def train(self, epoch):
self.model.train()
train_loss = MovingAverageMeter()
train_acc = AccuracyMeter()
for i, (x, y) in enumerate(self.train_loader):
x = Variable(x)
y = Variable(y)
if self.use_cuda:
x = x.cuda()
y = y.cuda()
output = self.model(x)
loss = F.cross_entropy(output, y)
self.optimizer.zero_grad()
loss.backward()
self.optimizer.step()
train_loss.update(float(loss.data))
y_pred = output.data.max(dim=1)[1]
correct = int(y_pred.eq(y.data).cpu().sum())
train_acc.update(correct, x.size(0))
return train_loss.average, train_acc.accuracy
def train(self, epoch):
self.model.train()
train_loss = AverageMeter()
train_acc = AccuracyMeter()
for i, (x, y) in enumerate(self.train_loader):
x = Variable(x)
y = Variable(y)
if self.use_cuda:
x = x.cuda()
y = y.cuda()
output = self.model(x)
loss = F.cross_entropy(output, y)
self.optimizer.zero_grad()
loss.backward()
clip_grad_norm(self.optimizer, max_norm=1) #防止梯度爆炸
self.optimizer.step()
train_loss.update(float(loss.data), x.size(0))
y_pred = output.data.max(dim=1)[1]
#correct = int(y_pred.eq(y.data).cpu().sum())
_, correct, _ = get_accuracy(y.data, y_pred)
train_acc.update(correct, x.size(0))
if i % 100 == 0:
print(
'\nTrain Epoch/batch| [{}/{}]: Average batch loss:{:.6f},acc: {:.6f}\n'
.format(epoch, i, train_loss.average, train_acc.accuracy))
#save_model_checkpoint(self.model,epoch,self.save)
return train_loss.average, train_acc.accuracy
def validate(self):
self.model.eval()
valid_loss = AverageMeter()
valid_acc = AccuracyMeter()
for i, (x, y) in enumerate(self.valid_loader):
x = Variable(x, volatile=True)
y = Variable(y).long()
if self.use_cuda:
x = x.cuda()
y = y.cuda()
output = self.model(x)
loss = F.cross_entropy(output, y)
valid_loss.update(float(loss.data))
y_pred = output.data.max(dim=1)[1]
correct = int(y_pred.eq(y.data).cpu().sum())
valid_acc.update(correct, x.size(0))
print('\nTrain Epoch [{}]: Average batch loss: {:.6f}\n'.format(epoch,valid_acc.accuracy))
return valid_loss.average, valid_acc.accuracy