Add upload study artifact api

[gen740]

Contributor License Agreement

This repository (optuna-dashboard) and Goptuna share common code.
This pull request may therefore be ported to Goptuna.
Make sure that you understand the consequences concerning licenses and check the box below if you accept the term before creating this pull request.

• I agree this patch may be ported to Goptuna by other Goptuna contributors.

Reference Issues/PRs

None.

What does this implement/fix? Explain your changes.

• Add the upload_study_artifact api.
• Add the study artifact cards in the StudyHistory page.

[HideakiImamura]

Let me read this PR for my study. I will not merge the PR with my approval.

[HideakiImamura]

Thanks for the PR. I have some early comments. PTAL.

[codecov]

Codecov Report

Attention: 24 lines in your changes are missing coverage. Please review.

Comparison is base (0755902) 62.88% compared to head (68e181e) 63.42%.
Report is 11 commits behind head on main.

Files	Patch %	Lines
optuna_dashboard/artifact/_backend.py	29.41%	24 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #674      +/-   ##
==========================================
+ Coverage   62.88%   63.42%   +0.53%     
==========================================
  Files          35       35              
  Lines        2250     2280      +30     
==========================================
+ Hits         1415     1446      +31     
+ Misses        835      834       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[keisuke-umezawa]

Sorry for being late. I just checked python API part, and left comments. Regarding python API parts, it looks almost good to me other than my comments.
If you divide this PR into python API change and adding new UI, I can merge the PR for python API change immediately.

[keisuke-umezawa]

[question] As study artifact is introduced in this PR, the metadata is always stored in ARTIFACTS_ATTR_PREFIX + artifact_id?

[gen740]

Yes, this behavior is the same as optuna/artifacts/_upload.py

[keisuke-umezawa]

In that case, can we remove those lines?

storage.set_study_system_attr(
            study_id, _dashboard_artifact_prefix(study_id) + artifact_id, json.dumps(None)
        )

[gen740]

It sounds good. I think I can remove that line from delete_study_artifact. However, I would not remove it from delete_trial_artifact due to backward compatibility.

[keisuke-umezawa]

It is not relevant to this PR, but should we also need to delete meta data in delete_all_artifacts?

[keisuke-umezawa]

Also, it is not relevant to this PR or optunda-dashboard, but should we prepare API for deleting data and meta together like delete_artifact similar to upload_artifact in optuna itself?

[HideakiImamura]

Thanks for the update. I added some additional comments. PTAL

[keisuke-umezawa]

I checked ui side and as a code, looks almost good to me. However, in my environment, I can only see blank study artifact without uploader. What should I do to show uploader?

[gen740]

However, in my environment, I can only see a blank study artifact without an uploader.

Good catch, that's because you did not pass the artifact_store instance to run_server (e.g., run_server(storage, artifact_store=artifact_store)).
Anyway, this is a bug because it should not appear when artifact_store isn't passed. Thanks!

[keisuke-umezawa]

LGTM!
Thank you for working on it.