Rerank scorer: опция для выбора источника для расчета вектора вероятностей #115
Conversation
This file contains 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
voorhs
previously approved these changes
Feb 3, 2025
voorhs
dismissed
their stale review
не подумал про масштаб скоров из кросс энкодера
voorhs
reviewed
Feb 3, 2025
Comment on lines
+168
to
+170
| if self.use_crosencoder_scores: | ||
| distances.append([rank["score"] for rank in cur_ranks]) | ||
| else: |
There was a problem hiding this comment.
надо посмотреть на масштаб скоров которые выходят из кросс-энкодера
дело в том что скоры из векторного индекса это значения косинуса (т.е. [-1,+1]), а в кросс-энкодере по-моему выдаются логиты (т.е. (-infty, +infty)) либо вероятности (т.е. [0,1]).
в общем надо в классе Ranker посмотреть на метод rank и сделать так чтобы
- всегда выдавалось [-1,1]
- либо чтобы можно было выбирать между [-1,1] и [0,1]
но сырые логиты точно не нужно выдавать
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Считать вектор на основе данных из kNN или из CrossEncoder.