Dear Author,
This is a great result, I am recently using TAPE for sequence design task, I have no problem in single card training process, but I encounter a little problem using distributed training.
I'm looking forward to hearing from you.
Problem Description:
- After the distributed training is finished, the program cannot be exited normally and remains in running state
- There have been no problems with the training process.
- 4 RTX6000 on one machine


I'm looking forward to hearing from you!