subprocess_runner

This file contains some utilities that are useful to run a function inside an independent sub-process.

When the subprocess is finished the memory is fully released, which is useful to avoid memory leaks, especially from tensorflow.

tensiometer.utilities.subprocess_runner.run_in_process(**kwargs)

Decorator to run a function in a subprocess with optional feedback, monitoring, and timeout capabilities.

Parameters:

• subprocess – whether to execute in a subprocess.

• feedback_level – verbosity level from 0 to 3.

• context – multiprocessing context (fork, spawn or forkserver).

• monitoring – enable monitoring of the subprocess.

• monitoring_frequency – interval, in seconds, for monitoring updates.

• timeout – maximum runtime in seconds; enables monitoring when set.

Returns:

wrapped function with subprocess capabilities.

Raises:

ValueError – if supplied configuration values are invalid.