Finish training if we cannot get better dev ppl. for N hours. #5
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
I thought that the early stopping is usually done based on number of evaluations (e.g. if there is no dev-loss improvement in recent K evaluations, then the trainer would stop). Why is PR based on the fixed time? |
odashi
reviewed
Jan 13, 2017
| @@ -180,6 +180,10 @@ dropout_ratio=0.3 | |||
| ; Maximum number of batch data to be trained. | |||
| max_iteration=10000 | |||
|
|
|||
| ; Training will be stopped if we cannot get the lowest dev ppl. for this hours. | |||
| ; If this is set to 0, we don't stop training with this criteria. | |||
| max_waiting_hour=24 | |||
There was a problem hiding this comment.
this method could not stop the trainer as soon as the dev-loss goes worse. Maybe it is better to remove this behavior and to recommend to set some large value (e.g. 114514) to disable early stopping virtually.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
I added the function to finish training if we cannot get better dev ppl. for N hours.
In the default settings I set it to 24 hours, so if 24 hours have passed after getting the best dev ppl.,
training will be finished automatically.
If set it to 0, we do not stop with this criteria.