-
Notifications
You must be signed in to change notification settings - Fork 22
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
Dropout used incorrectly! #27
Comments
Thanks for bringing this to my attention. I will make the fix this week. I like the idea of making it a placeholder. |
Would this change affect the accuracy of the model. What I mean to ask is, whether this is a code design change or model improvement change ? |
I think the error is still there, you need to set the dropout to 1 when forward_only (i.e during the testing step) anyway i did the change on my own and i can confirm an improvement on accuracy up to 1% on my best model with less overfitting (I have dropout_keep_prob = 0.7) |
It's probably an oversight on my part, I thought I had made that change a while ago but apparently not. Thanks for bringing this to my attention. |
during the test the
dropout_keep_prob
parameter is not set to 1 as it should, thedropout_keep_prob
parameter shouldn't be constant but a placeholder or at least atf.Variable
The text was updated successfully, but these errors were encountered: