def start(self):
train_loader, valid_loader = self.read_dataset()
#create network
self.NET = self.get_network()
print("#net params:", sum(p.numel() for p in self.NET.parameters()))
if self.USE_CUDA:
self.NET.cuda()
optimizer = Optimizer.get_optimizer(self.OPTIMIZER, self.NET,
self.LEARNING_RATE)
criterion = Loss.get_loss(self.LOSS)
scheduler = ExponentialLR(optimizer, gamma=0.95)
self.__train(train_loader, valid_loader, optimizer, criterion,
scheduler)
def train(self, loader):
net = Network_AE(vocab_size=len(self.VOCAB), drop_out=self.DROP_OUT)
if self.USE_CUDA:
net.cuda()
optimizer = Optimizer.get_optimizer(self.OPTIMIZER, net,
self.LEARNING_RATE)
criterion = Loss.get_loss(self.LOSS)
#train
plot_train_loss = []
net.train()
for epoch in range(1, self.EPOCH + 1):
train_loss = 0.0
for batch_idx, (X, y) in enumerate(loader):
y = torch.squeeze(y)
if self.USE_CUDA:
X = Variable(X.cuda())
y = Variable(y.cuda())
else:
X = Variable(X)
y = Variable(y)
output = net(X)
loss = criterion(output, y)
train_loss += loss.data[0]
optimizer.zero_grad()
loss.backward()
optimizer.step()
print("epoch:{}, train_loss:{}".format(epoch, train_loss))
plot_train_loss.append(train_loss)
#plot
self.plot(plot_train_loss)
def start(self):
train_loader, valid_loader = self.read_dataset()
optimizer = Optimizer.get_optimizer(self.OPTIMIZER, self.NET, self.LEARNING_RATE)
criterion = Loss.get_loss(self.LOSS)
scheduler = ExponentialLR(optimizer, gamma=0.95)
self.__train(train_loader, valid_loader, optimizer, criterion, scheduler)
