def main(args):
address_parser = AddressParser(model_type=args.model_type, device=0)
train_container = PickleDatasetContainer(args.train_dataset_path)
lr_scheduler = StepLR(step_size=20)
address_parser.retrain(train_container,
0.8,
epochs=args.epochs,
batch_size=args.batch_size,
num_workers=6,
learning_rate=args.learning_rate,
callbacks=[lr_scheduler],
logging_path=f"./chekpoints/{args.model_type}")
test_container = PickleDatasetContainer(args.test_dataset_path)
checkpoint = "best"
address_parser.test(test_container,
batch_size=args.batch_size,
num_workers=4,
logging_path=f"./chekpoints/{args.model_type}",
checkpoint=checkpoint)
def main(args):
address_parser = AddressParser(model_type=args.model_type, device=0)
if args.mode in ("train", "both"):
train_container = PickleDatasetContainer(args.train_dataset_path)
lr_scheduler = StepLR(step_size=20)
address_parser.retrain(train_container,
0.8,
epochs=100,
batch_size=1024,
num_workers=6,
learning_rate=0.001,
callbacks=[lr_scheduler],
logging_path=f"./chekpoints/{args.model_type}")
if args.mode in ("test", "both"):
test_container = PickleDatasetContainer(args.test_dataset_path)
if args.mode == "test":
checkpoint = handle_pre_trained_checkpoint(args.model_type)
else:
checkpoint = "best"
address_parser.test(test_container,
batch_size=2048,
num_workers=4,
logging_path=f"./chekpoints/{args.model_type}",
checkpoint=checkpoint)
def test_exception_is_thrown_on_optimizer_argument(self):
with self.assertRaises(ValueError):
StepLR(self.optimizer, step_size=3)
def test_step_lr_integration(self):
step_lr = StepLR(step_size=3)
self._fit_with_callback_integration(step_lr)