GPTQ vs Q4 GGUF

#2
by ciprianv - opened

Thank you for providing these GPTQ version. How should it compare as accuracy vs standard Q4 GGUF. Is there any post quantization training done? (QAT). In vllm/sglang does this has advantages over AWQ?

I also wonder about this, but from testing this model does very well. I imagine having the original training set to use as the dataset would provide a good model.

Sign up or log in to comment