def mlnet_method(X, Y, splits=5):
accuracy, cnt = 0, 0
mskf = my_SKF(X, Y, splits)
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
for X_train, Y_train, X_test, Y_test in mskf:
new_X = transform(X_train, Y_train)
dataset = ImageDataset(new_X)
dataloader = DataLoader(dataset,
batch_size=128,
shuffle=True,
num_workers=4)
net = SiameseNetwork().to(device)
net.fit(dataloader)
net.eval()
hit = 0
for i, x in enumerate(X_test):
y_pred = knn_pred(net, X_train, Y_train, x, batch_num=1)
y_real = Y_test[i]
if y_pred == y_real:
hit += 1
acc = hit / X_test.shape[0]
accuracy += acc
cnt += 1
print('Split:{}, Acc:{:.4f}'.format(cnt, acc))
return accuracy / splits
def get_siamese_model(nclasses):
device = torch.device("cuda")
model = SiameseNetwork()
# model.classifier.classifier = nn.Sequential()
checkpoint = torch.load('./model/siamese/net_59.pth')
model.load_state_dict(checkpoint['model_state_dict'])
model.to(device)
model.eval()
return model
print("metric gogo : {} ".format(last_loss))
#best model check - DBY
best_model = (last_loss > best_loss)
if best_model:
save_model("best_" + str(epoch + 1), model, optimizer,
scheduler)
best_loss = last_loss
print("best epoch {} best loss {} ".format(
str(epoch + 1), last_loss))
last_loss = validate(best_prediction_file, model,
validate_dataloader, validate_label_file,
cuda)
print("best prediction file save")
time_ = datetime.datetime.now()
elapsed = datetime.datetime.now() - time_
print('[epoch {}] elapsed: {}'.format(epoch + 1, elapsed))
# save model
#if (epoch+1) % 10 == 0:
# save_model(str(epoch + 1), model, optimizer)
elif mode == 'test':
model.eval()
# accuracy 확인
test_dataloader, _ = data_loader(root=DATASET_PATH,
phase='test',
batch_size=1)
prediction_file = "prediction_test.txt"
test(prediction_file, model, test_dataloader, cuda)