本文主要是介绍【神经网络与深度学习】Long short-term memory网络（LSTM)，希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！
简单介绍
API介绍：
nn.LSTM(input_size=100, hidden_size=10, num_layers=1,batch_first=True, bidirectional=True)
inuput_size: embedding_dim
hidden_size: 每一层LSTM单元的数量
num_layers: RNN中LSTM的层数
batch_first: True对应[batch_size, seq_len, embedding_dim]
bidiectional: True对应使用双向LSTM
实例化LSTM对象后，不仅要传入数据，还有传入前一次的h_0和c_0
lstm(input, (h_0, c_0))
LSTM默认输出(output, (h_n, c_n))
output: [ seq_len, batch, hidden_size*num_directions ] (若batch_first=false)
h_n: [num_directions, batch, hidden_size]
c_n : [num_directions, batch, hidden_size]
import torch.nn as nn
import torch.nn.functional as F
import torchbatch_size = 10
seq_len =20 #句子长度
vocab_size = 100 # 词典数量
embedding_dim = 30 # 用embedding_dim长度的向量表示一个词语
hidden_size = 18input = torch.randint(0, 100, [batch_size, seq_len])
print(input.size())
print("*"*100)
# 经过embedding
embed = nn.Embedding(vocab_size, embedding_dim)input_embed = embed(input) # [bs, seq_len, embedding_dim]
print(input_embed.size())
print("*"*100)
lstm = nn.LSTM(embedding_dim, hidden_size=hidden_size, num_layers=1, batch_first=True)
output,(h_n, c_n) = lstm(input_embed)
print(output.size())
print("*"*100)
print(h_n.size())
print("*"*100)
print(c_n.size())通常由最后一个输出代替整个句子
使用双向LSTM实现
"""
定义模型
"""
import torch
import torch.nn as nn
import torch.nn.functional as F
from lib import ws,max_len
from dataset import get_data
import lib
import os
import numpy as np
class MyModel(nn.Module):def __init__(self):super(MyModel, self).__init__()self.embedding = nn.Embedding(len(ws), 100)self.lstm = nn.LSTM(input_size=100, hidden_size=lib.hidden_size, num_layers=lib.num_layers,batch_first=True, bidirectional=lib.bidirectional, dropout=lib.dropout)self.fc = nn.Linear(lib.hidden_size*2, 2)def forward(self, input):""":param input: [batch_size, max_len]:return:"""x = self.embedding(input) # [batch_size, max_len, 100]x,(h_n,c_n)= self.lstm(x)output = torch.cat([h_n[-2,:,:],h_n[-1,:,:]],dim=-1)output = self.fc(output)return F.log_softmax(output,dim=-1)model = MyModel().to(lib.device)
optimizer = torch.optim.Adam(model.parameters(),lr=0.001)
if os.path.exists("./model0/model.pkl"):model.load_state_dict(torch.load("./model0/model.pkl"))optimizer.load_state_dict(torch.load("./model0/optimizer.pkl"))def train(epoch):for idx,(input,target) in enumerate(get_data(train=True)):input = input.to(lib.device)target = target.to(lib.device)# 梯度清零optimizer.zero_grad()output= model(input)loss = F.nll_loss(output,target)loss.backward()optimizer.step()print(epoch, idx, loss.item())if idx%100==0:torch.save(model.state_dict(),"./model0/model.pkl")torch.save(optimizer.state_dict(),"./model0/optimizer.pkl")def eval():loss_list = []acc_list = []for idx,(input,target) in enumerate(get_data(train=False, batch_size=lib.test_batch_size)):input = input.to(lib.device)target = target.to(lib.device)with torch.no_grad():output= model(input)loss = F.nll_loss(output,target)loss_list.append(loss.cpu().item())pre = output.max(dim=-1)[-1]acc = pre.eq(target).float().mean()acc_list.append(acc.cpu().item())print("total loss, acc:", np.mean(loss_list), np.mean(acc_list))if __name__ == '__main__':for i in range(10):train(epoch=i)eval()这篇关于【神经网络与深度学习】Long short-term memory网络（LSTM)的文章就介绍到这儿，希望我们推荐的文章对编程师们有所帮助！
