-
Notifications
You must be signed in to change notification settings - Fork 1.9k
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
fix: stash and restore prediction when calling learn during learn_returns_prediction==false calls #4632
fix: stash and restore prediction when calling learn during learn_returns_prediction==false calls #4632
Conversation
jackgerrits
commented
Aug 15, 2023
•
edited
Loading
edited
- Fixes Igl's usage of the prediction in learn
- Causes one less call to predict for cats which means the PRNG sequence changes and so do the predictions.
- Seems that expreplay used to be returning the final prediction that was coming out of GD in its learn call. This fixes that to be the initial predict.
…urns_prediction==false calls
30.00431,0.00625 | ||
0.51225615,0.40625 | ||
2.3816023,0.40625 | ||
25.271912,0.00625 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
numbers here are crazily different - maybe this test can be reconsidered (in next PRs)
|
||
finished run | ||
number of examples per pass = 200 | ||
passes used = 8 | ||
weighted example sum = 1600.000000 | ||
weighted label sum = 728.000000 | ||
average loss = 0.047746 | ||
average loss = 0.455000 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
so loss went up?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Previous loss was cheating since it was using predictions that were coming out of learn.