feat: support for pylate models #1
Open
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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.
Description
This PR adds support for pylate-based ColBERT models. The existing implementation in
main.pyshows suboptimal performance, particularly in recall, when using modern pylate-compatible models. This update implemented in the newmain_pylate.pyscript, correctly integrates the pylate library and demonstrates significant performance improvements.Benchmark Results
The following results were obtained using:
Model:
ayushexel/colbert-ModernBERT-base-1-neg-1-epoch-gooaq-1995000Dataset:
zeta-alpha-ai/NanoFiQA2018Before (Current
main.py)===================================================================================== FINAL REPORT (Dataset: zeta-alpha-ai/NanoFiQA2018) ===================================================================================== Retriever | Indexing Time (s) | Avg Query Time (ms) | Recall@10 ------------------------------------------------------------------------------------- 1. ColBERT (Native) | 26.55 | 395.44 | 0.1600 2. ColBERT + FDE | 261.37 | 111.83 | 0.0000 =====================================================================================After (This PR's
main_pylate.py)===================================================================================== FINAL REPORT (Dataset: zeta-alpha-ai/NanoFiQA2018) ===================================================================================== Retriever | Indexing Time (s) | Avg Query Time (ms) | Recall@10 ------------------------------------------------------------------------------------- 1. ColBERT (Native) | 12.55 | 447.60 | 0.4400 2. ColBERT + FDE | 188.83 | 146.80 | 0.1000 =====================================================================================Additional Notes
There is a known issue in pylate regarding an incompatibility with MUVERA and the normalization scores for models trained via distillation. Consequently, models trained with this method may exhibit poor performance.
For more details, see the related GitHub issue: lightonai/pylate#142