-
Notifications
You must be signed in to change notification settings - Fork 196
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
seems deadlock occurs when using multithread #75
Comments
I change the num_data_threads to 2. It's working now...But wondering why ? |
Thanks, this should be a bug. @8z8z8z please help check this |
It might be related to this horovod issue . the smaller the num_data_threads I set, the faster I wait for the start of evaluation stage. Here's the part of the log of evaluation:
|
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. |
Keeping this issue open. This is still a bug. |
This is a script for auto install horovod in centos and may help you |
Here's the mtl_transformer.json for finetuning stage. I set the num_data_threads:32, and it hangs up at the end of the first training epoch.
horovodrun -np 4 -H localhost:4 python athena/horovod_main.py examples/asr/hkust/configs/mtl_transformer.json
Here's part of the strace log of one of the process:
The text was updated successfully, but these errors were encountered: