This repository has been archived by the owner on Nov 22, 2022. It is now read-only.
Add support for length label smoothing #1308
Closed
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
Summary: Adding support for length label smoothing, which from prior experiments improves NAR performance (reducing overfitting on the length prediction task). This diff however runs into an interesting issue with CUDA OOM and throwing an exception here https://fburl.com/diffusion/akkh752n. I believe this is because with label smoothing and length beam size of 5, there is a chance that max target length is the predicted length size, which can cause OOM issues. Reducing max target lengths is critical to helping with this. Another issue is the exception being thrown. This is because of an off by one issue in conv_encoder.max_target_lengths vs conv_decoder.max_target_lengths. Instead of enforcing consistency this updates to make sure the predicted lengths (conv_encoder) are less than or equal to the (conv_decoder) ones. Reviewed By: arbabu123 Differential Revision: D20865186 fbshipit-source-id: dc8e5816c4b31d1adf123179b08cac3667655989
facebook-github-bot
added
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D20865186 |
|
This pull request has been merged in 9548c94. |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.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.
Summary:
Adding support for length label smoothing, which from prior experiments improves NAR performance (reducing overfitting on the length prediction task).
This diff however runs into an interesting issue with CUDA OOM and throwing an exception here https://fburl.com/diffusion/akkh752n.
I believe this is because with label smoothing and length beam size of 5, there is a chance that max target length is the predicted length size, which can cause OOM issues. Reducing max target lengths is critical to helping with this.
Another issue is the exception being thrown. This is because of an off by one issue in conv_encoder.max_target_lengths vs conv_decoder.max_target_lengths. Instead of enforcing consistency this updates to make sure the predicted lengths (conv_encoder) are less than or equal to the (conv_decoder) ones.
Reviewed By: arbabu123
Differential Revision: D20865186