Is "Prepare a config file for the CLI" out of date? about pytorch-lightning HOT 2 OPEN

If you define your cli as cli = LightningCLI(pl.LightningModule, BoringDataModule, subclass_mode_model=True) or cli = LightningCLI(datamodule_class=BoringDataModule) and then run:

python prepare_config.py fit --model DemoModel --print_config

you get the expected result. If you define the cli as you show, then you need to drop the --model DemoModel for it not to error and give you the expected result.

I think what the documentation is suggesting when it shows the example you're using is that if no model is passed to the LightningCLI -- "By default, no model is selected, meaning the printed config will not include model settings. To get a config with the default values of a particular model would be: python main.py fit --model DemoModel --print_config". I think what isn't clear is that the cli setup is different from what is shown directly above with the hardcoded DemoModel passed in.

from pytorch-lightning.

If you define your cli as cli = LightningCLI(pl.LightningModule, BoringDataModule, subclass_mode_model=True) or cli = LightningCLI(datamodule_class=BoringDataModule) and then run:

python prepare_config.py fit --model DemoModel --print_config

you get the expected result. If you define the cli as you show, then you need to drop the --model DemoModel for it not to error and give you the expected result.

I think what the documentation is suggesting when it shows the example you're using is that if no model is passed to the LightningCLI -- "By default, no model is selected, meaning the printed config will not include model settings. To get a config with the default values of a particular model would be: python main.py fit --model DemoModel --print_config". I think what isn't clear is that the cli setup is different from what is shown directly above with the hardcoded DemoModel passed in.

@synapticarbors Thanks! Now I totally get it.

from pytorch-lightning.