train_start = '2015-01-01 01:30:00'
train_end = '2019-03-01 15:00:00'
cv_start = '2019-03-02 01:30:00'
cv_end = '2019-12-01 15:00:00'
test_start = '2019-12-02 01:30:00'
test_end = '2021-04-20 15:00:00'
data_processor = DataProcessor()
train_loader = data_processor.get_data(train_start, train_end)
cv_loader = data_processor.get_data(cv_start, cv_end)
test_loader = data_processor.get_data(test_start, test_end)
model = CNN(transform='GAF')
model = model.double()
loss_fn = torch.nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(), lr=1e-3)
max_epochs = 100
accuracy_list = []
for t in range(max_epochs):
print(f"Epoch {t+1}\n-------------------------------")
train_loop(train_loader, model, loss_fn, optimizer)
acc = test_loop(cv_loader, model, loss_fn)
accuracy_list.append(acc)
if len(accuracy_list) > 8:
print(accuracy_list[-8:])
if len(accuracy_list) > 8 and max(
if __name__ == "__main__":
ip = IndexProcessor()
data, y = ip.get_data()
images = ip.apply_gaf(data)
print(images.shape)
print(y.shape)
train_loader, cv_loader, test_loader = ip.split_data(images, y)
use_gpu = torch.cuda.is_available()
print('gpu:', use_gpu)
if use_gpu:
model = CNN(transform='GAF')
model = model.double().cuda()
loss_fn = torch.nn.CrossEntropyLoss().cuda()
optimizer = torch.optim.Adam(model.parameters(), lr=1e-3)
max_epochs = 100
accuracy_list = []
for t in range(max_epochs):
print(f"Epoch {t+1}\n-------------------------------")
train_loop(train_loader, model, loss_fn, optimizer)
acc = test_loop(cv_loader, model, loss_fn)
accuracy_list.append(acc)
if len(accuracy_list) > 10:
print(accuracy_list[-10:])
if len(accuracy_list) > 10 and max(