模型:
mayaeary/pygmalion-6b_dev-4bit-128g
GPTQ量化 https://huggingface.co/PygmalionAI/pygmalion-6b/commit/30e2405100eac6bd53f75964cc7345eeafd19f7d
使用此存储库: https://github.com/mayaeary/GPTQ-for-LLaMa/tree/gptj-v2
命令:
python3 gptj.py models/pygmalion-6b_dev c4 --wbits 4 --groupsize 128 --save_safetensors models/pygmalion-6b_dev-4bit-128g.safetensors