def main():
tf.set_random_seed(1231)
np.random.seed(1231)
try:
args = get_args()
config = process_config(args.config)
except:
print("Add a config file using \'--config file_name.json\'")
exit(1)
makedirs(config.summary_dir)
makedirs(config.checkpoint_dir)
# set logger
path = os.path.dirname(os.path.abspath(__file__))
path1 = os.path.join(path, 'core/model.py')
path2 = os.path.join(path, 'core/train.py')
logger = get_logger('log', logpath=config.summary_dir+'/',
filepath=os.path.abspath(__file__), package_files=[path1, path2])
logger.info(config)
# load data
train_loader, test_loader = load_pytorch(config)
# define computational graph
sess = tf.Session()
model_ = Model(config, _INPUT_DIM[config.dataset], len(train_loader.dataset))
trainer = Trainer(sess, model_, train_loader, test_loader, config, logger)
trainer.train()
def main():
config = None
try:
args = get_args()
config = process_config(args.config)
if config is None:
raise Exception()
except:
print("Add a config file using \'--config file_name.json\'")
exit(1)
makedirs(config.summary_dir)
makedirs(config.checkpoint_dir)
logger = get_logger('log', logpath=config.summary_dir, filepath=os.path.abspath(__file__))
train_labelled_data_loader, train_unlabelled_data_loader, test_loader = load_pytorch(config)
model = SDNet(config.image_size, config.num_anatomical_factors, config.num_modality_factors, config.num_classes)
print(model)
trainer = Trainer(model, train_labelled_data_loader, train_unlabelled_data_loader, test_loader, config, logger)
if config.train:
trainer.train()
if config.validation:
trainer.resume(os.path.join(config.checkpoint_dir, 'model.pth'))
trainer.test_epoch(debug=False)
def main():
tf.set_random_seed(1231)
np.random.seed(1231)
config = None
try:
args = get_args()
config = process_config(args.config)
if config is None:
raise Exception()
except:
print("Add a config file using \'--config file_name.json\'")
exit(1)
makedirs(config.summary_dir)
makedirs(config.checkpoint_dir)
# set logger
path = os.path.dirname(os.path.abspath(__file__))
path1 = os.path.join(path, 'classification/model.py')
path2 = os.path.join(path, 'classification/train.py')
path3 = os.path.join(path, 'regression/model.py')
path4 = os.path.join(path, 'regression/train.py')
logger = get_logger('log',
logpath=config.summary_dir + '/',
filepath=os.path.abspath(__file__),
package_files=[path1, path2, path3, path4])
logger.info(config)
# Define computational graph.
sess = tf.Session()
if config.mode == "classification":
train_loader, test_loader = load_pytorch(config)
model_ = ClassificationModel(config,
_CLASSIFICATION_INPUT_DIM[config.dataset],
len(train_loader.dataset),
attack=False)
trainer = ClassificationTrainer(sess, model_, train_loader,
test_loader, config, logger)
elif config.mode == "regression":
train_loader, test_loader, std_train = generate_data_loader(config)
config.std_train = std_train
model_ = RegressionModel(config, _REGRESSION_INPUT_DIM[config.dataset],
len(train_loader.dataset))
trainer = RegressionTrainer(sess, model_, train_loader, test_loader,
config, logger)
else:
print("Please choose either 'classification' or 'regression'.")
raise NotImplementedError()
# choose one of the following four functions
# 1. train the model
trainer.train()
def main():
config = None
try:
args = get_args()
config = process_config(args.config)
if config is None:
raise Exception()
except:
print("Add a config file using \'--config file_name.json\'")
exit(1)
makedirs(config.summary_dir)
makedirs(config.checkpoint_dir)
logger = get_logger('log',
logpath=config.summary_dir,
filepath=os.path.abspath(__file__))
train_loader, test_loader = load_pytorch(config)
model = PolyGNN(state_dim=128,
n_adj=4,
coarse_to_fine_steps=config.coarse_to_fine_steps,
get_point_annotation=False)
trainer = Trainer(model, train_loader, test_loader, config, logger)
if config.train:
trainer.train()
if config.validation:
trainer.resume(os.path.join(config.checkpoint_dir, 'model.pth'))
trainer.test_epoch(cur_epoch=999, plot=True)
def gradient_check():
tf.set_random_seed(1231)
np.random.seed(1231)
try:
args = get_args()
config = process_config(args.config)
except:
print("Add a config file using \'--config file_name.json\'")
exit(1)
# set logger
path = os.path.dirname(os.path.abspath(__file__))
path1 = os.path.join(path, 'core/model.py')
path2 = os.path.join(path, 'core/train.py')
logger = get_logger('log', logpath=config.summary_dir+'/',
filepath=os.path.abspath(__file__), package_files=[path1, path2])
logger.info(config)
batch_sizes = [1,4,16,32,64,128,256,512,1024]
precon = False
for bs in batch_sizes:
start_time = time.time()
print("processing batch size {}".format(bs))
# load data
train_loader, test_loader = load_pytorch(config)
# define computational graph
sess = tf.Session()
model_ = Model(config, _INPUT_DIM[config.dataset], len(train_loader.dataset))
trainer = Trainer(sess, model_, train_loader, test_loader, config, logger)
trainer.grad_check(sess, bs, precon)
print('batch size {} takes {} secs to finish'.format(
bs, time.time()-start_time))
tf.reset_default_graph()
precon = True
for bs in batch_sizes:
start_time = time.time()
print("processing batch size {}".format(bs))
# load data
train_loader, test_loader = load_pytorch(config)
# define computational graph
sess = tf.Session()
model_ = Model(config, _INPUT_DIM[config.dataset], len(train_loader.dataset))
trainer = Trainer(sess, model_, train_loader, test_loader, config, logger)
trainer.grad_check(sess, bs, precon)
print('batch size {} takes {} secs to finish'.format(
bs, time.time()-start_time))
tf.reset_default_graph()
def main():
tf.set_random_seed(1231)
np.random.seed(1231)
config = None
try:
args = get_args()
config = process_config(args.config)
if config is None:
raise Exception()
except:
print("Add a config file using \'--config file_name.json\'")
exit(1)
makedirs(config.summary_dir)
makedirs(config.checkpoint_dir)
# set logger
path = os.path.dirname(os.path.abspath(__file__))
path1 = os.path.join(path, 'classification/model.py')
path2 = os.path.join(path, 'classification/train.py')
path3 = os.path.join(path, 'regression/model.py')
path4 = os.path.join(path, 'regression/train.py')
path5 = os.path.join(path, args.config)
logger = get_logger('log',
logpath=config.summary_dir + '/',
filepath=os.path.abspath(__file__),
package_files=[path1, path2, path3, path4, path5])
logger.info(config)
# Define computational graph.
sess = tf.Session()
if config.mode == "classification":
train_loader, test_loader = load_pytorch(config)
model_ = ClassificationModel(config,
_CLASSIFICATION_INPUT_DIM[config.dataset],
len(train_loader.dataset), config.mode)
trainer = ClassificationTrainer(sess, model_, train_loader,
test_loader, config, logger)
elif config.mode == "segmentation":
train_loader, test_loader = load_pytorch(config)
model_ = ClassificationModel(
config, [config.image_size, config.image_size, 1],
config.total_num_images * config.image_size * config.image_size,
config.mode, config.num_classes)
trainer = ClassificationTrainer(sess, model_, train_loader,
test_loader, config, logger)
elif config.mode == "regression":
train_loader, test_loader, std_train = generate_data_loader(config)
config.std_train = std_train
model_ = RegressionModel(config, _REGRESSION_INPUT_DIM[config.dataset],
len(train_loader.dataset))
trainer = RegressionTrainer(sess, model_, train_loader, test_loader,
config, logger)
else:
print("Please choose either 'classification' or 'regression'.")
raise NotImplementedError()
if config.train:
trainer.train()
if config.validation:
trainer.load_checkpoint(config.checkpoint)
trainer.test_epoch_with_misc_metrics()
def main():
tf.set_random_seed(1231)
np.random.seed(1231)
config = None
try:
args = get_args()
config = process_config(args.config)
if config is None:
raise Exception()
except:
print("Add a config file using \'--config file_name.json\'")
exit(1)
makedirs(config.summary_dir)
makedirs(config.checkpoint_dir)
# set logger
path = os.path.dirname(os.path.abspath(__file__))
path3 = os.path.join(path, 'regression/model.py')
path4 = os.path.join(path, 'regression/train.py')
logger = get_logger('log',
logpath=config.summary_dir + '/',
filepath=os.path.abspath(__file__),
package_files=[path3, path4])
logger.info(config)
# Define computational graph
rmse_results, ll_results = [], []
n_runs = 10
for i in range(1, n_runs + 1):
sess = tf.Session()
# Perform data splitting again with the provided seed.
train_loader, test_loader, std_train = generate_data_loader(config,
seed=i)
config.std_train = std_train
model_ = RegressionModel(config, _REGRESSION_INPUT_DIM[config.dataset],
len(train_loader.dataset))
trainer = RegressionTrainer(sess, model_, train_loader, test_loader,
config, logger)
trainer.train()
rmse, ll = trainer.get_result()
rmse_results.append(float(rmse))
ll_results.append(float(ll))
tf.reset_default_graph()
for i, (rmse_result, ll_result) in enumerate(zip(rmse_results,
ll_results)):
logger.info("\n## RUN {}".format(i))
logger.info('# Test rmse = {}'.format(rmse_result))
logger.info('# Test log likelihood = {}'.format(ll_result))
logger.info("Results (mean/std. errors):")
logger.info("Test rmse = {}/{}".format(np.mean(rmse_results),
np.std(rmse_results) / n_runs**0.5))
logger.info("Test log likelihood = {}/{}".format(
np.mean(ll_results),
np.std(ll_results) / n_runs**0.5))