How to load a config.yaml file for training?

[SilverStarCoder]

In the older version of anomalib, I could load a config.yaml file using the get_configurable_parameters function from anomalib.config, but this is no longer available in the latest version. Will this be added in the future? Or alternatively, how else can I train using the config file, because this code doesn't work: engine.fit(datamodule=datamodule, model=model, config="config.yaml"). In the engine.py file, for fit, it says. If you have a ready configuration file, run it like this: anomalib fit --config <config_file_path, but surely that doesn't work as config is not a function argument.

[samet-akcay]

Hi @SilverStarCoder, yes, this feature is currently not available, but will be introduced. It will be something like;

data.from_config()
model.from_config()
engine.from_config()

so each component could be instantiated from a config file.

@ashwinvaidya17 and @harimkang, looks like this is of interest

[SilverStarCoder]

Thanks @samet-akcay for the response, so just double checking: I cannot at the moment train using a config.yaml file via API? I will have to input each one of the parameters (excluding the datamodule relevant ones) into the Engine class like this:

engine = Engine(
    callbacks=callbacks,
    pixel_metrics="AUROC",
    accelerator="auto", 
    devices=1,
    logger=False,
    ..........
)

[samet-akcay]

If you want to use the API, yes, this is the current approach. @ashwinvaidya17, do you want to add anything here?