from mxnet import gluon,initfrom mxnet.gluon import loss as gloss, nnfrom mxnet.gluon import data as gdatafrom mxnet import nd,autogradimport gluonbook as gbimport sys# 读取数据# 读取数据mnist_train = gdata.vision.FashionMNIST(train=True)mnist_test = gdata.vision.FashionMNIST(train=False)batch_size = 256transformer = gdata.vision.transforms.ToTensor()if sys.platform.startswith('win'):    num_workers = 0else:    num_workers = 4# 小批量数据迭代器train_iter = gdata.DataLoader(mnist_train.transform_first(transformer),batch_size=batch_size,shuffle=True,num_workers=num_workers)test_iter = gdata.DataLoader(mnist_test.transform_first(transformer),batch_size=batch_size,shuffle=False,num_workers=num_workers)# 定义网络net = nn.Sequential()net.add(nn.Dense(256,activation='relu'),nn.Dense(10))net.initialize(init.Normal(sigma=0.01))# 损失函数loss = gloss.SoftmaxCrossEntropyLoss()trainer = gluon.Trainer(net.collect_params(),'sgd',{
   'learning_rate':0.5})def accuracy(y_hat, y):    return (y_hat.argmax(axis=1) == y.astype('float32')).mean().asscalar()def evaluate_accuracy(data_iter, net):    acc = 0    for X, y in data_iter:        acc += accuracy(net(X), y)    return acc / len(data_iter)num_epochs = 5def train(net,train_iter,test_iter,loss,num_epochs,batch_size,params=None,lr=None,trainer=None):    for epoch in range(num_epochs):        train_l_sum = 0        train_acc_sum = 0        for X,y in train_iter:            with autograd.record():                y_hat = net(X)                l = loss(y_hat,y)            l.backward()            if trainer is None:                gb.sgd(params,lr,batch_size)            else:                trainer.step(batch_size)            train_l_sum += l.mean().asscalar()        test_acc = evaluate_accuracy(test_iter,net)        print('epoch %d,loss %.4f,test acc %.3f'%(epoch+1,train_l_sum / len(train_iter),test_acc))train(net,train_iter,test_iter,loss,num_epochs,batch_size,None,None,trainer)