Trm adjustment#88
Merged
Merged
Conversation
Eliminates abbreviation-based normalization in filename boosting and related tests, relying instead on simple morphological normalization (e.g., plural/singular). Updates tests to reflect the removal of abbreviation handling and improves robustness of the ConfidenceEstimator tests. Also adds persistence, hot-reload, and online learning support to LatentRefiner, and improves numerical stability in score blending.
Updated CollectionLearner to train and save weights for both TinyScorer and LatentRefiner after each batch. Enhanced logging to include refiner version, and improved the training loop to refine latent state toward teacher-optimal targets. Also ensured LatentRefiner checks for hot-reloaded weights and uses collection-specific weight paths.
🤖 Augment PR SummarySummary: This PR refines the TRM-style reranker by adding a trainable/persisted latent refiner, improving robustness around weight/checkpoint handling, and tightening numerical stability/thread safety. Changes:
Technical Notes: Weight persistence now uses advisory file locks to reduce partial reads during hot reload; learning worker writes weights/checkpoints via atomic rename semantics. 🤖 Was this summary useful? React with 👍 or 👎 |
![augmentcode[bot]](https://avatars.githubusercontent.com/in/1027498?s=60&v=4){kind=small}
Update logger calls from debug to warning level and include collection context in log messages for better traceability of errors and skipped events during reranker processing.
Updates _save_weights to use a base path without the .npz extension for np.savez, mirroring TinyScorer._save_weights. Ensures the temporary file is correctly named and atomically renamed into place, preventing file extension issues.
Renamed 'learning_iterations' to 'refinement_iterations' in mcp_indexer_server.py for clarity. Updated the default RERANK_EVENTS_SAMPLE_RATE from 0.33 to 0.5 in rerank_events.py to increase the proportion of logged events.
The README has been rewritten for clarity and conciseness, focusing on quick start instructions, key differentiators, supported clients, endpoints, and extension usage. Outdated or redundant sections were removed, tables and guides were updated, and the architecture diagram and documentation links were improved for easier onboarding and reference.
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.
No description provided.