that's alright, it's just the description of the plugin that had me confused 🙂

But, if you like, you can connect a remote interpreter and debug with PyCharm, locally, without clearml-agent

TL;DR - no 😕

when using task.execute_remotely() , as you wrote, it won’t run on you local environment but with the clearml-agent. When running on the clearml-agent, you task should be stable for running and create a model - for optimize, pipeline and more, but not for debugging.

When running development mode, on your local machine (w/ or w/o remote interpreter), you can debug you code. There are some differences between the two runs - running from a clearml-agent won’t re using last task ID like dev mode is, the configuration will be like in the UI and not according to the original task and some more differences.

I am wondering whether this scenario (executing remotely on an agent that spawns a docker container + debugging from pycharm) is supported at all

https://allegro.ai/docs/faq/faq/#remote-debugging-trains-pycharm-plugin this is for not getting the git information in the task, but I understand your task logged the changes and the git information.

Are you looking for ssh interpreter in your PyCharm?

Yes, this is the very plugin. Everything syncs correctly and is started remotely on the respective agent (including my local changes).
What has me confused is that the Github page of the plugin, and also https://allegro.ai/docs/faq/faq/ , sort of imply, that it is used for "debugging".

I am wondering what the workflow in that case is. Would I start a python debug server and have the code connect back to it?

Hi ElegantDeer55 ,

Are you referring to https://github.com/allegroai/trains-pycharm-plugin ? If so, it should sync you .git folder to the remote machine so the task will log the git.

Basically I am confused if “remote debugging” should work / kick in automatically when running in docker mode and starting a task like this:

 from trains import Task
task = Task.init(project_name=“my project”, task_name=“my task”)
task.execute_remotely()

When you are running this code from you PyCharm, it should not run it from you local machine, but create a task and run it with the clearml-agent. Is this what you a looking for? If not, can you share about your use-case?