产品摄影网站,网站二级栏目如何调用,微网站特点,长沙装修公司招聘信息目录
1.创建 添加了Bahdanau的decoder
2. 训练 3.定义评估函数BLEU 4.预测 5.知识点个人理解 1.创建 添加了Bahdanau的decoder
import torch
from torch import nn
import dltools#定义注意力解码器基类
class AttentionDecoder(dltools.Decoder): #继承dltools.Decoder写…目录
1.创建 添加了Bahdanau的decoder
2. 训练 3.定义评估函数BLEU 4.预测 5.知识点个人理解 1.创建 添加了Bahdanau的decoder
import torch
from torch import nn
import dltools#定义注意力解码器基类
class AttentionDecoder(dltools.Decoder): #继承dltools.Decoder写注意力编码器的基类def __init__(self, **kwargs):super().__init__(**kwargs)property #装饰器 定义的函数方法可以像类的属性一样被调用def attention_weights(self):#raise用于引发(或抛出)异常raise NotImplementedError #通常用于抽象基类中作为占位符提醒子类必须实现这个方法。 #创建 添加了Bahdanau的decoder
#继承AttentionDecoder这个基类创建Seq2SeqAttentionDecoder子类 子类必须实现父类中NotImplementedError占位的方法
class Seq2SeqAttentionDecoder(AttentionDecoder): #初始化属性和方法def __init__(self, vocab_size, embed_size, num_hiddens, num_layers, dropout0, **kwargs):vocab_size:此表大小, 相当于输入数据的特征数features, 也是输出数据的特征数embed_size嵌入层的大小将输入数据处理成小批量的数据num_hiddens隐藏层神经元的数量num_layers循环网络的层数dropout0不释放模型的参数比如神经元super().__init__(**kwargs)#初始化注意力机制的评分函数方法self.attention dltools.AdditiveAttention(key_sizenum_hiddens,query_sizenum_hiddens, num_hiddensnum_hiddens,dropoutdropout)#初始化嵌入层将输入的数据处理成小批量的tensor数据 文本---数值的映射转化self.embedding nn.Embedding(num_embeddingsvocab_size, embedding_dimembed_size)#初始化循环网络self.rnn nn.GRU(embed_sizenum_hiddens, num_hiddens, num_layers, dropoutdropout)#初始化线性层 输出层self.dense nn.Linear(num_hiddens, vocab_size)#初始化隐藏层的状态state (计算state需要编码器的输出结果、序列的有效长度)def init_state(self, enc_outputs, enc_valid_lens, *args):#enc_outputs是一个元组输出结果隐藏状态#outputs的shapebatch_size, num_steps, num_hiddens#hidden_state的shapenum_layers, batch_size, num_hiddensoutputs, hidden_state enc_outputs#返回一个元组可以用一个变量接收#outputs.permute(1, 0, 2)转换数据的维度是因为rnn循环神经网络的输入要求是先num_steps再batch_size,return (outputs.permute(1, 0, 2), hidden_state, enc_valid_lens)#定义前向传播 (输入数据Xstate)def forward(self, X, state):#变量赋值接收编码器encoder的输出结果、隐藏状态、序列有效长度#enc_outputs的shapebatch_size, num_steps, num_hiddens#hidden_state的shapenum_layers, batch_size, num_hiddensenc_outputs, hidden_state, enc_valid_lens state#X的shapebatch_size, num_steps, vocab_sizeX self.embedding(X) #将X输入embedding嵌入层后 X的shapebatch_size, num_steps, embed_size#调换X的0维度和1维度数据X X.permute(1, 0, 2) #X的shapenum_steps, batch_size, embed_sizeoutputs, self._attention_weights [], [] #创建空列表用于存储数据for x in X: #遍历每一批数据#获取query#hidden_state[-1]表示最后一层循环网络的隐藏层状态 有两层循环网络#hidden_state[-1]的shapebatch_size, num_hiddens #dim1表示在原索引1的维度增加一个维度query torch.unsqueeze(hidden_state[-1], dim1)
# print(query的shape:, query.shape) #query的shape(batch_size, 1, num_hiddens)#通过注意力机制获取上下文序列context self.attention(query, enc_outputs, enc_outputs, enc_valid_lens)
# print(context的shape:, context.shape) #context的shape(batch_size, 1, num_hiddens)#用最后一个维度 拼接context, x 数据x torch.cat((context, torch.unsqueeze(x, dim1)), dim-1)
# print(x的shape:, x.shape) #x的shape(batch_size, 1, num_hiddensembed_size)#将x和hidden_state输入循环神经网络中获取输出结果和新的hidden_stateout, hidden_state self.rnn(x.permute(1, 0, 2), hidden_state)
# print(out的shape:, out.shape) #out的shape(1, batch_size, num_hiddens)
# print(hidden_state的shape:, hidden_state.shape) #两层循环层hidden_state的shape(2, batch_size, num_hiddens)#将输出结果添加到列表中outputs.append(out)self._attention_weights.append(self.attention_weights)outputs self.dense(torch.cat(outputs, dim0))
# print(outputs的shape:, outputs.shape) #outputs的shape(num_steps, batch_size, vocab_size)return outputs.permute(1, 0, 2), [enc_outputs, hidden_state, enc_valid_lens]propertydef attention_weights(self):return self._attention_weights#测试代码
#创建编码器对象
encoder dltools.Seq2SeqEncoder(vocab_size10, embed_size8, num_hiddens16, num_layers2)
#需要预测 要加encoder.eval()
encoder.eval()
#创建解码器对象
decoder Seq2SeqAttentionDecoder(vocab_size10, embed_size8, num_hiddens16, num_layers2)
decoder.eval()#假设数据
batch_size, num_steps 4, 7
X torch.zeros((4, 7), dtype torch.long)
#初始化状态state
state decoder.init_state(encoder(X), None)
outputs, state decoder(X, state)
#state包含三个东西enc_outputs, hidden_state, enc_valid_lens
#state[0]是 enc_outputs
#state[1]是 hidden_state 两层循环层就会有两个hidden_state state[1][0]是第一层的hidden_state
outputs.shape, len(state), state[0].shape, len(state[1]), state[1][0].shape query的shape: torch.Size([4, 1, 16])
context的shape: torch.Size([4, 1, 16])
x的shape: torch.Size([4, 1, 24])
out的shape: torch.Size([1, 4, 16])
hidden_state的shape: torch.Size([2, 4, 16])
query的shape: torch.Size([4, 1, 16])
context的shape: torch.Size([4, 1, 16])
x的shape: torch.Size([4, 1, 24])
out的shape: torch.Size([1, 4, 16])
hidden_state的shape: torch.Size([2, 4, 16])
query的shape: torch.Size([4, 1, 16])
context的shape: torch.Size([4, 1, 16])
x的shape: torch.Size([4, 1, 24])
out的shape: torch.Size([1, 4, 16])
hidden_state的shape: torch.Size([2, 4, 16])
query的shape: torch.Size([4, 1, 16])
context的shape: torch.Size([4, 1, 16])
x的shape: torch.Size([4, 1, 24])
out的shape: torch.Size([1, 4, 16])
hidden_state的shape: torch.Size([2, 4, 16])
query的shape: torch.Size([4, 1, 16])
context的shape: torch.Size([4, 1, 16])
x的shape: torch.Size([4, 1, 24])
out的shape: torch.Size([1, 4, 16])
hidden_state的shape: torch.Size([2, 4, 16])
query的shape: torch.Size([4, 1, 16])
context的shape: torch.Size([4, 1, 16])
x的shape: torch.Size([4, 1, 24])
out的shape: torch.Size([1, 4, 16])
hidden_state的shape: torch.Size([2, 4, 16])
query的shape: torch.Size([4, 1, 16])
context的shape: torch.Size([4, 1, 16])
x的shape: torch.Size([4, 1, 24])
out的shape: torch.Size([1, 4, 16])
hidden_state的shape: torch.Size([2, 4, 16])
outputs的shape: torch.Size([7, 4, 10])Out[11]: (torch.Size([4, 7, 10]), 3, torch.Size([4, 7, 16]), 2, torch.Size([4, 16]))2. 训练
#声明变量
embed_size, num_hiddens, num_layers, dropout 32, 32, 2, 0.1
batch_size, num_steps 64, 10
lr, num_epochs, device 0.005, 200, dltools.try_gpu()#加载数据
train_iter, src_vocab, tgt_vocab dltools.load_data_nmt(batch_size, num_steps)#创建编辑器对象
encoder dltools.Seq2SeqEncoder(len(src_vocab), embed_size, num_hiddens, num_layers, dropout)
#创建编辑器对象
decoder Seq2SeqAttentionDecoder(len(tgt_vocab), embed_size, num_hiddens, num_layers, dropout)#创建网络模型
net dltools.EncoderDecoder(encoder, decoder)#模型训练
dltools.train_seq2seq(net, train_iter, lr, num_epochs, tgt_vocab, device) 3.定义评估函数BLEU def bleu(pred_seq, label_seq, k):print(pred_seq:, pred_seq)print(label_seq:, label_seq)#将pred_seq, label_seq分别进行空格分隔pred_tokens, label_tokens pred_seq.split( ), label_seq.split( )#获取pred_seq, label_seq的长度len_pred, len_label len(pred_seq), len(label_seq)score math.exp(min(0, 1 - (len_label / len_pred)))for n in range(1, k1): #n的取值范围 range左闭右开num_matches, label_subs 0, collections.defaultdict(int)for i in range(len_label - n 1):label_subs[ .join(label_tokens[i: in])] 1for i in range(len_pred - n 1):if label_subs[ .join(pred_tokens[i: in])] 0:num_matches 1label_subs[ .join(pred_tokens[i: in])] -1score * math.pow(num_matches / (len_pred -n 1), math.pow(0.5, n))return score 4.预测
import math
import collectionsengs [go ., i lost ., he\s calm ., i\m home .]
fras [va !, j\ai perdu ., il est calme ., je suis chez moi .]
for eng, fra in zip(engs, fras):translation dltools.predict_seq2seq(net, eng, src_vocab, tgt_vocab, num_steps, device)print(f{eng} {translation}, bleu {dltools.bleu(translation[0], fra, k2):.3f}) go . (va !, []), bleu 1.000
i lost . (jai perdu ., []), bleu 1.000
hes calm . (il est bon ., []), bleu 0.658
im home . (je suis chez moi ., []), bleu 1.000 5.知识点个人理解
