metadata
license: other
4-bit (128 groupsize) quantized files for upstage/llama-30b-instruct
Quantized using GPTQ-for-LLaMa.
Command used to quantize: python llama.py /my/model/directory c4 --wbits 4 --true-sequential --act-order --groupsize 128 --save_safetensors /my/output/file.safetensors