The linked Huging Face page in the README doesn't have any .bin files

RWKV / rwkv.cpp

INT4/INT5/INT8 and FP16 inference on CPU for RWKV language model

MIT License

1.41k stars 95 forks source link

The linked Huging Face page in the README doesn't have any .bin files #159

Closed krikru closed 8 months ago

krikru commented 8 months ago

Under "Option 3.1. Download pre-quantized Raven model" in README.md, the instructions tell you to "check that you are downloading .bin file, not .pth," but on the linked page, there are .pth files but no .bin files.

saharNooby commented 8 months ago

Blink does not provide pre-quantized models for rwkv.cpp anymore, so I've removed the section.

I'm not comfortable officially recommending third-party pre-quantized models on Hugging Face, and instead recommend to do the conversion yourself.

However, looks like latestissue on Hugging Face has uploaded all RWKV models in all available formats -- you may find these useful, but use them at your own risk.