senseable

/

WestLake-7B-v2

like 108

This is the imatrix for WestLake-7B-v2. I computed it using WikiText-2-raw (10,800 chunks = ~2.5 hrs). Please use this to make IQuants that have much closer perplexity and accuracy to the F16 of the model. Translation for the uninformed: lower perplexity and better accuracy = better performing llm. Thanks 🙏
https://huggingface.co/Joseph717171/Imatrices/blob/main/WestLake-7B-v2_wikitext-2-raw.imatrix