Skip to content
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

acc and loss remains the same #3

Open
teddyjkwang opened this issue Dec 21, 2017 · 3 comments
Open

acc and loss remains the same #3

teddyjkwang opened this issue Dec 21, 2017 · 3 comments

Comments

@teddyjkwang
Copy link

THX for your great effort.
I try to run the code but found that acc remains 0.119 for a long time and loss remains about 2.3000. How to optimize it?

@xyyu-callen
Copy link

@daviddao
Many thanks.
I have got the same problem.
The accuracy is always 0.119 and the loss is about 2.3000.
Can you help us to solve it?

@daviddao
Copy link
Owner

Try to restart the training and tune hyperparameters. One case why the accuracy and loss get stuck might be because the transformer zoomed in too radical into an empty part of the image!

@chl916185
Copy link

@SwaggyK93 @xyyu-kevin
Accuracy (498): 0.957
Iteration: 0 Loss: 0.0012747037
Iteration: 10 Loss: 0.0026973477
Iteration: 20 Loss: 0.0013430312
Iteration: 30 Loss: 0.005542839
Iteration: 40 Loss: 0.00043637512
Iteration: 50 Loss: 0.0073211356
Iteration: 60 Loss: 0.0051542623
Iteration: 70 Loss: 0.007538586
Iteration: 80 Loss: 0.00058482925
Iteration: 90 Loss: 0.00042093257
Accuracy (499): 0.957
i try to run the code ,result is 0.957

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

4 participants