This repository was archived by the owner on Nov 22, 2022. It is now read-only.
Always create non-empty logits / targets for MLM #979
Closed
Conversation
This file contains hidden or 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
facebook-github-bot
added
the
CLA Signed
mwu1993
pushed a commit
to mwu1993/pytext-1
that referenced
this pull request
Sep 16, 2019
Summary: Pull Request resolved: facebookresearch#979 D17241503 selects only masked tokens for the final logits / targets during MLM. This fails when there are no masked tokens (e.g. at the end of a file there can be a very short batch). In this case, select just the first token in the first batch. BTW, the new masking strategy is faster (20%) than the old one - f138005051 vs f138005044. Differential Revision: D17370855 fbshipit-source-id: 0304da89c0ccd99083e761cb8f648e01456ebdd8
mwu1993
force-pushed
the
export-D17370855
branch
from
eedd38c to
7d51646
Compare
Summary: Pull Request resolved: facebookresearch#979 D17241503 selects only masked tokens for the final logits / targets during MLM. This fails when there are no masked tokens (e.g. at the end of a file there can be a very short batch). In this case, select just the first token in the first batch. BTW, the new masking strategy is faster (20%) than the old one - f138005051 vs f138005044. Reviewed By: borguz Differential Revision: D17370855 fbshipit-source-id: 62f0540fb94819c6a269ea067b1c0b9e08c82119
mwu1993
force-pushed
the
export-D17370855
branch
from
7d51646 to
89a45ec
Compare
|
This pull request has been merged in 513abc9. |
rutyrinott
pushed a commit
to rutyrinott/pytext-1
that referenced
this pull request
Sep 25, 2019
Summary: Pull Request resolved: facebookresearch#979 D17241503 selects only masked tokens for the final logits / targets during MLM. This fails when there are no masked tokens (e.g. at the end of a file there can be a very short batch). In this case, select just the first token in the first batch. BTW, the new masking strategy is faster (20%) than the old one - f138005051 vs f138005044. Differential Revision: D17370855 fbshipit-source-id: 453ccc4797c0d976281aa14212eba43b6f9b5d6f
rutyrinott
pushed a commit
to rutyrinott/pytext-1
that referenced
this pull request
Sep 25, 2019
Summary: Pull Request resolved: facebookresearch#979 D17241503 selects only masked tokens for the final logits / targets during MLM. This fails when there are no masked tokens (e.g. at the end of a file there can be a very short batch). In this case, select just the first token in the first batch. BTW, the new masking strategy is faster (20%) than the old one - f138005051 vs f138005044. Differential Revision: D17370855 fbshipit-source-id: 6c8dbd6346e8f8db029047171cb81f147dc70a43
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:
D17241503 selects only masked tokens for the final logits / targets during MLM. This fails when there are no masked tokens (e.g. at the end of a file there can be a very short batch). In this case, select just the first token in the first batch.
BTW, the new masking strategy is faster (20%) than the old one - f138005051 vs f138005044.
Differential Revision: D17370855