Hi. Does ClearML support auto-magical (scalars) for scenarios where we need to execute third party installed toolkits, which runs training with common ML frameworks such as pytorch and tensorflow? Two examples below. ClearML can capture the console output but does not do so for the scalar outputs.

Example one. We are only allowed to execute externalScriptWithouClearMLInitcodes.py but has no write access.
` from clearml import Task
import subprocess
import os

task = Task.init(....)
subprocess.run(['python', 'externalScriptWithouClearMLInitcodes.py']) Example two. We are given a executable that runs a pytorch training code in the backend. from clearml import Task
import subprocess
import os

task = Task.init(....)
subprocess.run(['executablefile']) `