Skip to content
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

Improve dpr conversion #826

Merged
merged 4 commits into from
Mar 18, 2021
Merged

Improve dpr conversion #826

merged 4 commits into from
Mar 18, 2021

Conversation

Timoeller
Copy link
Contributor

@Timoeller Timoeller commented Feb 12, 2021

implemented changes

  • The key value pair "passage_id" was missing during conversion of squad to DPR.
  • Added some minor improvements to the code.

open TODOs

  • preprocessing of positive context (currently the whole passage is taken instead of a 100 word passage containing any of the annotated answers)
  • the option for additional data besides SQuAD training set, such that hard negatives can be chosen from a larger pool of passages
  • [optional] adding of no_answers as negative passages only (this change requires adjustments to the training code that always expects a positive context)

@brandenchan
Copy link
Contributor

Tried it out and it seems to work well! LGTM

@Timoeller Timoeller changed the title WIP: Improve dpr conversion Improve dpr conversion Mar 18, 2021
@Timoeller Timoeller merged commit 7b559fa into master Mar 18, 2021
@Timoeller Timoeller deleted the fix_squad_dpr_conversion branch March 18, 2021 13:51
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

2 participants