result is nan

Open knight9630 opened this issue 2 years ago • 0 comments

Run the Conformer model, python -u train.py --data ECL, the output result is all nan, and the loss is also nan, what is the reason?

Dec 06 '23 10:12 knight9630