Training on Custom Dataset.

[Amith4504]

After going through the training code in train_script.py I came across the following piece of code. trainer.train(cfg.TRAIN.EPOCH, load_latest=True, fail_safe=True) found here. This defaults to loading the previous checkpoints stored when you have to resume training. Am i wrong in my assumption? If I have to train the network from scratch should I set it to False? And further make it configurable.