git-cloner / llama-lora-fine-tuning

llama fine-tuning with lora
https://gitclone.com/aiit/chat/
MIT License
136 stars 14 forks source link

down llama model so slowly #7

Closed mz2sj closed 1 year ago

mz2sj commented 1 year ago

hi~ I used llama.download to download model but the speed is so slow.Any advice for acceleration? By the way,I used command pip install pyllama -U install pyllama.Is there any diffrence beween your suggesion command. Thanks~

little51 commented 1 year ago

There is no necessary to download LLAMA native model files and convert using convert_llama_weights_to_hf.py,please download llama huggingface format file direct: git clone https://huggingface.co/decapoda-research/llama-7b-hf download "lfs" files by browser,put these file together with llama-7b-hf

mz2sj commented 1 year ago

Thank you.I'll try this method!