def run():
global model
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# Create dataset
template_dataset = TemplateDataset(config)
training_loader, validation_loader, test_loader = template_dataset.get_loaders(
)
# Create the neural network
model = NN(net, optimizer, loss_function, lr_scheduler, metric, device,
config).to(device)
# Create the data handler
data_handler = DataHandler(training_loader, validation_loader, test_loader)
for epoch in range(config['epochs']):
# Training
model.train()
for i, data in enumerate(training_loader, 0):
x, y = data
x, y = x.to(device), y.to(device)
y_hat = model(x)
loss = model.backpropagate(y_hat, y)
result = model.evaluate(y_hat, y)
data_handler.train_loss.append(loss)
data_handler.train_metric.append(result)
with torch.no_grad():
model.eval()
# Validating
if validation_loader is not None:
for i, data in enumerate(validation_loader, 0):
x, y = data
x, y = x.to(device), y.to(device)
y_hat = model(x)
_, loss = model.calculate_loss(y_hat, y)
result = model.evaluate(y_hat, y)
data_handler.valid_loss.append(loss)
data_handler.valid_metric.append(result)
# Testing
if test_loader is not None:
for i, data in enumerate(test_loader, 0):
x, y = data
x, y = x.to(device), y.to(device)
y_hat = model(x)
_, loss = model.calculate_loss(y_hat, y)
result = model.evaluate(y_hat, y)
data_handler.test_loss.append(loss)
data_handler.test_metric.append(result)
model.lr_scheduler_step()
data_handler.epoch_end(epoch, model.get_lr())
data_handler.plot(loss=config['plot']['loss'],
metric=config['plot']['metric'])
