Repository navigation
fix(projection): detect numpy and polars list columns as vectors - #271
Merged
Merged
Conversation
Auto modality detection only accepted a Python list of int/float. Lists of numpy scalars (e.g. float32 embeddings turned into lists) and polars List/Array columns were treated as text, so the vectors were either stringified and embedded again or failed the text cast.
donghaoren
approved these changes
Oct 1, 2026
Collaborator
|
Thanks for fixing this! |
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Problem
_infer_modalityonly calls a column "vector" when the first value is a Python list of int/float or an ndarray. Two common cases miss that and get treated as text:A pandas column built like
[list(v) for v in emb]from a float32 array holdsnp.float32values, which aren'tfloatsubclasses. The vectors get stringified to'[np.float32(0.31...), ...]'and embedded again with sentence-transformers. No error, just a wrong projection.A polars
ListorArraycolumn, wherenon_null[0]comes back as a narwhals Series instead of a list. That one ends in a List to String cast error.test_vector_polarspasses only because it setsmodality="vector".Fix
Check the column dtype first and return "vector" for
nw.List/nw.Arraywith a numeric inner type, and acceptnp.numberin the list check. Lists of strings still fall through to text. Added tests for both cases plus a polars list of strings.