Hi!
I'm trying to implement a "keep N best checkpoints" logic in my project. (It was discussed here some time ago but I can't see the thread due to Slack limitations) In that thread, a function https://github.com/allegroai/clearml/blob/4e2e6a4d6b43b6c7cd9ab8ad8dcb9f1b415946f5/clearml/backend_interface/task/task.py#L1523 https://github.com/allegroai/clearml/blob/4e2e6a4d6b43b6c7cd9ab8ad8dcb9f1b415946f5/clearml/backend_interface/task/task.py#L1523 https://github.com/allegroai/clearml/blob/4e2e6a4d6b43b6c7cd9ab8ad8dcb9f1b415946f5/clearml/backend_interface/task/task.py#L1523 was recommended which appeared to be quite helpful to me.
At the same time I found out that the responsibility for artifacts saving lies with task._artifacts_manager, which has its own list of artifacts which in turn isn't impacted by the task._ delete_ artifacts() function. This results in the "resurrection" of the artifacts deleted using the function in the following scenario: task.upload_artifact(artifact_A) -> task. delete_ artifacts([artifact_A]) -> task.upload_artifact(artifact_B) . In this case, in spite of the removal of artifact_A, it will show again in the WEB UI for the task.
So could you kindly tell me if there's a way to remove an artifact from the task altogether?
Thank you! I'll try it out and let you know the result
What about Task.unregister_artifact ?
https://clear.ml/docs/latest/docs/references/sdk/task#unregister_artifact
Tried it, the outcome's still the same though: the artifacts deleted using the task._delete_artifacts() function resurrect on further calls of task.upload_artifact() on new artifacts
