Proper way to upload artifacts
Good day to all. I am running a pytorch lightning notebook with ClearML integration. As was advised in lightning example just invoke task.init and all the rest will be logged. With regards to scalars and hyperparameters all fine but I have two problems:
To complete the task I used task.mark_completed()and got the following message:<clearml.backend_api.session.callresult.CallResult at 0x7f12f7c7b6d0> 2022-05-25 16:54:57,273 - clearml.Task - WARNING - ### TASK STOPPED - USER ABORTED - STATUS CHANGED ###Whatever I do after that - I can not restart jupyter kernel - it's dead.
2. All of lightning checkpoints are stored locally and were not uploaded to the server.
Please advise what's the proper way to handle those issues.
Thank you!
Hi GentleSwallow91 ,
task.close() - It will bring the same affect you're interested inTask.init() The parameter is output_uri. You can read more here - https://clear.ml/docs/latest/docs/references/sdk/task#taskinitTrue or provide a path to a bucket. The simplest usage would be Task.init(..., output_uri=True)