We've also added a new command for fetching `examples` and `deepspeed_configs` to your
local machine. This will come in handy when installing `axolotl` from PyPI.
```bash
# Fetch example YAML files (stores in "examples/" folder)
```shell
# Fetch axolotl examples
axolotl fetch examples
# Fetch deepspeed config files (stores in "deepspeed_configs/" folder)
axolotl fetch deepspeed_configs
# Optionally, specify a destination folder
# Or, specify a custom path
axolotl fetch examples --dest path/to/folder
# Train a model using LoRA
axolotl train examples/llama-3/lora-1b.yml
```
### Legacy Usage
<details>
That's it! Check out our [Getting Started Guide](https://axolotl-ai-cloud.github.io/axolotl/docs/getting-started.html) for a more detailed walkthrough.
<summary>Click to Expand</summary>
## ✨ Key Features
While the Axolotl CLI is the preferred method for interacting with axolotl, we
still support the legacy `-m axolotl.cli.*` usage.
- **Multiple Model Support**: Train various models like LLaMA, Mistral, Mixtral, Pythia, and more
- **Training Methods**: Full fine-tuning, LoRA, QLoRA, and more
- **Easy Configuration**: Simple YAML files to control your training setup
- **Performance Optimizations**: Flash Attention, xformers, multi-GPU training
- **Flexible Dataset Handling**: Use various formats and custom datasets
- **Cloud Ready**: Run on cloud platforms or local hardware
Contributions are welcome! Please see our [Contributing Guide](https://github.com/axolotl-ai-cloud/axolotl/blob/main/.github/CONTRIBUTING.md) for details.
## Badge ❤🏷️
Building something cool with Axolotl? Consider adding a badge to your model card.
```markdown
[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
```
[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
## Sponsors 🤝❤
If you love axolotl, consider sponsoring the project by reaching out directly to [wing@axolotl.ai](mailto:wing@axolotl.ai).
---
- [Modal](https://www.modal.com?utm_source=github&utm_medium=github&utm_campaign=axolotl) Modal lets you run data/AI jobs in the cloud, by just writing a few lines of Python. Customers use Modal to deploy Gen AI models at large scale, fine-tune large language models, run protein folding simulations, and much more.
---
## Contributing 🤝
Please read the [contributing guide](./.github/CONTRIBUTING.md)
Bugs? Please check the [open issues](https://github.com/axolotl-ai-cloud/axolotl/issues/bug) else create a new Issue.
PRs are **greatly welcome**!
Please run the quickstart instructions followed by the below to setup env:
@@ -272,523 +136,16 @@ Thanks to all of our contributors to date. Help drive open source AI progress fo
❌: not supported
❓: untested
## Advanced Setup
## ❤️ Sponsors
### Environment
Thank you to our sponsors who help make Axolotl possible:
#### Docker
- [Modal](https://www.modal.com?utm_source=github&utm_medium=github&utm_campaign=axolotl) - Modal lets you run
jobs in the cloud, by just writing a few lines of Python. Customers use Modal to deploy Gen AI models at large scale,
fine-tune large language models, run protein folding simulations, and much more.
```bash
docker run --gpus '"all"' --rm -it axolotlai/axolotl:main-latest
```
Interested in sponsoring? Contact us at [wing@axolotl.ai](mailto:wing@axolotl.ai)
Or run on the current files for development:
## 📜 License
```sh
docker compose up -d
```
>[!Tip]
> If you want to debug axolotl or prefer to use Docker as your development environment, see the [debugging guide's section on Docker](docs/debugging.qmd#debugging-with-docker).
<details>
<summary>Docker advanced</summary>
A more powerful Docker command to run would be this:
* Prevents memory issues when running e.g. deepspeed (e.g. you could hit SIGBUS/signal 7 error) through `--ipc` and `--ulimit` args.
* Persists the downloaded HF data (models etc.) and your modifications to axolotl code through `--mount`/`-v` args.
* The `--name` argument simply makes it easier to refer to the container in vscode (`Dev Containers: Attach to Running Container...`) or in your terminal.
* The `--privileged` flag gives all capabilities to the container.
* The `--shm-size 10g` argument increases the shared memory size. Use this if you see `exitcode: -7` errors using deepspeed.
[More information on nvidia website](https://docs.nvidia.com/deeplearning/frameworks/user-guide/index.html#setincshmem)
Use a Deeplearning linux OS with cuda and pytorch installed. Then follow instructions on quickstart.
Make sure to run the below to uninstall xla.
```bash
pip uninstall -y torch_xla[tpu]
```
</details>
#### Windows
Please use WSL or Docker!
#### Mac
Use the below instead of the install method in QuickStart.
```
pip3 install --no-build-isolation -e '.'
```
More info: [mac.md](/docs/mac.qmd)
#### Google Colab
Please use this example [notebook](examples/colab-notebooks/colab-axolotl-example.ipynb).
#### Launching on public clouds via SkyPilot
To launch on GPU instances (both on-demand and spot instances) on 7+ clouds (GCP, AWS, Azure, OCI, and more), you can use [SkyPilot](https://skypilot.readthedocs.io/en/latest/index.html):
```bash
pip install "skypilot-nightly[gcp,aws,azure,oci,lambda,kubernetes,ibm,scp]" # choose your clouds
sky check
```
Get the [example YAMLs](https://github.com/skypilot-org/skypilot/tree/master/llm/axolotl) of using Axolotl to finetune `mistralai/Mistral-7B-v0.1`:
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/).
then, simply run the job with `dstack run` command. Append `--spot` option if you want spot instance. `dstack run` command will show you the instance with cheapest price across multi cloud services:
```bash
pip install dstack
HUGGING_FACE_HUB_TOKEN=xxx WANDB_API_KEY=xxx dstack run . -f dstack.yaml # --spot
```
For further and fine-grained use cases, please refer to the official [dstack documents](https://dstack.ai/docs/) and the detailed description of [axolotl example](https://github.com/dstackai/dstack/tree/master/examples/fine-tuning/axolotl) on the official repository.
### Dataset
Axolotl supports a variety of dataset formats. It is recommended to use a JSONL. The schema of the JSONL depends upon the task and the prompt template you wish to use. Instead of a JSONL, you can also use a HuggingFace dataset with columns for each JSONL field.
See [the documentation](https://axolotl-ai-cloud.github.io/axolotl/docs/dataset-formats/) for more information on how to use different dataset formats.
### Config
See [examples](examples) for quick start. It is recommended to duplicate and modify to your needs. The most important options are:
- model
```yaml
base_model: ./llama-7b-hf # local or huggingface repo
```
Note: The code will load the right architecture.
- dataset
```yaml
datasets:
# huggingface repo
- path: vicgalle/alpaca-gpt4
type: alpaca
# huggingface repo with specific configuration/subset
- path: EleutherAI/pile
name: enron_emails
type: completion # format from earlier
field: text # Optional[str] default: text, field to use for completion data
# huggingface repo with multiple named configurations/subsets
> You can also reference a config file that is hosted on a public URL, for example `accelerate launch -m axolotl.cli.train https://yourdomain.com/your_config.yml`
#### Preprocess dataset
You can optionally pre-tokenize dataset with the following before finetuning.
This is recommended for large datasets.
- Set `dataset_prepared_path:` to a local folder for saving and loading pre-tokenized dataset.
- (Optional): Set `push_dataset_to_hub: hf_user/repo` to push it to Huggingface.
- (Optional): Use `--debug` to see preprocessed examples.
```bash
python -m axolotl.cli.preprocess your_config.yml
```
#### Multi-GPU
Below are the options available in axolotl for training with multiple GPUs. Note that DeepSpeed
is the recommended multi-GPU option currently because FSDP may experience
Axolotl supports training with FSDP and QLoRA, see [these docs](docs/fsdp_qlora.qmd) for more information.
##### Weights & Biases Logging
Make sure your `WANDB_API_KEY` environment variable is set (recommended) or you login to wandb with `wandb login`.
- wandb options
```yaml
wandb_mode:
wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:
```
##### Comet Logging
Make sure your `COMET_API_KEY` environment variable is set (recommended) or you login to wandb with `comet login`.
- wandb options
```yaml
use_comet:
comet_api_key:
comet_workspace:
comet_project_name:
comet_experiment_key:
comet_mode:
comet_online:
comet_experiment_config:
```
##### Special Tokens
It is important to have special tokens like delimiters, end-of-sequence, beginning-of-sequence in your tokenizer's vocabulary. This will help you avoid tokenization issues and help your model train better. You can do this in axolotl like this:
```yml
special_tokens:
bos_token: "<s>"
eos_token: "</s>"
unk_token: "<unk>"
tokens: # these are delimiters
- "<|im_start|>"
- "<|im_end|>"
```
When you include these tokens in your axolotl config, axolotl adds these tokens to the tokenizer's vocabulary.
##### Liger Kernel
Liger Kernel: Efficient Triton Kernels for LLM Training
https://github.com/linkedin/Liger-Kernel
Liger (LinkedIn GPU Efficient Runtime) Kernel is a collection of Triton kernels designed specifically for LLM training.
It can effectively increase multi-GPU training throughput by 20% and reduces memory usage by 60%. The Liger Kernel
composes well and is compatible with both FSDP and Deepspeed.
```yaml
plugins:
- axolotl.integrations.liger.LigerPlugin
liger_rope: true
liger_rms_norm: true
liger_glu_activation: true
liger_layer_norm: true
liger_fused_linear_cross_entropy: true
```
### Inference Playground
Axolotl allows you to load your model in an interactive terminal playground for quick experimentation.
The config file is the same config file used for training.
Pass the appropriate flag to the inference command, depending upon what kind of model was trained:
Please use `--sample_packing False` if you have it on and receive the error similar to below:
> RuntimeError: stack expects each tensor to be equal size, but got [1, 32, 1, 128] at entry 0 and [1, 32, 8, 128] at entry 1
### Merge LORA to base
The following command will merge your LORA adapater with your base model. You can optionally pass the argument `--lora_model_dir` to specify the directory where your LORA adapter was saved, otherwhise, this will be inferred from `output_dir` in your axolotl config file. The merged model is saved in the sub-directory `{lora_model_dir}/merged`.
You may need to use the `gpu_memory_limit` and/or `lora_on_cpu` config options to avoid running out of memory. If you still run out of CUDA memory, you can try to merge in system RAM with
although this will be very slow, and using the config options above are recommended instead.
## Common Errors 🧰
See also the [FAQ's](./docs/faq.qmd) and [debugging guide](docs/debugging.qmd).
> If you encounter a 'Cuda out of memory' error, it means your GPU ran out of memory during the training process. Here's how to resolve it:
Please reduce any below
- `micro_batch_size`
- `eval_batch_size`
- `gradient_accumulation_steps`
- `sequence_len`
If it does not help, try running without deepspeed and without accelerate (replace "accelerate launch" with "python") in the command.
Using adamw_bnb_8bit might also save you some memory.
> `failed (exitcode: -9)`
Usually means your system has run out of system memory.
Similarly, you should consider reducing the same settings as when you run out of VRAM.
Additionally, look into upgrading your system RAM which should be simpler than GPU upgrades.
> RuntimeError: expected scalar type Float but found Half
Try set `fp16: true`
> NotImplementedError: No operator found for `memory_efficient_attention_forward` ...
Try to turn off xformers.
> accelerate config missing
It's safe to ignore it.
> NCCL Timeouts during training
See the [NCCL](docs/nccl.qmd) guide.
### Tokenization Mismatch b/w Inference & Training
For many formats, Axolotl constructs prompts by concatenating token ids _after_ tokenizing strings. The reason for concatenating token ids rather than operating on strings is to maintain precise accounting for attention masks.
If you decode a prompt constructed by axolotl, you might see spaces between tokens (or lack thereof) that you do not expect, especially around delimiters and special tokens. When you are starting out with a new format, you should always do the following:
1. Materialize some data using `python -m axolotl.cli.preprocess your_config.yml --debug`, and then decode the first few rows with your model's tokenizer.
2. During inference, right before you pass a tensor of token ids to your model, decode these tokens back into a string.
3. Make sure the inference string from #2 looks **exactly** like the data you fine tuned on from #1, including spaces and new lines. If they aren't the same, adjust your inference server accordingly.
4. As an additional troubleshooting step, you can look at the token ids between 1 and 2 to make sure they are identical.
Having misalignment between your prompts during training and inference can cause models to perform very poorly, so it is worth checking this. See [this blog post](https://hamel.dev/notes/llm/finetuning/05_tokenizer_gotchas.html) for a concrete example.
## Debugging Axolotl
See [this debugging guide](docs/debugging.qmd) for tips on debugging Axolotl, along with an example configuration for debugging with VSCode.
## Need help? 🙋
Join our [Discord server](https://discord.gg/HhrNrHJPRb) where our community members can help you.
Need dedicated support? Please contact us at [✉️wing@axolotl.ai](ailto:wing@axolotl.ai) for dedicated support options.
This project is licensed under the Apache 2.0 License - see the [LICENSE](LICENSE) file for details.
Chat Template strategy uses a jinja2 template that converts a list of messages into a prompt. Support using tokenizer's template, a supported template, or custom jinja2.
The stepwise supervised format is designed for chain-of-thought (COT) reasoning datasets where each example contains multiple completion steps and a preference label for each step.
### ExampleHere's a simple example of a stepwise supervised dataset entry:```json
The stepwise supervised format is designed for chain-of-thought (COT) reasoning
datasets where each example contains multiple completion steps and a preference label
for each step.
### Example
Here's a simple example of a stepwise supervised dataset entry:
```json
{
"prompt": "Which number is larger, 9.8 or 9.11?",
"completions": [
@@ -16,3 +23,4 @@ The stepwise supervised format is designed for chain-of-thought (COT) reasoning
This guide covers advanced training configurations for multi-GPU setups using Axolotl.
## Overview {#sec-overview}
Axolotl supports several methods for multi-GPU training:
- DeepSpeed (recommended)
- FSDP (Fully Sharded Data Parallel)
- FSDP + QLoRA
## DeepSpeed {#sec-deepspeed}
DeepSpeed is the recommended approach for multi-GPU training due to its stability and performance. It provides various optimization levels through ZeRO stages.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Dan Saunders
GitHub