Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

save function does save modules.json incorectly for jina-embeddings-v3 #3111

Open
guenthermi opened this issue Dec 3, 2024 · 3 comments · May be fixed by #3112
Open

save function does save modules.json incorectly for jina-embeddings-v3 #3111

guenthermi opened this issue Dec 3, 2024 · 3 comments · May be fixed by #3112
Labels
bug Something isn't working

Comments

@guenthermi
Copy link

We noticed that when saving a jina-embeddings-v3 model as in the code example below the, the kwargs attribute is missing in modules.json of the saved model. When calling the encode function after loading a model, the task attribute is not passed to the the model anymore.

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("jinaai/jina-embeddings-v3", trust_remote_code=True)
model.save('test_model')
@tomaarsen tomaarsen added the bug Something isn't working label Dec 3, 2024
@tomaarsen tomaarsen linked a pull request Dec 3, 2024 that will close this issue
@danielbichuetti
Copy link

There are issues with jinaai/jina-clip-v2 too. When you save and try to load again, custom_st is not found.

@tomaarsen
Copy link
Collaborator

@danielbichuetti I chased this down - there's 2 issues currently, one must be fixed in jina-clip-v2 itself (https://huggingface.co/jinaai/jina-clip-v2/discussions/17), and the other one I'll fix in #3112.
Thanks for mentioning!

  • Tom Aarsen

@danielbichuetti
Copy link

@tomaarsen That's great! Thank you for the fast fix. That's amazing, as always.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug Something isn't working
Projects
None yet
Development

Successfully merging a pull request may close this issue.

3 participants