4.65 bits/bpw quantization of sheep-duck-llama-2 (https://huggingface.co/Riiid/sheep-duck-llama-2) with exllamav2.
Update on 21/09/2023: Re-did the quant with newest exllamav2 version, which fixed some measurement issues.
4.65 bits/bpw quantization of sheep-duck-llama-2 (https://huggingface.co/Riiid/sheep-duck-llama-2) with exllamav2.
Update on 21/09/2023: Re-did the quant with newest exllamav2 version, which fixed some measurement issues.