# ## Create Model # config # If number of epochs has been passed in use that, otherwise default to 50 parser = argparse.ArgumentParser() parser.add_argument("-e", "--epochs", help="number of epochs to train", type=int) args = parser.parse_args() if args.epochs: epochs = args.epochs else: epochs = 50 batch_size = 64 train_files, total_records = get_training_data(what=dataset) ## Hyperparameters # Small epsilon value for the BN transform epsilon = 1e-8 # learning rate epochs_per_decay = 5 starting_rate = 0.001 decay_factor = 0.85 staircase = True # learning rate decay variables steps_per_epoch = int(total_records / batch_size) print("Steps per epoch:", steps_per_epoch)
# ## Create Model # config # If number of epochs has been passed in use that, otherwise default to 50 parser = argparse.ArgumentParser() parser.add_argument("-e", "--epochs", help="number of epochs to train", type=int) args = parser.parse_args() if args.epochs: epochs = args.epochs else: epochs = 50 batch_size = 64 train_files, total_records = get_training_data(type="new") ## Hyperparameters # Small epsilon value for the BN transform epsilon = 1e-8 # learning rate epochs_per_decay = 5 starting_rate = 0.001 decay_factor = 0.85 staircase = True # learning rate decay variables steps_per_epoch = int(total_records / batch_size) print("Steps per epoch:", steps_per_epoch)