model: Add video support to qwen3-vl embedding#4699
Merged
Conversation
AdnanElAssadi56
approved these changes
May 21, 2026
Member
Author
|
@KennethEnevoldsen Can you review this when you have time? |
KennethEnevoldsen
left a comment
Contributor
There was a problem hiding this comment.
Sorry I didn't see this earlier - I think we can avoid the wrapper class (in general I think we could combine most of the ST encoder wrapper into one)
| return embeddings | ||
|
|
||
|
|
||
| class MultimodalInstructSentenceTransformerModel(InstructSentenceTransformerModel): |
Contributor
There was a problem hiding this comment.
Why not simply integrate this into InstructSentenceTransformerModel, seems like we don't need two classes for this
| training_datasets=None, | ||
| citation=QWEN3_VL_EMBEDDING_CITATION, | ||
| extra_requirements_groups=["qwen-vl"], | ||
| extra_requirements_groups=["multimodal-sbert"], |
Contributor
There was a problem hiding this comment.
multimodal-sbert seems a bit funny to me (it is a while ago that people called it sbert). I would use multimodal-sentence-transformer, but it would be breaking so maybe not worth changing.
KennethEnevoldsen
approved these changes
Jun 9, 2026
6 tasks
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
3 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
I got
0.74249, onVidore3ComputerScienceRetrieval.v2 (eng subset)in results we have0.74409. I think new implementation is close enough.I run BreakfastClassification and got:
{ "dataset_revision": "59a874899eb241993794a3454c37829727c3b559", "task_name": "BreakfastClassification", "mteb_version": "2.13.1", "scores": { "test": [ { "scores_per_experiment": [ ... ], "accuracy": 0.471024, "f1": 0.448974, "f1_weighted": 0.467435, "precision": 0.482644, "precision_weighted": 0.503809, "recall": 0.459089, "recall_weighted": 0.471024, "ap": NaN, "ap_weighted": NaN, "main_score": 0.471024, "hf_subset": "default", "languages": [ "eng-Latn" ], "mteb_version": "2.13.1" } ] }, "evaluation_time": 101.02676701545715, "kg_co2_emissions": null, "date": 1779312638.304263 }