Hello everyone.
I would like to ask about the rapid decline in acuracy of opportunity learning.
When I run the code for learning locally, it is fine, but when I run it in queue, the accuracies suddenly drop.
No error is displayed and I want to find out the cause. If you know the cause or know how to identify the cause, please advise me.
The agent on the server side to which it was assigned remained activated.
(6:14 PM is the end of the 100 epoch, so workers worked until the end.)
This script uses dataset (torchvision.datasets.CIFAR10) for a three-level classification model with pytorch.
Thank you!!
Hi @<1761199244808556544:profile|SarcasticHare65> , and if you run locally for the same amount of iterations this does not happen?
@<1523701070390366208:profile|CostlyOstrich36>
Thanks for the reply!
Sorry it took me so long to reply as I was able to solve the problem myself right away.
The problem is that I let the learning rate get too big.
I reduced the learning rate and that solved the problem.
Thanks!
