search

unslothai / unsloth

Finetune Llama 3.1, Mistral, Phi & Gemma LLMs 2-5x faster with 80% less memory
https://unsloth.ai
Apache License 2.0
15.08k stars 1.01k forks source link

Yi 1.5 Model Support #523

Open suptejas opened 3 months ago

suptejas commented 3 months ago

Yi 1.5 models are some of the most capable < 10B LLMs out there. It would be amazing to get fine-tuning capabilities for the model via Unsloth.

avcode-exe commented 3 months ago

I think it is possible to fine tune Yi-1.5, just replace the model name to be 01-ai/Yi-1.5-9B or unsloth/Yi-1.5-6B-bnb-4bit or other models. I used unsloth to fine tune 01-ai/Yi-1.5-9B, here is the model NotAiLOL/Athena-Yi-1.5-9B-v0.1.

shimmyshimmer commented 3 months ago

You will need to change the model name like @avcode-exe said.

We have more details about Yi in our blog post: https://unsloth.ai/blog/phi3

danielhanchen commented 3 months ago

@suptejas @avcode-exe Yes it should be supported now! Simply change the model name to any model from Yi.

Also update Unsloth for local machines (Colab / Kaggle just refresh and restart the env):

pip uninstall unsloth -y
pip install --upgrade --force-reinstall --no-cache-dir git+https://github.com/unslothai/unsloth.git