代码练习day03-深度循环网络和编码器解码器架构

李沐-《动手学深度学习》

1.深度循环网络

import torch
from torch import nn
from d2l import torch as d2l

banch_size,num_steps=32,35
train_iter,vocab=d2l.load_data_time_machine(banch_size,num_steps)

#num_layers设定隐藏层数
vocab_size,num_hiddens,num_layers=len(vocab),256,2
num_inputs=vocab_size
device=d2l.try_gpu()
lstm_layer=nn.LSTM(num_inputs,num_hiddens,num_layers)

"""input_size - 输入中预期的特征数量`x`。
hidden_size - 隐藏状态下的特征数量`h`。
num_layers - 循环层的数量。例如，设置 "num_layers=2 "意味着将两个LSTM堆叠在一起，形成一个 "堆叠LSTM"，第二个LSTM接收第一个LSTM的输出并计算最终结果。默认值：1
bias - 如果`False`，那么该层不使用偏置权重`b_ih`和`b_hh`。"""
model=d2l.RNNModel(lstm_layer,len(vocab))
model=model.to(device)

#训练
num_epochs,lr=500,2
d2l.train_ch8(model,train_iter,vocab,lr,num_epochs,device)

2.编码器解码器架构\

from torch import nn

#@save
class Encoder(nn.Module):
    """编码器-解码器的基本编码器接口"""
    def __init__(self,**kwangs):
        super(Encoder, self).__init__(**kwangs)

    def forward(self, X, *args):
        raise NotImplementedError

#@save
class Decoder(nn.Module):
    """编码器-解码器的基本解码器接口"""
    def __init__(self,**kwargs):
        super(Decoder, self).__init__(**kwargs)

    #enc_output输出，
    def init_state(self,enc_output,*args):
        raise NotImplementedError

    def forward(self, X, state):
        raise NotImplementedError

#@save
class EncoderDecoder(nn.Module):
    """编码器-解码器架构的基类"""
    def __init__(self,encoder,decoder,**kwangs):
        super(EncoderDecoder, self).__init__(**kwangs)
        self.encoder=encoder
        self.decoder=decoder

    def forward(self, enc_X,dec_X,*args):
        #编码器的输出
        enc_outputs=self.encoder(enc_X,*args)
        #解码器的状态
        dec_state=self.decoder.init_state(enc_outputs,*args)
        return self.decoder(dec_X,dec_state)