-
Notifications
You must be signed in to change notification settings - Fork 1.4k
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
too high loss #416
Comments
Thanks for the bug report @dccho. Unfortunately, I don't know what would cause this. Has this happened to you more than once? Is the problem reproducible? Try it again. If you can provide some extra information, the the engineers working on
|
Probably it is problem with combination of several unstable libraries. Now I change DIGITS to 3.0-rc3 with cuDNN v4, NVCaffe-0.14, then it works fine. |
Glad you got your problem worked out - sorry I wasn't more help debugging! I'll close this now, but anyone who can add to the discussion should still feel free to do so. |
It would be interesting to know if the high loss issue with NVCaffe-0.13 also occurred during single-GPU training? |
i'm using DIGIT 3.0.0-rc1 + NVIDIA/caffe v0.13.2.
Loss2 for training set (green graph) suddenly went to high. Here is log when it was happened.
The text was updated successfully, but these errors were encountered: