Wrong predicted
computation for multi-label datasets
#1527
Labels
area: server
Indicates that an issue or pull request is related to the server
type: bug
Indicates an unexpected problem or unintended behavior
Projects
Describe the bug
A record annotated/predicted without labels (a special allowed case for the text classification multi-label problem) does not reflect the predicted ok/ko status in the filters.
Also, the f1 metrics are affected.
To Reproduce
Steps to reproduce the behavior:
Expected behavior
No labeled annotated/predicted records should be included as part of predicted ok/ko metrics.
Screenshots
Environment (please complete the following information):
Additional context
The text was updated successfully, but these errors were encountered: