Fix Semantic Feature Extraction from Incorrect Layer #333
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
# for free
to join this conversation on GitHub.
Already have an account?
# 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Outdated suggestions cannot be applied.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.
This PR addresses an inconsistency between the code and the paper regarding which layer is used for semantic feature extraction in W2v-BERT 2.0. The paper specifies:
However, in the current code, the features are extracted using
feat = vq_emb.hidden_states[17], which actually points to the 18th layer due to zero-based indexing in Python. This PR updates the code to usevq_emb.hidden_states[16], aligning it with the paper's description if it indeed intended to refer to the 17th layer.Question: Could you clarify if this is an error in the paper or the code? This PR assumes the paper’s statement is correct, but confirmation would be helpful.