fix(response): clean up chunk text #16
Merged
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
clean up chunk text in the response to remove prefix for easier downstream processing
mlissner
reviewed
Mar 25, 2025
Comment on lines
+167
to
+169
| clean_chunks = [ | ||
| chunk.replace("search_document: ", "") for chunk in all_chunks | ||
| ] |
There was a problem hiding this comment.
Hm, will this use a bunch more memory where a different solution could remove search_document further upstream so that a full copy (and manipulation) of the response won't be necessary?
There was a problem hiding this comment.
The line right above this is where we pass all_chunks to the model to generate the embeddings, so we cannot do this upstream, unless one the three options:
- We carry two copies of the chunks, one with the prefix, one without, and pass the one with prefix to the model and return the one without as the text
- We create the chunks without the prefix but before passing the chunks to the models, we add the prefix, which is equivalent to this but in reverse
- or, we simply do not use the prefix. The nomic_ai model is designed for query-context with prefix so this will impact the performance slightly
I don't think either of the first two options will be any better, this list comprehension will consume CPU memory, it will not impact GPU memory.
There was a problem hiding this comment.
Cool. Got it. Let's stick with what you've got. Thanks.
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Outdated suggestions cannot be applied.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.
clean up chunk text in the response to remove prefix for easier downstream processing
Per discussion: freelawproject/courtlistener#5298 (comment)
We should clean up the chunk text in the response to remove the prefix for easier downstream processing and keeping the opinion text clean