Skip to content

Latest commit

 

History

History
66 lines (51 loc) · 1.94 KB

fine-tuning.md

File metadata and controls

66 lines (51 loc) · 1.94 KB

Axolotl: https://github.com/OpenAccess-AI-Collective/axolotl

  • GCP:
pip install "skypilot-nightly[gcp]"
pip install google-api-python-client
conda install -c conda-forge google-cloud-sdk -y
conda activate
sky check
gcloud init
gcloud auth application-default login    
git clone https://github.com/skypilot-org/skypilot.git
cd skypilot/llm/axolotl
# On-demand
HF_TOKEN=xx sky launch axolotl.yaml --env HF_TOKEN

FastChat: https://github.com/lm-sys/FastChat/blob/main/docs/training.md

  • Apple M1
pip install -U mlx
git clone https://github.com/lm-sys/FastChat.git
cd FastChat
brew install rust 
brew install cmake
pip3 install --upgrade pip
pip3 install -e ".[model_worker,webui]"
python3 -m fastchat.serve.cli --model-path lmsys/vicuna-7b-v1.5 --device mps --load-8bit
pip install "mlx-lm>=0.0.6"
python3 -m fastchat.serve.controller
python3 -m fastchat.serve.gradio_web_server
python3 -m fastchat.serve.mlx_worker --model-path mistralai/Mistral-7B-v0.1

pip3 install -e ".[train]"

LM Buddy: https://github.com/mozilla-ai/lm-buddy

Levanter: https://levanter.readthedocs.io/en/latest/Fine-Tuning/

"While Levanter's main focus is pretraining, we can also use it for fine-tuning. As an example, we'll show how to reproduce Stanford Alpaca, using Levanter and either Llama 1 or Llama 2 7B. The script we develop will be designed for Alpaca, defaulting to using its dataset and prompts, but it should work for any single-turn instruction-following task."

  • Apple M1
conda create -n levanter-metal python=3.10 pip
conda activate levanter-metal

# Installing a compatible version of jax and jaxlib
pip install jax-metal==0.0.5

# Installing Levanter from source to get the latest updates
git clone https://github.com/stanford-crfm/levanter.git
cd levanter
pip install -e .

# Run a test
python -m levanter.main.train_lm --config config/gpt2_nano.yaml