Can't override Trainer defaults via env variables for LightningCLI

[awaelchli]

Bug description

Using the env vars PL_TRAINER_... to override trainer defaults doesn't work if you are using the LightningCLI. This is because the cli handles the defaults differently and prioritizes them over the env variable.

I need to make the logic to determine these settings consistent with how we do it in Fabric.

What version are you seeing the problem on?

v2.1

How to reproduce the bug

I will immediately work on it and know how to repro/fix it.

Error messages and logs

# Error messages and logs here please

Environment

Current environment

#- Lightning Component (e.g. Trainer, LightningModule, LightningApp, LightningWork, LightningFlow):
#- PyTorch Lightning Version (e.g., 1.5.0):
#- Lightning App Version (e.g., 0.5.2):
#- PyTorch Version (e.g., 2.0):
#- Python version (e.g., 3.9):
#- OS (e.g., Linux):
#- CUDA/cuDNN version:
#- GPU models and configuration:
#- How you installed Lightning(`conda`, `pip`, source):
#- Running environment of LightningApp (e.g. local, cloud):

More info

No response

cc @justusschock @awaelchli

[carmocca]

The CLI uses its own environment variables: https://lightning.ai/docs/pytorch/stable/cli/lightning_cli_advanced_2.html#run-from-environment-variables-directly

[awaelchli]

Good to know @carmocca. This feature is only enabled when you explicitly pass parser_kwargs={"default_env": True} to the CLI. For the fix I took a different approach now, because the env vars need to apply without the user having to change the code.