Examples: query, "exact match", wildcard*, wild?ard, wild*rd
Fuzzy search: cake~ (finds cakes, bake)
Term boost: "red velvet"^4, chocolate^2
Field grouping: tags:(+work -"fun-stuff")
Escaping: Escape characters +-&|!(){}[]^"~*?:\ with \, e.g. \+
Range search: properties.timestamp:[1587729413488 TO *] (inclusive), properties.title:{A TO Z}(excluding A and Z)
Combinations: chocolate AND vanilla, chocolate OR vanilla, (chocolate OR vanilla) NOT "vanilla pudding"
Field search: properties.title:"The Title" AND text
If You Rely On Task To Automatically Upload Artifacts, It Seems You Cannot Retrieve Them Using

If you rely on task to automatically upload artifacts, it seems you cannot retrieve them using task.artifacts - The task.artifacts is empty post upload, and when querying using Task.get_task(task_id='<SAME_ID>') the artifacts are also empty.

Only when explicitly calling task.upload_artifact(<artifact>,'artifact.pq') you can access it in the same session.

Logger shows S3 upload as expected, and the object is confirmed to exist in S3 and in ClearML GUI.

Is this intended?


Posted one year ago
Votes Newest

Answers 3

Thanks for the rapid reply

Posted one year ago

TrickyRaccoon92 , after looking in the documentation a bit more it seems I am a bit wrong.

If you look here: https://clear.ml/docs/latest/docs/references/sdk/task/#upload_artifact

there is a parameter called wait_on_upload which is set by default to false. If you set it to 'True' then you should get the artifact object as well. This is due to the artifacts upload being asynchronous and I'm guessing the artifact didn't finish uploading by the time you called it.

Posted one year ago

TrickyRaccoon92 , Hi!

Yes I believe this is the intended behavior. Since if you upload automatically you can upload many artifacts during a single run, whereas when you upload manually you create the object yourself.

Posted one year ago
3 Answers
one year ago
8 months ago