Executing clearml-task from cli with "-m" modules

[EladDvash]

Describe the bug

When trying to use "python -m ..." to execute a module in the clearml-task cli tool it rejects the task even though the backend can run that task when cloning and executing from the web UI

To reproduce

Exact steps to reproduce the bug. Provide example code if possible.

1. Create a new folder with the name "mymodule"
2. Add an empty init.py file in the folder
3. Add a file callde train.py in the folder with some clearml code (e.g. pytorch mnist example)
4. Add a requirements.txt
5. Run:
   clearml-task --project test --name task_test --skip-task-init --folder . --requirements requirements.txt --script "-m mymodule.train" --args epochs=2 --queue default

Expected behaviour

What is the expected behaviour? What should've happened but didn't?

• The task should launch without issue and be added to the queue

Environment

• Server type: self hosted
• ClearML SDK Version: 1.16.1
• ClearML Server Version (Only for self hosted): 1.13.0-414
• Python Version: 3.10.12
• OS: Linux

[eugen-ajechiloae-clearml]

Hi @EladDvash ! We are taking a look. It seems that we can actually support modules to run (the agent can properly run them and the backend allows them as well), but the SDK can't create a task from a module yet. We will soon patch this.