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
12.99k stars 843 forks source link

docker images #697

Open skunkwerk opened 3 weeks ago

skunkwerk commented 3 weeks ago

Could someone please share a working Docker image for CUDA 12 and PyTorch 2.3? I've tried building my own, but fails to run properly (the conda version complains there's no nvidia driver, the pip version results in a messed up pytorch).

thanks

danielhanchen commented 3 weeks ago

Apologies on the delay - just relocated to SF, hence the delay!

If you're using Conda - try copying the Kaggle install instructions here: https://www.kaggle.com/code/danielhanchen/kaggle-llama-3-8b-unsloth-notebook

dsingal0 commented 2 weeks ago

@skunkwerk I have a dockerfile and docker image with torch 2.3 and cuda 12.1 https://github.com/unslothai/unsloth/pull/738