"validate_test_split": args.validate_test_split,
"augment": False,
"shuffle": False,
"seed": args.random_seed
}
validation_generator = DataGenerator("validate", args.data_path,
**validation_data_params)
if (hvd.rank() == 0):
validation_generator.print_info()
# Fit the model
# Do at least 3 steps for training and validation
steps_per_epoch = max(
3,
training_generator.get_length() // (args.bz * hvd.size()))
validation_steps = max(
3, 3 * training_generator.get_length() // (args.bz * hvd.size()))
unet_model.model.fit_generator(
training_generator,
steps_per_epoch=steps_per_epoch,
epochs=args.epochs,
verbose=verbose,
validation_data=validation_generator,
#validation_steps=validation_steps,
callbacks=callbacks,
max_queue_size=1, #args.num_prefetched_batches,
workers=1, #args.num_data_loaders,
use_multiprocessing=True)
"validate_test_split": args.validate_test_split,
"augment": False,
"shuffle": False,
"seed": args.random_seed
}
validation_generator = DataGenerator("validate", args.data_path,
**validation_data_params)
if (hvd.rank() == 0):
validation_generator.print_info()
# Fit the model
# Do at least 3 steps for training and validation
steps_per_epoch = max(
3,
training_generator.get_length() // (args.bz * hvd.size()))
validation_steps = max(
3,
validation_generator.get_length() // (args.bz * hvd.size()))
unet_model.model.fit_generator(training_generator,
steps_per_epoch=steps_per_epoch,
epochs=args.epochs,
verbose=verbose,
validation_data=validation_generator,
validation_steps=validation_steps,
callbacks=callbacks,
max_queue_size=args.num_prefetched_batches,
workers=args.num_data_loaders,
use_multiprocessing=False)
