import d2lzh as d2l
from mxnet import gluon,init
from mxnet.gluon import loss as gloss,nn
batch_size = 256
train_iter,test_iter = d2l.load_data_fashion_mnist(batch_size)
net = nn.Sequential()
net.add(nn.Dense(10))
net.initialize(init.Normal(sigma=0.01))
loss = gloss.SoftmaxCrossEntropyLoss()
trainer = gluon.Trainer(net.collect_params(),'sgd',{'learning_rate':0.1})
num_epochs = 5
d2l.train_ch3(net,train_iter,test_iter,loss,num_epochs,batch_size,None,None,trainer)
epoch 1, loss 0.7890, train acc 0.747, test acc 0.806