wandb config for models

[espinielli]

from run_wandb.py it is not clear which models are defined in wandb_config...or at least I was not able to find the relevant info. (yes I see there is s models/ directory but it is not clear what is used)

[wurzelDeveloper]

run_wandb.py by default uses the AutogluonModel (see lines 63 to 70):

model_config = {
    "time_limit": args.time,
    "preset": args.quality,
    "verbosity": 2,
}

evaluator = MetricEvals()
model = AutogluonModel(**model_config)
loader = DataLoader(Path("data"))

Where args are command line arguments to configure the training (use run_wandb.py --help for more info).

The model is also explained in the Project Overview. I'll keep this issue open as a reminder for us to improve the documentation, as it seems not clear enough right now.

[wurzelDeveloper]

• [ ] Consider updating the readme with more information on how to run the training with command line arguments and with information about what model is used

[niclaswue]

Yes, just as an additional comment, we started with the simpler sklearn model class and the average ensemble, however autogluon does all that under the hood and outperformed the sklearn models by a large margin. Therefore we only used Autogluon afterwards. It would also be an option to delete the old model classes.