How to redirect all error messages to logger

Question

I want to capture all error messages and warnings that are triggered from a library (in my case pytorch_lightning) in a log file. My current setup only captures some of the warnings, but I would like to capture all the warnings/errors that PyTorch sends to stdout. How can I do this?

import logging

# Config logger
logging.basicConfig(
  encoding='utf-8',
  level=logging.DEBUG,
  filename=log_path,)

# Redirect all warnings to log file
logging.captureWarnings(True)

# Redirect any error messages to log file???
# ....
# ....

# Pytorch code that will lead to warnings/errors that I
# want stored in the logger file
# ....

The above code will catch only warnings like the following:

WARNING:py.warnings:C:\Users\jared\Anaconda3\envs\base-torch\lib\site-packages orch\functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at ..\aten\src\ATen ative\TensorShape.cpp:2157.) return _VF.meshgrid(tensors, **kwargs) # type: ignore[attr-defined]

but will not capture warnings like the following:

WARNING: The shape inference of prim::Constant type is missing, so it may result in wrong shape inference for the exported graph. Please consider adding it in symbolic functio in symbolic function.

How to redirect all error messages to logger

0 Answers