注：书中对代码的讲解并不详细，本文对很多细节做了详细注释。另外，书上的源代码是在Jupyter Notebook上运行的，较为分散，本文将代码集中起来，并加以完善，全部用vscode在python 3.9.18下测试通过，同时对于书上部分章节也做了整合。

Chapter8 Recurrent Neural Networks

8.6 Concise Implementation of RNN

import torch
from torch import nn
from torch.nn import functional as F
from d2l import torch as d2l
import matplotlib.pyplot as pltbatch_size, num_steps = 32, 35
train_iter, vocab = d2l.load_data_time_machine(batch_size, num_steps)num_hiddens = 256
rnn_layer = nn.RNN(len(vocab), num_hiddens)state = torch.zeros((1, batch_size, num_hiddens))
print(state.shape)X = torch.rand(size=(num_steps, batch_size, len(vocab)))
Y, state_new = rnn_layer(X, state)#Y不涉及输出层的计算
print(Y.shape, state_new.shape)class RNNModel(nn.Module):#@save"""循环神经网络模型"""def __init__(self, rnn_layer, vocab_size, **kwargs):super(RNNModel, self).__init__(**kwargs)self.rnn = rnn_layerself.vocab_size = vocab_sizeself.num_hiddens = self.rnn.hidden_size# 如果RNN是双向的（之后将介绍），num_directions应该是2，否则应该是1if not self.rnn.bidirectional:self.num_directions = 1self.linear = nn.Linear(self.num_hiddens, self.vocab_size)else:self.num_directions = 2self.linear = nn.Linear(self.num_hiddens * 2, self.vocab_size)def forward(self, inputs, state):X = F.one_hot(inputs.T.long(), self.vocab_size)X = X.to(torch.float32)Y, state = self.rnn(X, state)# 全连接层首先将Y的形状改为(时间步数*批量大小,隐藏单元数)，它的输出形状是(时间步数*批量大小,词表大小)。output = self.linear(Y.reshape((-1, Y.shape[-1])))return output, statedef begin_state(self, device, batch_size=1):if not isinstance(self.rnn, nn.LSTM):#nn.GRU以张量作为隐状态#GRU为门控循环单元(Gated Recurrent Unit)，是一种流行的循环神经网络变体。#GRU使用了一组门控机制来控制信息的流动，包括更新门(update gate)和重置门(reset gate)，以更好地捕捉长期依赖关系return  torch.zeros((self.num_directions * self.rnn.num_layers,batch_size, self.num_hiddens),device=device)else:#nn.LSTM以元组作为隐状态#LSTM代表长短期记忆网络(Long Short-Term Memory)，是另一种常用的循环神经网络类型。#相比于简单的循环神经网络，LSTM引入了三个门控单元：输入门(input gate)、遗忘门(forget gate)和输出门(output gate)，以及一个记忆单元(cell state)，可以更有效地处理长期依赖性。return (torch.zeros((self.num_directions * self.rnn.num_layers,batch_size, self.num_hiddens), device=device),torch.zeros((self.num_directions * self.rnn.num_layers,batch_size, self.num_hiddens), device=device))device = d2l.try_gpu()
net = RNNModel(rnn_layer, vocab_size=len(vocab))
net = net.to(device)
d2l.predict_ch8('time traveller', 10, net, vocab, device)num_epochs, lr = 500, 1
d2l.train_ch8(net, train_iter, vocab, lr, num_epochs, device)
plt.show()

训练结果:

与上一节相比，由于pytorch的高级API对代码进行了更多的优化，该模型在较短的时间内达到了较低的困惑度。