[pytorch、学习]3.9多重感知机的从零开始实现

import torch
import numpy as np
import sys
sys.path.append("..")
import d2lzh_pytorch as d2l


3.9.1. 获取和读取数据

batch_size &＃61; 256
train_iter, test_iter &＃61; d2l.load_data_fashion_mnist(batch_size)


3.9.2. 定义模型参数

num_inputs, num_outputs, num_hiddens &＃61; 784, 10, 256W1 &＃61; torch.tensor(np.random.normal(0, 0.01, (num_inputs, num_hiddens)), dtype&＃61;torch.float)
b1 &＃61; torch.zeros(num_hiddens, dtype&＃61;torch.float)
W2 &＃61; torch.tensor(np.random.normal(0, 0.01, (num_hiddens, num_outputs)), dtype&＃61;torch.float)
b2 &＃61; torch.zeros(num_outputs, dtype&＃61;torch.float)params &＃61; [W1, b1, W2, b2]
for param in params:param.requires_grad_(requires_grad&＃61;True)


3.9.3. 定义激活函数

def relu(X):return torch.max(input&＃61;X, other&＃61;torch.tensor(0.0))


3.9.4. 定义模型

def net(X):X &＃61; X.view((-1, num_inputs))H &＃61; relu(torch.matmul(X, W1) &＃43; b1)return torch.matmul(H, W2) &＃43; b2


3.9.5. 定义损失函数

loss &＃61; torch.nn.CrossEntropyLoss()


3.9.6. 训练模型

num_epochs, lr &＃61; 5, 100.0
d2l.train_ch3(net, train_iter, test_iter, loss, num_epochs, batch_size, params, lr)


3.9.7. 预测

X, y &＃61; iter(test_iter).next()true_labels &＃61; d2l.get_fashion_mnist_labels(y.numpy())
pred_labels &＃61; d2l.get_fashion_mnist_labels(net(X).argmax(dim&＃61;1).numpy())
titles &＃61; [true &＃43; &＃39;\n&＃39; &＃43; pred for true, pred in zip(true_labels, pred_labels)]d2l.show_fashion_mnist(X[0:9], titles[0:9])