Issues
- 1
训练损失降到很低,但predict.py的输出还是<s></s>
#5 opened by woaipichuli - 1
运行chatglm的SFT代码,loss显示NAN
#4 opened by woaipichuli - 1
请问pytorch_bin下载很慢有办法解决不
#3 opened by libertyarow - 0
finetune LLama
#2 opened by xzdong-2019 - 1
Why can not use multi GPUs to train even though set CUDA_ENVISION_DEVICES='0,1,2,3'?
#1 opened by GaoXinJian-USTC