James Liu issues

Repositories
Issues
Comments

Results 1 issues of


                                            James Liu

Dequantize to fp16?

Is there an easy way to dequantize the model weights back to fp16 format? Specifically for Exllama I'm looking for something similar to what [bitsandbytes does](https://github.com/huggingface/blog/blob/main/hf-bitsandbytes-integration.md), reproduced below: > The...