yu1ut / multi-task-curriculum-framework-for-open-set-ssl Goto Github PK

In line 23 of the train.py, which is 'from utils import Bar, Logger, AverageMeter, accuracy, mkdir_p, savefig'
The code report the error:
Traceback (most recent call last):
File "train.py", line 23, in
from utils import Bar, Logger, AverageMeter, accuracy, mkdir_p, savefig
File "C:\code\Multi-Task-Curriculum-Framework-for-Open-Set-SSL-master\utils_init_.py", line 10, in
from progress.bar import Bar as Bar
ModuleNotFoundError: No module named 'progress'

'_SingleProcessDataLoaderIter' object has no attribute 'next'

I got this error that the s of all unlabeled samples would converge to the same value so that the threshold_otsu cannot be solved with one value when the epoch reached about 700. And the test ACC had been kept very low.

The following is the log file:
log.txt

Hi
Thank you for your excellent fundamental work in the robust SSL.
I have a question regarding the experiment setting in your paper:
In Table1 of your paper, https://arxiv.org/pdf/2007.11330.pdf, for different labeled data setting, you use validation set of 5000 for the validation set, and you tune hyperparameters for the different labeled data setting, is this correct?

Looking forward to hearing from you!