3.6节train_ch3函数loss计算和sgd问题 #155
Description
bug描述
3.6节train_ch3函数，如果传入的loss已经是求过平均的，train_l_sum每次只累加一个batch的平均值，最后却除以总样本数，打印的loss结果就会很小，例如3.9节的调用。
如果每个batch不一样大（例如Fashion-MNIST设置batch_size=256时，最后一个batch是96），当optimizer=None时，默认的sgd传入batch_size应该会在最后一个batch造成误差，似乎应该使用y.shape[0]。
另外train_ch5是用batch_count，如果batch大小不一致，最后打印的loss也应该会有微小误差。
def train_ch3(net, train_iter, test_iter, loss, num_epochs, batch_size,
params=None, lr=None, optimizer=None):
for epoch in range(num_epochs):
train_l_sum, train_acc_sum, n = 0.0, 0.0, 0
for X, y in train_iter:
y_hat = net(X)
l = loss(y_hat, y).sum()
# 梯度清零
if optimizer is not None:
optimizer.zero_grad()
elif params is not None and params[0].grad is not None:
for param in params:
param.grad.data.zero_()
l.backward()
if optimizer is None:
sgd(params, lr, batch_size)
else:
optimizer.step() # “softmax回归的简洁实现”一节将用到
train_l_sum += l.item()
train_acc_sum += (y_hat.argmax(dim=1) == y).sum().item()
n += y.shape[0]
test_acc = evaluate_accuracy(test_iter, net)
print('epoch %d, loss %.4f, train acc %.3f, test acc %.3f'
% (epoch + 1, train_l_sum / n, train_acc_sum / n, test_acc))版本信息
pytorch:
torchvision:
torchtext:
...