Open wxskl opened 1 year ago
Hi, I'm not sure what is going on here, but you can safely ignore the eval during training by commenting out eval_dataset=eval_dataset,
in the trainer argument for now.
Hello, this problem is solved?
you can use model.module.generate, this problem can be solved
AttributeError: 'DataParallel' object has no attribute 'generate'![image](https://user-images.githubusercontent.com/30484996/209766535-ef8b6191-54cd-4490-9855-c3eadd7559fc.png)