return t, [index]

def __len__(self):

return len()

def cnn_classification():

batch_size = 256

trainDataLoader = DataLoader(TrainingDataSet(), batch_size=batch_size, shuffle=False)

testDataLoader = DataLoader(TestDataSet(), batch_size=batch_size, shuffle=False)

epoch_num = 200

#lr = 

lr = 

net = VGGBaseSimpleS2().to(device)

print(net)

# loss

loss_func = ()

# optimizer

optimizer = ((), lr=lr)

# optimizer = ((), lr=lr, momentum=, weight_decay=5e-4)

scheduler = .StepLR(optimizer, step_size=5, gamma=)

if not (“logCNN“):

(“logCNN“)

writer = (“logCNN“)

for epoch in range(epoch_num):

train_sum_loss = 0

train_sum_correct = 0

train_sum_fp = 0

train_sum_fn = 0

train_sum_tp = 0

train_sum_tn = 0

for i, data in enumerate(trainDataLoader):

()

inputs, labels = data

inputs = (1).to()

labels = ()

inputs, labels = (device), (device)

outputs = net(inputs)

loss = loss_func(outputs, labels)

()

()

()

_, pred = (, dim=1)

acc = ().cpu().sum()

one = (labels)

zero = (labels)

tn = ((labels == zero) * (pred == zero)).sum()

tp = ((labels == one) * (pred == one)).sum()

fp = ((labels == zero) * (pred == one)).sum()

fn = ((labels == one) * (pred == zero)).sum()

train_sum_fn += ()

train_sum_fp += ()

train_sum_tn += ()

train_sum_tp += ()

train_sum_loss += ()

train_sum_correct += ()

train_loss = train_sum_loss *  / len(trainDataLoader)

train_correct = train_sum_correct *  / len(trainDataLoader) / batch_size

train_precision = train_sum_tp *  / (train_sum_fp + train_sum_tp)

train_recall = train_sum_tp *  / (train_sum_fn + train_sum_tp)

(“train loss“, train_loss, global_step=epoch)

(“train correct“,

train_correct, global_step=epoch)

(“train precision“,

train_precision, global_step=epoch)

(“train recall“, train_recall, global_step=epoch)

if not (“models_aug_CNN“):

(“models_aug_CNN“)

((), “models_aug_CNN/{}.pth“.format(epoch + 1))