    --lora_model_dir="./outputs/lora-out"
-
-# gradio
-accelerate launch -m axolotl.cli.inference examples/openllama-3b/lora.yml \
-    --lora_model_dir="./outputs/lora-out" --gradio
-
-# remote yaml files - the yaml config can be hosted on a public URL
-# Note: the yaml config must directly link to the **raw** yaml
-accelerate launch -m axolotl.cli.train https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/examples/openllama-3b/lora.yml
-```
-
-### Axolotl CLI
-
-If you've installed this package using `pip` from source, we now support a new, more
-streamlined CLI using [click](https://click.palletsprojects.com/en/stable/). Rewriting
-the above commands:
+### Axolotl CLI Usage
+We now support a new, more streamlined CLI using [click](https://click.palletsprojects.com/en/stable/).
 
 ```bash
 # preprocess datasets - optional but recommended
-CUDA_VISIBLE_DEVICES="0" axolotl preprocess examples/openllama-3b/lora.yml
+CUDA_VISIBLE_DEVICES="0" axolotl preprocess examples/llama-3/lora-1b.yml
 
 # finetune lora
-axolotl train examples/openllama-3b/lora.yml
+axolotl train examples/llama-3/lora-1b.yml
 
 # inference
-axolotl inference examples/openllama-3b/lora.yml \
+axolotl inference examples/llama-3/lora-1b.yml \
     --lora-model-dir="./outputs/lora-out"
 
 # gradio
-axolotl inference examples/openllama-3b/lora.yml \
+axolotl inference examples/llama-3/lora-1b.yml \
     --lora-model-dir="./outputs/lora-out" --gradio
 
 # remote yaml files - the yaml config can be hosted on a public URL
 # Note: the yaml config must directly link to the **raw** yaml
-axolotl train https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/examples/openllama-3b/lora.yml
+axolotl train https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/examples/llama-3/lora-1b.yml
 ```
 
 We've also added a new command for fetching `examples` and `deepspeed_configs` to your
@@ -175,6 +171,36 @@ axolotl fetch deepspeed_configs
 axolotl fetch examples --dest path/to/folder
 ```
 
+### Legacy Usage
+<details>
+
+<summary>Click to Expand</summary>
+
+While the Axolotl CLI is the preferred method for interacting with axolotl, we
+still support the legacy `-m axolotl.cli.*` usage.
+
+```bash
+# preprocess datasets - optional but recommended
+CUDA_VISIBLE_DEVICES="0" python -m axolotl.cli.preprocess examples/llama-3/lora-1b.yml
+
+# finetune lora
+accelerate launch -m axolotl.cli.train examples/llama-3/lora-1b.yml
+
+# inference
+accelerate launch -m axolotl.cli.inference examples/llama-3/lora-1b.yml \
+    --lora_model_dir="./outputs/lora-out"
+
+# gradio
+accelerate launch -m axolotl.cli.inference examples/llama-3/lora-1b.yml \
+    --lora_model_dir="./outputs/lora-out" --gradio
+
+# remote yaml files - the yaml config can be hosted on a public URL
+# Note: the yaml config must directly link to the **raw** yaml
+accelerate launch -m axolotl.cli.train https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/examples/llama-3/lora-1b.yml
+```
+
+</details>
+
 ## Badge ❤🏷️
 
 Building something cool with Axolotl? Consider adding a badge to your model card.
@@ -294,7 +320,7 @@ docker run --privileged --gpus '"all"' --shm-size 10g --rm -it --name axolotl --
   3. Install Axolotl along with python dependencies
         ```bash
         pip3 install packaging
-        pip3 install -e '.[flash-attn,deepspeed]'
+        pip3 install --no-build-isolation -e '.[flash-attn,deepspeed]'
         ```
   4. (Optional) Login to Huggingface to use gated models/datasets.
         ```bash
@@ -373,7 +399,7 @@ Please use WSL or Docker!
 
 Use the below instead of the install method in QuickStart.
 ```
-pip3 install -e '.'
+pip3 install --no-build-isolation -e '.'
 ```
 More info: [mac.md](/docs/mac.qmd)
 

cicd/Dockerfile.jinja

@@ -31,9 +31,9 @@ RUN if [ "$NIGHTLY_BUILD" = "true" ] ; then \
     fi
 
 RUN if [ "$AXOLOTL_EXTRAS" != "" ] ; then \
-        pip install -e .[deepspeed,flash-attn,optimizers,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
+        pip install --no-build-isolation -e .[deepspeed,flash-attn,optimizers,$AXOLOTL_EXTRAS] $AXOLOTL_ARGS; \
     else \
-        pip install -e .[deepspeed,flash-attn,optimizers] $AXOLOTL_ARGS; \
+        pip install --no-build-isolation -e .[deepspeed,flash-attn,optimizers] $AXOLOTL_ARGS; \
     fi
 
 RUN python scripts/unsloth_install.py | sh

cicd/cicd.sh

@@ -1,7 +1,10 @@
 #!/bin/bash
 set -e
 
+python -c "import torch; assert '$PYTORCH_VERSION' in torch.__version__"
+
 pytest -v --durations=10 -n8 --ignore=tests/e2e/ --ignore=tests/patched/ /workspace/axolotl/tests/
-pytest -v --durations=10 -n1 --dist loadfile /workspace/axolotl/tests/patched/
-pytest -v --durations=10 -n1 --dist loadfile /workspace/axolotl/tests/e2e/patched/ /workspace/axolotl/tests/e2e/integrations/
+# pytest -v --durations=10 -n8 --dist loadfile /workspace/axolotl/tests/patched/
+pytest -v --durations=10 /workspace/axolotl/tests/e2e/patched/
+pytest -v --durations=10 /workspace/axolotl/tests/e2e/integrations/
 pytest -v --durations=10 --ignore=tests/e2e/patched/ --ignore=tests/e2e/multigpu/ --ignore=tests/e2e/integrations/ /workspace/axolotl/tests/e2e/

deepspeed_configs/zero1_torch_compile.json

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

docker/Dockerfile

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

docker/Dockerfile-base

@@ -16,7 +16,7 @@ ENV PYTHON_VERSION=$PYTHON_VERSION
 ENV TORCH_CUDA_ARCH_LIST=$TORCH_CUDA_ARCH_LIST
 
 RUN apt-get update \
-    && apt-get install -y wget git build-essential ninja-build git-lfs libaio-dev && rm -rf /var/lib/apt/lists/* \
+    && apt-get install -y wget git build-essential ninja-build git-lfs libaio-dev pkg-config && rm -rf /var/lib/apt/lists/* \
     && wget \
     https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh \
     && mkdir /root/.conda \

docker/Dockerfile-cloud

@@ -2,7 +2,7 @@ ARG BASE_TAG=main
 FROM axolotlai/axolotl:$BASE_TAG
 
 ENV HF_DATASETS_CACHE="/workspace/data/huggingface-cache/datasets"
-ENV HUGGINGFACE_HUB_CACHE="/workspace/data/huggingface-cache/hub"
+ENV HF_HUB_CACHE="/workspace/data/huggingface-cache/hub"
 ENV HF_HOME="/workspace/data/huggingface-cache/hub"
 ENV HF_HUB_ENABLE_HF_TRANSFER="1"
 

docker/Dockerfile-cloud-no-tmux

@@ -2,7 +2,7 @@ ARG BASE_TAG=main
 FROM axolotlai/axolotl:$BASE_TAG
 
 ENV HF_DATASETS_CACHE="/workspace/data/huggingface-cache/datasets"
-ENV HUGGINGFACE_HUB_CACHE="/workspace/data/huggingface-cache/hub"
+ENV HF_HUB_CACHE="/workspace/data/huggingface-cache/hub"
 ENV HF_HOME="/workspace/data/huggingface-cache/hub"
 ENV HF_HUB_ENABLE_HF_TRANSFER="1"
 

docker/Dockerfile-tests

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

docs/amd_hpc.qmd

@@ -52,7 +52,7 @@ export GPU_ARCHS="gfx90a"
 cd flash-attention
 export PYTHON_SITE_PACKAGES=$(python -c 'import site; print(site.getsitepackages()[0])')
 patch "${PYTHON_SITE_PACKAGES}/torch/utils/hipify/hipify_python.py" hipify_patch.patch
-pip install .
+pip install --no-build-isolation .
 ```
 
 ### 6. Install Axolotl
@@ -63,7 +63,7 @@ Clone and install Axolotl:
 git clone https://github.com/axolotl-ai-cloud/axolotl
 cd axolotl
 pip install packaging ninja
-pip install -e .
+pip install --no-build-isolation -e .
 ```
 
 ### 7. Apply xformers Workaround

docs/config.qmd

@@ -127,34 +127,40 @@ datasets:
     # - tokenizer_default_fallback_*: where * is the name of the chat template to fallback to if the tokenizer does not have a chat template else default to tokenizer. E.g. tokenizer_default_fallback_chatml.
     # - jinja: Uses a custom jinja template for the chat template. The custom jinja template should be provided in the chat_template_jinja field.
     chat_template: tokenizer_default
-    # Custom jinja template for chat template. This will be only used if `chat_template` is set to `jinja` or empty (in which case chat_template is automatically set to `jinja`).
+
+    # Custom jinja chat template. Used only if `chat_template: jinja` or empty.
     chat_template_jinja:
-    # The key in the data example that contains the messages. Default is "messages".
+
+    # Key containing the messages (default: "messages")
     field_messages: messages
-    # The key in the message turn that contains the role. Default is "role".
+    # Key for role in each message (default: "role")
     message_field_role: role
-    # The key in the message turn that contains the content. Default is "content".
+    # Key for content in each message (default:  "content")
     message_field_content: content
-    # Optional[Dict[str, List]]. Roles mapping for the messages.
+
+    # Optional[Dict[str, List]]. Roles mapping in the messages. The default is:
     roles:
       user: ["human", "user"]
-      assistant: ["gpt", "assistant", "ai"]
+      assistant: ["gpt", "assistant"]
       system: ["system"]
+      tool: ["tool"]
 
-    ## NOTE: Leaving the below empty will default to using the simple legacy tokenization strategy where only last message is trained on.
+    # IMPORTANT: The following fields determine which parts of the conversation to train on.
+    # Priority order: message_field_training > message_field_training_detail > train_on_inputs or role in roles_to_train
+    # See examples at `docs/dataset-formats/conversation.qmd`
+    # Note: If the below 4 fields are empty, defaults to training only on the last message.
 
     # Optional[List[str]]. Roles to train on. The tokens from these roles will be considered for the loss.
-    roles_to_train: ["gpt", "assistant"]
+    roles_to_train: ["assistant"]  # default
     # Optional[str]. Which EOS tokens to train on in the conversation. Possible values are:
     # - all: train on all EOS tokens
-    # - turn: train on the EOS token at the end of each trainable turn
+    # - turn (default): train on the EOS token at the end of each trainable turn
     # - last: train on the last EOS token in the conversation
     train_on_eos: last
     # The key in the message turn that indicates via boolean whether tokens of a turn should be considered for training. Useful to selectively train on certain turns besides the `roles_to_train`.
     message_field_training: training
     # The key in the message turn that contains the training details. Useful to selectively train on certain tokens in a turn.
     # The value of the key is a List[Dict] containing `begin_offset` (start character index in content), `end_offset` (end character index in content), and `train` (boolean whether to train).
-    # See example at `docs/dataset-formats/conversation.qmd`
     message_field_training_detail: train_detail
 
 
@@ -239,6 +245,9 @@ 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
 
+# Use batch flattening for speedups when not using sample_packing
+batch_flattening:
+
 # Passed through to transformers when loading the model when launched without accelerate
 # Use `sequential` when training w/ model parallelism to limit memory
 device_map:
@@ -331,7 +340,8 @@ comet_experiment_config: # Dictionary for additional configuration settings, see
 output_dir: ./completed-model
 
 # Whether to use torch.compile and which backend to use
-torch_compile:  # bool
+# setting to `auto` will enable torch compile when torch>=2.5.1
+torch_compile:  # Optional[Union[Literal["auto"], bool]]
 torch_compile_backend:  # Optional[str]
 
 # Training hyperparameters
@@ -363,6 +373,10 @@ eval_table_size: # Approximate number of predictions sent to wandb depending on
 eval_max_new_tokens: # Total number of tokens generated for predictions sent to wandb. Default is 128
 eval_causal_lm_metrics: # HF evaluate metrics used during evaluation. Default is ["sacrebleu", "comet", "ter", "chrf", "perplexity"]
 
+profiler_steps: # enable the pytorch profiler to capture the first N steps of training to the output_dir.
+                # see https://pytorch.org/blog/understanding-gpu-memory-1/ for more information
+                # snapshots can be visualized @ https://pytorch.org/memory_viz
+
 loss_watchdog_threshold: # High loss value, indicating the learning has broken down (a good estimate is ~2 times the loss at the start of training)
 loss_watchdog_patience: # Number of high-loss steps in a row before the trainer aborts (default: 3)
 

docs/dataset-formats/conversation.qmd

@@ -68,6 +68,8 @@ We recommend checking the below examples for other usecases.
 datasets:
   - path: ...
     type: chat_template
+    roles_to_train:
+    train_on_eos:
 ```
 
 2. Using the `gemma` chat template to override the tokenizer_config.json's chat template on OpenAI messages format, training on all assistant messages.
@@ -77,7 +79,7 @@ chat_template: gemma # this overwrites the tokenizer's chat_template
 datasets:
   - path: ...
     type: chat_template
-    roles_to_train: ["assistant"]
+    roles_to_train: ["assistant"]  # default value
 ```
 
 3. Using the tokenizer_config.json's chat template or `chatml` as fallback if the former's chat template does not exist, on OpenAI messages format, training on all assistant messages.
@@ -87,7 +89,6 @@ chat_template: tokenizer_default_fallback_chatml # this overwrites the tokenizer
 datasets:
   - path: ...
     type: chat_template
-    roles_to_train: ["assistant"]
 ```
 
 4. Using a custom jinja template on OpenAI messages format, training on all assistant messages.
@@ -99,7 +100,6 @@ chat_template_jinja: "{{ bos_token }}{% for message in messages %}{% if (message
 datasets:
   - path: ...
     type: chat_template
-    roles_to_train: ["assistant"]
 ```
 
 5. (Advanced) Using fine-grained control over tokens and turns to train in a conversation

docs/debugging.qmd

@@ -71,7 +71,7 @@ Make sure you have an [editable install](https://setuptools.pypa.io/en/latest/us
 
 ```bash
 pip3 install packaging
-pip3 install -e '.[flash-attn,deepspeed]'
+pip3 install --no-build-isolation -e '.[flash-attn,deepspeed]'
 ```
 
 #### Remote Hosts
@@ -212,7 +212,7 @@ You will now be in the container.  Next, perform an editable install of Axolotl:
 
 ```bash
 pip3 install packaging
-pip3 install -e '.[flash-attn,deepspeed]'
+pip3 install --no-build-isolation -e '.[flash-attn,deepspeed]'
 ```
 
 ### Attach To Container

docs/rlhf.qmd

@@ -52,6 +52,26 @@ datasets:
     type: chat_template.argilla
 ```
 
+
+#### KTO
+
+```yaml
+rl: kto
+rl_beta: 0.5
+kto_desirable_weight: 0.2
+
+remove_unused_columns: false
+
+datasets:
+  - path: argilla/ultrafeedback-binarized-preferences-cleaned-kto
+    type: llama3.ultra
+    split: train
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: true
+```
+
 #### Using local dataset files
 ```yaml
 datasets:

examples/cerebras/btlm-ft.yml

@@ -1,6 +1,10 @@
 base_model: cerebras/btlm-3b-8k-base
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: GPT2Tokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 tokenizer_use_fast: true
 tokenizer_legacy: true

examples/cerebras/qlora.yml

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

examples/dbrx/16bit-lora.yaml

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

examples/dbrx/8bit-lora.yaml

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

examples/dbrx/fft-ds-zero3.yaml

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

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

@@ -1,4 +1,6 @@
 base_model: deepseek-ai/DeepSeek-V2-Lite
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 trust_remote_code: true
 
 load_in_8bit: false

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

@@ -1,4 +1,7 @@
 base_model: axolotl-quants/DeepSeek-V2.5-bnb-nf4-bf16
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: false

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

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

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

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

examples/falcon/config-7b.yml

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

examples/gemma/qlora.yml

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

examples/gemma2/qlora.yml

@@ -1,6 +1,9 @@
 base_model: google/gemma-2-9b
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true

examples/gemma2/reward-model.yaml

@@ -1,6 +1,9 @@
 base_model: google/gemma-2-2b
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForSequenceClassification
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: false

examples/gptj/qlora.yml

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

examples/jamba/qlora.yaml

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

examples/jamba/qlora_deepspeed.yaml

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

examples/jamba/qlora_fsdp_large.yaml

@@ -1,5 +1,8 @@
 base_model: ai21labs/AI21-Jamba-1.5-Large
+# optionally might have model_type or tokenizer_type
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_4bit: true
 strict: false

examples/jeopardy-bot/config.yml

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

examples/llama-2/fft_optimized.yml

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

examples/llama-2/gptq-lora.yml

@@ -1,8 +1,13 @@
 base_model: TheBloke/Llama-2-7B-GPTQ
-gptq: true
-gptq_disable_exllama: true
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+gptq: true
+gptq_disable_exllama: true
+
 tokenizer_use_fast: true
 tokenizer_legacy: true
 load_in_8bit: false

examples/llama-2/lisa.yml

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

examples/llama-2/loftq.yml

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

examples/llama-2/lora.yml

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

examples/llama-2/qlora-fsdp.yml

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

examples/llama-2/qlora.yml

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

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

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

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

@@ -1,4 +1,6 @@
 base_model: NousResearch/Meta-Llama-3.1-8B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 plugins:
   - axolotl.integrations.liger.LigerPlugin

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

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

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

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

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

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

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

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

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

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

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

@@ -0,0 +1,76 @@
+base_model: NousResearch/Llama-3.2-1B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: false
+strict: false
+
+datasets:
+  - path: teknium/GPT4-LLM-Cleaned
+    type: alpaca
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.1
+output_dir: ./outputs/lora-out
+
+adapter: lora
+lora_model_dir:
+
+sequence_len: 2048
+sample_packing: true
+eval_sample_packing: true
+pad_to_sequence_len: true
+
+lora_r: 16
+lora_alpha: 32
+lora_dropout: 0.05
+lora_fan_in_fan_out:
+lora_target_modules:
+  - gate_proj
+  - down_proj
+  - up_proj
+  - q_proj
+  - v_proj
+  - k_proj
+  - o_proj
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 2
+micro_batch_size: 2
+num_epochs: 1
+optimizer: adamw_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: false
+
+gradient_checkpointing: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+loss_watchdog_threshold: 5.0
+loss_watchdog_patience: 3
+
+warmup_steps: 10
+evals_per_epoch: 4
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:
+  pad_token: "<|end_of_text|>"

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

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

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

@@ -0,0 +1,77 @@
+base_model: meta-llama/Llama-3.2-1B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
+load_in_8bit: false
+load_in_4bit: true
+strict: false
+
+rl: kto
+rl_beta: 0.5
+kto_desirable_weight: 0.2
+
+datasets:
+  - path: argilla/ultrafeedback-binarized-preferences-cleaned-kto
+    type: llama3.ultra
+    split: train
+dataset_prepared_path: last_run_prepared
+val_set_size: 0.0
+output_dir: ./outputs/qlora-out
+
+remove_unused_columns: false
+
+adapter: qlora
+lora_model_dir:
+
+sequence_len: 2048
+sample_packing: false  # not supported with kto
+eval_sample_packing: false
+pad_to_sequence_len: false
+
+lora_r: 32
+lora_alpha: 64
+lora_dropout: 0.05
+lora_target_linear: true
+lora_fan_in_fan_out:
+
+wandb_project:
+wandb_entity:
+wandb_watch:
+wandb_name:
+wandb_log_model:
+
+gradient_accumulation_steps: 1
+micro_batch_size: 2
+num_epochs: 1
+optimizer: adamw_8bit
+lr_scheduler: cosine
+learning_rate: 0.0002
+
+train_on_inputs: false
+group_by_length: false
+bf16: auto
+fp16:
+tf32: true
+
+gradient_checkpointing: true
+gradient_checkpointing_kwargs:
+  use_reentrant: true
+early_stopping_patience:
+resume_from_checkpoint:
+local_rank:
+logging_steps: 1
+xformers_attention:
+flash_attention: true
+
+warmup_steps: 20
+evals_per_epoch: 4
+eval_table_size:
+eval_max_new_tokens: 128
+saves_per_epoch: 1
+debug:
+deepspeed:
+weight_decay: 0.0
+fsdp:
+fsdp_config:
+special_tokens:
+  pad_token: "<|end_of_text|>"

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

@@ -1,4 +1,6 @@
-base_model: meta-llama/Llama-3.2-1B
+base_model: NousResearch/Llama-3.2-1B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true
@@ -22,7 +24,6 @@ pad_to_sequence_len: true
 lora_r: 32
 lora_alpha: 16
 lora_dropout: 0.05
-lora_target_linear: true
 lora_fan_in_fan_out:
 lora_target_modules:
   - gate_proj

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

@@ -1,5 +1,8 @@
 base_model: hugging-quants/Meta-Llama-3.1-405B-BNB-NF4-BF16
+# optionally might have model_type or tokenizer_type
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_4bit: true
 strict: false

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

@@ -1,6 +1,9 @@
 base_model: casperhansen/llama-3-70b-fp16
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: AutoTokenizer  # PreTrainedTokenizerFast
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true

examples/llama-3/qlora.yml

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

examples/mamba/config.yml

@@ -1,7 +1,10 @@
 base_model: state-spaces/mamba-2.8b
+# optionally might have model_type or tokenizer_type or tokenizer_config
 model_type: MambaLMHeadModel
 tokenizer_type: AutoTokenizer
 tokenizer_config: EleutherAI/gpt-neox-20b
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: false

examples/mistral/bigstral-ds-zero3.yaml

@@ -1,6 +1,10 @@
 base_model: mistral-community/Mixtral-8x22B-v0.1
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: false

examples/mistral/config.yml

@@ -1,6 +1,9 @@
 base_model: mistralai/Mistral-7B-v0.1
+# optionally might have model_type or tokenizer_type
 model_type: MistralForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: false

examples/mistral/lora-mps.yml

@@ -1,6 +1,9 @@
 base_model: mistralai/Mistral-7B-v0.1
+# optionally might have model_type or tokenizer_type
 model_type: MistralForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: false

examples/mistral/lora.yml

@@ -1,6 +1,9 @@
 base_model: mistralai/Mistral-7B-v0.1
+# optionally might have model_type or tokenizer_type
 model_type: MistralForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: true
 load_in_4bit: false

examples/mistral/mistral-dpo-qlora.yml

@@ -4,8 +4,11 @@
 #face problems with the special tokens.
 
 base_model: mistralai/Mistral-7B-Instruct-v0.2
+# optionally might have model_type or tokenizer_type
 model_type: MistralForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true

examples/mistral/mistral-qlora-fsdp.yml

@@ -1,6 +1,10 @@
 base_model: mistralai/Mixtral-8x7B-v0.1
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: false

examples/mistral/mistral-qlora-orpo.yml

@@ -1,6 +1,9 @@
 base_model: mistralai/Mistral-7B-v0.1
+# optionally might have model_type or tokenizer_type
 model_type: MistralForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true

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

@@ -1,6 +1,9 @@
 base_model: mistral-community/Mixtral-8x22B-v0.1
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true

examples/mistral/mixtral-qlora-fsdp.yml

@@ -1,6 +1,10 @@
 base_model: mistralai/Mixtral-8x7B-v0.1
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: false

examples/mistral/mixtral.yml

@@ -1,6 +1,10 @@
 base_model: mistralai/Mixtral-8x7B-v0.1
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: false

examples/mistral/mixtral_22.yml

@@ -1,6 +1,10 @@
 base_model: mistral-community/Mixtral-8x22B-v0.1
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: false

examples/mistral/qlora.yml

@@ -1,6 +1,9 @@
 base_model: mistralai/Mistral-7B-v0.1
+# optionally might have model_type or tokenizer_type
 model_type: MistralForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true

examples/mpt-7b/config.yml

@@ -1,5 +1,9 @@
 base_model: mosaicml/mpt-7b
+# optionally might have model_type or tokenizer_type
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true  # required for mpt as their model class is not merged into transformers yet
 load_in_8bit: false
 datasets:

examples/openllama-3b/config.yml

@@ -1,6 +1,10 @@
 base_model: openlm-research/open_llama_3b_v2
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 load_in_8bit: false
 load_in_4bit: false
 strict: false

examples/openllama-3b/lora.yml

@@ -1,6 +1,10 @@
 base_model: openlm-research/open_llama_3b_v2
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 load_in_8bit: true
 load_in_4bit: false
 strict: false

examples/openllama-3b/qlora.yml

@@ -1,6 +1,10 @@
 base_model: openlm-research/open_llama_3b_v2
+# optionally might have model_type or tokenizer_type
 model_type: LlamaForCausalLM
 tokenizer_type: LlamaTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 load_in_8bit: false
 load_in_4bit: true
 strict: false

examples/phi/lora-3.5.yaml

@@ -1,6 +1,9 @@
 base_model: microsoft/Phi-3.5-mini-instruct
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: true
 load_in_4bit: false

examples/phi/phi-ft.yml

@@ -1,6 +1,9 @@
 base_model: microsoft/phi-1_5
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: false

examples/phi/phi-qlora.yml

@@ -1,6 +1,9 @@
 base_model: microsoft/phi-1_5
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: true

examples/phi/phi2-ft.yml

@@ -1,6 +1,9 @@
 base_model: microsoft/phi-2
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: false

examples/phi/phi3-ft-fsdp.yml

@@ -1,6 +1,9 @@
 base_model: microsoft/Phi-3-mini-4k-instruct
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 load_in_8bit: false
 load_in_4bit: false

examples/phi/phi3-ft.yml

@@ -1,7 +1,11 @@
 base_model: microsoft/Phi-3-mini-4k-instruct
+# optionally might have model_type or tokenizer_type
 trust_remote_code: true
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 chat_template: phi_3
 
 load_in_8bit: false

examples/pythia-12b/config.yml

@@ -1,7 +1,11 @@
 base_model: EleutherAI/pythia-12b-deduped
 base_model_ignore_patterns: pytorch*  # prefer safetensors
+# optionally might have model_type or tokenizer_type
 model_type: GPTNeoXForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 load_in_8bit: false
 load_in_4bit: false
 gptq: false

examples/pythia/lora.yml

@@ -1,4 +1,7 @@
 base_model: EleutherAI/pythia-1.4b-deduped
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 load_in_8bit: true
 datasets:
   - path: teknium/GPT4-LLM-Cleaned

examples/qwen/lora.yml

@@ -1,6 +1,9 @@
 base_model: Qwen/Qwen-7B
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 trust_remote_code: true
 

examples/qwen/qlora.yml

@@ -1,6 +1,9 @@
 base_model: Qwen/Qwen-7B
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 trust_remote_code: true
 

examples/qwen/qwen2-moe-lora.yaml

@@ -1,4 +1,7 @@
 base_model: Qwen/Qwen1.5-MoE-A2.7B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: false

examples/qwen/qwen2-moe-qlora.yaml

@@ -1,4 +1,7 @@
 base_model: Qwen/Qwen1.5-MoE-A2.7B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: false

examples/qwen2/dpo.yaml

@@ -1,4 +1,6 @@
 base_model: Qwen/Qwen2.5-0.5B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
 
 strict: false
 

examples/qwen2/qlora-fsdp.yaml

@@ -1,4 +1,7 @@
 base_model: Qwen/Qwen2-7B
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: false

examples/redpajama/config-3b.yml

@@ -1,6 +1,10 @@
 base_model: togethercomputer/RedPajama-INCITE-Chat-3B-v1
+# optionally might have model_type or tokenizer_type
 model_type: GPTNeoXForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code:
 load_in_8bit: false
 datasets:

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

@@ -1,4 +1,7 @@
 base_model: replit/replit-code-v1-3b
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 load_in_8bit: false
 datasets:

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

@@ -1,6 +1,10 @@
 base_model: stabilityai/stablelm-2-1_6b
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: false

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

@@ -1,6 +1,10 @@
 base_model: stabilityai/stablelm-2-1_6b
+# optionally might have model_type or tokenizer_type
 model_type: AutoModelForCausalLM
 tokenizer_type: AutoTokenizer
+# Automatically upload checkpoint and final model to HF
+# hub_model_id: username/custom_model_name
+
 trust_remote_code: true
 
 load_in_8bit: true