Add WandB group name to training runs

[bfhealy]

As we run multiple iterations of training for each binary classifier, we will want to compare the different runs while maintaining organization between them. I think we should associate each run with a group that shares a common DNN architecture, hyperparameters and training set using e.g. wandb.init(group="experiment_1", job_type="eval") (https://docs.wandb.ai/guides/track/advanced/grouping).

[AshishMahabal]

Good idea

[guiga004]

@bfhealy - is there a default group and job_type we want to use when not specified?

[bfhealy]

@guiga004 Perhaps the default group could be 'experiment' and job_type 'train'?