zero-lora
zero-lora copied to clipboard
xturing 低 VRAM 微调
https://github.com/oobabooga/text-generation-webui/issues/1088
似乎可以用低至 6Gib VRAM 微调 INT4 LLM,这要归功于随机/xturing:https://github.com/stochasticai/xturing#-new-feature---int4-fine-tuning-with-llama-lora
将此功能集成到文本生成中会很棒 webui .
谢谢!
其他上下文
https://github.com/stochasticai/xturing/blob/main/examples/int4_finetuning/README.md