Facico / Chinese-Vicuna

Chinese-Vicuna: A Chinese Instruction-following LLaMA-based Model —— 一个中文低资源的llama+lora方案,结构参考alpaca
https://github.com/Facico/Chinese-Vicuna
Apache License 2.0
4.14k stars 422 forks source link

你好,训练是正常的,但是一到模型保存的step就会显存爆炸,很奇怪,保存模型为啥会额外占用显存呢 #178

Closed wangfudong65 closed 1 year ago

Facico commented 1 year ago

检查一下bitsandbytes版本 你的问题可以参考这里:https://github.com/Facico/Chinese-Vicuna/blob/master/docs/notes.md#oomout-of-memory%E7%88%86%E6%98%BE%E5%AD%98

wangfudong65 commented 1 year ago

检查一下bitsandbytes版本 你的问题可以参考这里:https://github.com/Facico/Chinese-Vicuna/blob/master/docs/notes.md#oomout-of-memory%E7%88%86%E6%98%BE%E5%AD%98

我看看,感谢

xuehunmu commented 1 year ago

我保存的时候也是显存爆炸了,重新建了个虚拟环境就正常了