georgesung

/

llama2_7b_chat_uncensored

like 326

Is there a way I can convert the .safetensors files back to the .pth (pytorch) files meta used on their models (like llama-2-7b-chat which contain a consolidated.00.pth file and params.json)?
I would like to use different models but cannot simply torch.load them like I would with the .pth file.

Tldr;
Any way I can convert or use this model the way I can with the consolidated.00.pth file without lmstudio, llama2.cpp, etc. but just with what I got when cloning meta's llama2 gitHub repo?