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Outdated suggestions cannot be applied.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.
Overview
This PR implements entity pre-training. Simply put, entity pre-training involves pre-training the entity module before training the entire NER + RE model jointly.
Previously, we were implementing this by delaying the training of the RE module by a full epoch. I have found a different yet equally simple technique that works quite well. Basically, the loss function is weighted according to
where
0 <= delay_coef <= 1
and is computed ascurrent_step / total_steps
if this is the first epoch, or1
otherwise. I saw gains up to 1% on the validation set using this scheme.