def predict(net, test_iter):
"""模型预测"""
X, y = iter(test_iter).next()
print("X shape:", X.shape)
print("y shape:", y.shape)
if isinstance(net, torch.nn.Module):
net.eval() # 评估模式，这会关闭dropout
y_hat = net(X).argmax(dim=1)
net.train()
else: # 自定义的模型, func.__code__.co_varnames:将函数局部变量以元组的形式返回
if ('is_training' in net.__code__.co_varnames): # 如果有is_training这个参数
# 将is_training设置成False
y_hat = net(X, is_training=False).argmax(dim=1)
else:
y_hat = net(X).argmax(dim=1)
true_labels = get_fashion_mnist_labels(y.numpy())
predict_labels = get_fashion_mnist_labels(y_hat.numpy())
titles = [
true + '\n' + pred for true, pred in zip(true_labels, predict_labels)
]
show_fashion_mnist(X[10:19],
titles[10:19],
imagepath='../figures/3.13_dropout.jpg')
def predict(net, test_iter):
X, y = iter(test_iter).next()
y_hat = net(X).argmax(dim=1)
true_labels = get_fashion_mnist_labels(y.numpy())
predict_labels = get_fashion_mnist_labels(y_hat.numpy())
titles = [
true + '\n' + pred for true, pred in zip(true_labels, predict_labels)
]
show_fashion_mnist(
X[10:19],
titles[10:19],
imagepath='../figures/3.6_softmax-regression-scratch.jpg')
def predict(net, test_iter):
X, y = iter(test_iter).next()
print("X shape:", X.shape)
print("y shape:", y.shape)
y_hat = net(X).argmax(dim=1)
true_labels = get_fashion_mnist_labels(y.numpy())
predict_labels = get_fashion_mnist_labels(y_hat.numpy())
titles = [
true + '\n' + pred for true, pred in zip(true_labels, predict_labels)
]
show_fashion_mnist(X[10:19],
titles[10:19],
imagepath='../figures/3.9_mlp-scratch.jpg')
def main():
# Step 1：获取数据集
mnist_train = torchvision.datasets.FashionMNIST(
root='./Datasets/',
train=True,
download=True,
transform=transforms.ToTensor())
mnist_test = torchvision.datasets.FashionMNIST(
root='./Datasets/',
train=False,
download=True,
transform=transforms.ToTensor())
### begin Unit Test ###
# output FashionMNIST
print(type(mnist_train))
print(len(mnist_train), len(mnist_test))
feature, label = mnist_train[0]
print("feature.shape:", feature.shape, '\n' + "feature.dtype:",
feature.dtype) # Channel * Height * Width -> torch.Size([1, 28, 28])
print("label:", label) # 9 -> torch.float32
### end Unit Test ###
X, y = [], []
for i in range(10):
X.append(mnist_train[i][0]) # features
y.append(mnist_train[i][1]) # labels
show_fashion_mnist(X, get_fashion_mnist_labels(y), '../figures/test.jpg')
train_iter = data_iter(batch_size=256, data=mnist_train, shuffle=True)
test_iter = data_iter(batch_size=256, data=mnist_test, shuffle=False)
import time
start = time.time()
for X, y in train_iter:
continue
print("%.2f sec" % (time.time() - start))