ColossalAI icon indicating copy to clipboard operation
ColossalAI copied to clipboard

Published 20 hours ago • verified publisher icon hpcaitech

[BUG]: The LLama model trained in Lora mode is unable to perform normal reasoning

Open tianbuwei opened this issue 2 years ago • 1 comments

🐛 Describe the bug

In this path 'ColossalAI/applications/Chat/examples/train_sft.sh', LLama-7B model is trained with Lora training method, but there is a problem in the reasoning process, is it because Lora parameters are not loaded? May I ask how to solve it? Snipaste_2023-04-12_14-03-34 Snipaste_2023-04-12_14-03-55

Environment

Python 3.9.16 torch 1.12.1 torchaudio 0.12.1 torchvision 0.13.1 transformers 4.28.0.dev0

tianbuwei avatar Apr 12 '23 06:04 tianbuwei

Hi @tianbuwei Thanks for the feedback, we are fixing LoRA bug. #3439

binmakeswell avatar Apr 17 '23 08:04 binmakeswell