Training curve from the current MAAS deep-policy demo run.
{
"epoch": 6.0,
"train_loss": 0.02796567932647817,
"val_loss": 0.08078384151061375,
"train_condition_acc": 0.9963235294117647,
"val_condition_acc": 0.9791666666666666,
"train_urgency_acc": 1.0,
"val_urgency_acc": 1.0,
"lr": 0.0
}