if seed is not None:
train_config["seed"] = seed
train_config["gqcnn"]["seed"] = seed
if tensorboard_port is not None:
train_config["tensorboard_port"] = tensorboard_port
gqcnn_params = train_config["gqcnn"]
# Create a unique output folder based on the date and time.
if save_datetime:
# Create output dir.
unique_name = time.strftime("%Y%m%d-%H%M%S")
output_dir = os.path.join(output_dir, unique_name)
utils.mkdir_safe(output_dir)
# Set visible devices.
if "gpu_list" in train_config:
gqcnn_utils.set_cuda_visible_devices(train_config["gpu_list"])
# Fine-tune the network.
start_time = time.time()
gqcnn = get_gqcnn_model(backend)(gqcnn_params)
trainer = get_gqcnn_trainer(backend)(gqcnn,
dataset_dir,
split_name,
output_dir,
train_config,
name=name)
trainer.finetune(model_dir)
logger.info("Total Fine-tuning Time: " +
str(utils.get_elapsed_time(time.time() - start_time)))
if not os.path.isabs(dataset_dir):
dataset_dir = os.path.join(os.getcwd(), dataset_dir)
if not os.path.isabs(output_dir):
output_dir = os.path.join(os.getcwd(), output_dir)
if not os.path.isabs(config_filename):
config_filename = os.path.join(os.getcwd(), config_filename)
# create output dir if necessary
utils.mkdir_safe(output_dir)
# open train config
train_config = YamlConfig(config_filename)
train_config["tensorboard_port"] = tensorboard_port
network_params = train_config["siamese_net"]
if name is None:
# create a unique model name with a timestamp
timestamp = time.strftime("%Y-%m-%d_%H-%M-%S")
name = "model_{}".format(timestamp)
# train the network
start_time = time.time()
network = SiameseNet(network_params)
trainer = SiameseTrainer(network,
dataset_dir,
output_dir,
train_config,
model_name=name)
trainer.train()
logger.info("Total Training Time:" + str(utils.get_elapsed_time(time.time() - start_time)))
