031392ff2cbb5703558d17a676a456521f754190,mnist/main.py,,train,#Any#,67

Before Change

        loss = loss.data[0]
        optimizer.step()
        print("Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.4f}".format(epoch,
            i+BATCH_SIZE, training_data.size(0),
            float(i+BATCH_SIZE)/training_data.size(0)*100, loss))

def test(epoch):

After Change

def train(epoch):
    // create buffers for mini-batch
    batch_data = torch.FloatTensor(opt.batchSize, 1, 28, 28)
    batch_targets = torch.LongTensor(opt.batchSize)
    if cuda:
        batch_data, batch_targets = batch_data.cuda(), batch_targets.cuda()

    // create autograd Variables over these buffers
    batch_data, batch_targets = Variable(batch_data), Variable(batch_targets)

    for i in range(0, training_data.size(0)-opt.batchSize+1, opt.batchSize):
        start, end = i, i+opt.batchSize
        optimizer.zero_grad()
        batch_data.data[:] = training_data[start:end]
        batch_targets.data[:] = training_labels[start:end]
        output = model(batch_data)

Instances