self.record_config()
def record_config(self):
with open(os.path.join(self.log_dir, 'TrainConfig.json'), 'w') as f:
f.write(json.dumps(self.args))
def get_log_dir(self):
if hasattr(self, 'log_dir'):
return self.log_dir
else:
return None
def get_model_dir(self):
if hasattr(self, 'model_dir'):
return self.model_dir
else:
return None
def get_result_dir(self):
if hasattr(self, 'result_dir'):
return self.result_dir
else:
return None
if __name__ == "__main__":
args = Config().get_config()
logger = TrainLogger(args)
logger.record_config()
model_path = logger.get_model_dir()
print(model_path)
running_loss.update(loss.item(), label.size(0))
running_cindex.update(cindex, pair)
epoch_loss = running_loss.get_average()
epoch_cindex = running_cindex.get_average()
running_loss.reset()
running_cindex.reset()
model.train()
return epoch_loss, epoch_cindex
# %%
for fold in range(5):
config = Config()
args = config.get_config()
args['fold'] = fold
logger = TrainLogger(args)
logger.info(__file__)
data_root = args.get("data_root")
DATASET = args.get("dataset")
split_type = args.get("split_type")
save_model = args.get("save_model")
fold = args.get("fold")
fpath = os.path.join(data_root, DATASET)
dp = DataPrepared(fpath)
train_index, val_index, test_index = dp.read_sets(fold, split_type=split_type)
df = dp.get_data()
self.result_dir = os.path.join(load_dir, 'result')
create_dir([self.result_dir])
log_path = os.path.join(self.log_dir, 'Test.log')
super().__init__(log_path)
self.record_config()
def record_config(self):
with open(os.path.join(self.log_dir, 'TestConfig.json'), 'w') as f:
f.write(json.dumps(self.args))
def get_model_path(self):
if hasattr(self, 'model_path'):
return self.model_path
else:
return None
def get_result_dir(self):
if hasattr(self, 'result_dir'):
return self.result_dir
else:
return None
if __name__ == "__main__":
args = Config(train=False).get_config()
logger = TestLogger(args)
logger.record_config()
model_path = logger.get_model_path()
print(model_path)
label_list.append(label.detach().cpu().numpy())
running_loss.update(loss.item(), label.size(0))
pred = np.concatenate(pred_list, axis=0)
label = np.concatenate(label_list, axis=0)
epoch_cindex = get_cindex(label, pred)
epoch_loss = running_loss.get_average()
running_loss.reset()
return epoch_loss, epoch_cindex
# %%
config = Config(train=False)
args = config.get_config()
logger = TestLogger(args)
logger.info(__file__)
data_root = args.get("data_root")
DATASET = args.get("dataset")
split_type = args.get("split_type")
save_model = args.get("save_model")
fold = args.get("fold")
# %%
fpath = os.path.join(data_root, DATASET, 'CNN-CNN')
dp = DataPrepared(fpath)
train_index, val_index, test_index = dp.read_sets(fold, split_type=split_type)