How to solve the problem of continuing pre-training the XLSR-53 2B model

facebookresearch / fairseq

Facebook AI Research Sequence-to-Sequence Toolkit written in Python.

MIT License

30.57k stars 6.41k forks source link

How to solve the problem of continuing pre-training the XLSR-53 2B model #5554

Open Learneducn opened 1 month ago

Learneducn commented 1 month ago

🐛 Bug

When I use 8 A40 images to continue pre-training the XLSR-53 2B pre-training model, I always get an error message: Expected is_sm80 to be true, but got false. (could this error message be improved? If so, please report an enhancement request to pytorch) How can I solve this problem?