Wang-Tianwei/Decoupled-attention-network

Bug in `dataset_hw.py`

Opened this issue · 2 comments

Hi, thanks for your impressive work, I am really interested in it.
But when I test your code with the released model, I found that there is a small bug in dataset_hw.py.
Specifically, no locking mechanism for self.idx in class LineGenerate, which will cause an error when running in parallel.
The true testing results may be a little lower than your paper reported.

num_workers is set to 2 as your code
image

when num_workers = 1
image

Thank you for discovering this bug, I didn't notice it before~

Thanks for your reply,
and I wonder if a similar situation applies to RIMES?
If so, could you please provide the corrected result on RIMES or the corresponding checkpoints?