Label-studio: Slow UI response times after ML backend connection

[atakanokan]

Describe the bug
UI response have slowed down very much after connecting a ML backend to the annotation project. Disabling the ML backend doesn't make response times return to normal.

To Reproduce
Steps to reproduce the behavior:

1. Create a new annotation project
2. Open a task -> fast (<0.3 seconds)
3. Create a new ML backend
4. Connect the ML backend to the annotation project
5. Open the annotation project again (doesn't matter with ML backend or not)
6. Open a task -> slow (5> seconds)

Expected behavior
For the task to open fast as it was before the ML backend was connected.

Screenshots
ML backend disabled:

Environment (please complete the following information):

• OS: Windows 10
• Label Studio Version: 0.9.1.post1

Additional context
NA

[atakanokan]

Update:
After deleting the ML backend connection (screenshot below), the task opening is now fast again. I am guessing when there is a ML backend added to the annotation project, label-studio tries to get predictions even if the ML backend is not connected. Would it be possible to bypass that so that the UI works much faster when a ML backend is added but not connected?

[Rishabh-hyperverge]

I am facing the exact same problem

[niklub]

Hey, guys, that issue has been fixed in https://github.com/heartexlabs/label-studio/pull/681/commits/33eb3e21e08a037d7d246bc87752a81b66470fc4 and will be released as 1.0.1 version