-
Notifications
You must be signed in to change notification settings - Fork 1.2k
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
Training speed of mxnet-ssd slows down? #135
Comments
Can you elaborate how do you use record file with gluon data loader? |
and read recordiofile ,
I pull the new code , change the function to use RecordFileDetection,
The error info is ,
Stack trace returned 10 entries: |
The latest error seems to be unrelated to data loader, did you changed the network training part? |
sorry ,The error before may be caused by gluoncv not updated correctly. I updated gluoncv again and the error is as following,
Stack trace returned 10 entries: |
Try again: |
Can you use num-workers=0 to disable multiprocessing and make sure the record file is good? |
I suspect it might relate to multi-worker but cannot confirm. |
The record file is good because I have tested with my own transform with num-workers = 1 or num-workers= 0. Traceback (most recent call last): Stack trace returned 10 entries: |
I am confused. Can you post your training script? |
OK, but I haven't change training script:
I writed the classes directly in main function :
|
@zhreshold The previous error occured because I have change position format from float to int(original ,not normalized) in record file for my transformer. sorry .
Stack trace returned 10 entries: |
Tracked to apache/mxnet#9974 |
@zhreshold |
@zhreshold |
@WalterMa Yes, this bug should be easy to fix, but need to be careful not to change current api, so we are still discussing. |
An temporary solution is added to RecordFileDetection so multi worker can be enabled. |
It seems a multi-process problem with old rec file dataset?
The text was updated successfully, but these errors were encountered: