在Sentence to Sentence中使用带注意力机制的Encoder和Decoder实现机器翻译

目录

实验所需

实验环境

数据集

数据预处理

模型实现

编码器(Encoder)的实现

实现注意力机制

解码器(Decoder)的实现

训练模型

定义损失函数

训练

模型性能检测

序列预测输出

使用bleu评价翻译结果

使用更大的数据集进行训练

数据来源

代码修正


实验所需

实验环境

python=3.6.8

python库:d2l

                  pytorch=1.5.0

                  torchtext=0.6.0

import collections
import os
import io
import math
import torch
from torch import nn
import torch.nn.functional as F
import torchtext.vocab as Vocab
import torch.utils.data as Data

import sys
# sys.path.append("..") 
import d2lzh_pytorch as d2l

os.environ["CUDA_VISIBLE_DEVICES"] = "0"
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print(torch.__version__, device)

数据集

一个较小的法语--英语数据集

elle est vieille .	she is old .
elle est tranquille .	she is quiet .
elle a tort .	she is wrong .
elle est canadienne .	she is canadian .
elle est japonaise .	she is japanese .
ils sont russes .	they are russian .
ils se disputent .	they are arguing .
ils regardent .	they are watching .
ils sont acteurs .	they are actors .
elles sont crevees .	they are exhausted .
il est mon genre !	he is my type !
il a des ennuis .	he is in trouble .
c est mon frere .	he is my brother .
c est mon oncle .	he is my uncle .
il a environ mon age .	he is about my age .
elles sont toutes deux bonnes .	they are both good .
elle est bonne nageuse .	she is a good swimmer .
c est une personne adorable .	he is a lovable person .
il fait du velo .	he is riding a bicycle .
ils sont de grands amis .	they are great friends .

数据预处理

构造序列中出现词的词典,并用<pad>将序列补至等长,并为每句加上结束标记

使用词典将每句序列转换为对应索引

# 将一个序列中所有的词记录在all_tokens中以便之后构造词典,然后在该序列后面添加PAD直到序列
# 长度变为max_seq_len,然后将序列保存在all_seqs中
PAD, BOS, EOS = '<pad>', '<bos>', '<eos>'
def process_one_seq(seq_tokens, all_tokens, all_seqs, max_seq_len):
    all_tokens.extend(seq_tokens)
    seq_tokens += [EOS] + [PAD] * (max_seq_len - len(seq_tokens) - 1)
    all_seqs.append(seq_tokens)

# 使用所有的词来构造词典。并将所有序列中的词变换为词索引后构造Tensor
def build_data(all_tokens, all_seqs):
    vocab = Vocab.Vocab(collections.Counter(all_tokens),
                        specials=[PAD, BOS, EOS])
    indices = [[vocab.stoi[w] for w in seq] for seq in all_seqs]
    return vocab, torch.tensor(indices)

分别建立法语和英语词典,并利用对应数据创建数据集类为模型生成输入数据

def read_data(max_seq_len):
    # in和out分别是input和output的缩写
    in_tokens, out_tokens, in_seqs, out_seqs = [], [], [], []
    with io.open('fr-en-small.txt') as f:
        lines = f.readlines()
    for line in lines:
        in_seq, out_seq = line.rstrip().split('\t')
        in_seq_tokens, out_seq_tokens = in_seq.split(' '), out_seq.split(' ')
        if max(len(in_seq_tokens), len(out_seq_tokens)) > max_seq_len - 1:
            continue  # 如果加上EOS后长于max_seq_len,则忽略掉此样本
        process_one_seq(in_seq_tokens, in_tokens, in_seqs, max_seq_len)
        process_one_seq(out_seq_tokens, out_tokens, out_seqs, max_seq_len)
    in_vocab, in_data = build_data(in_tokens, in_seqs)
    out_vocab, out_data = build_data(out_tokens, out_seqs)
    return in_vocab, out_vocab, Data.TensorDataset(in_data, out_data)
max_seq_len = 7
in_vocab, out_vocab, dataset = read_data(max_seq_len)

其中一组数据的样子,前是法语输入,后是英语标签

dataset[0]

使用 torch中的DataLoader类以dataset为数据,为训练模型创建一个迭代器

data_iter = DataLoader(dataset, batch_size, shuffle=True)

 至此,数据预处理部分结束

模型实现

编码器(Encoder)的实现

class Encoder(nn.Module):
    def __init__(self, vocab_size, embed_size, num_hiddens, num_layers,
                 drop_prob=0, **kwargs):
        super(Encoder, self).__init__(**kwargs)
        self.embedding = nn.Embedding(vocab_size, embed_size)
        self.rnn = nn.GRU(embed_size, num_hiddens, num_layers, dropout=drop_prob)

    def forward(self, inputs, state):
        # 输入形状是(批量大小, 时间步数)。将输出互换样本维和时间步维
        embedding = self.embedding(inputs.long()).permute(1, 0, 2) # (seq_len, batch, input_size)
        return self.rnn(embedding, state)

    def begin_state(self):
        return None
encoder = Encoder(vocab_size=10, embed_size=8, num_hiddens=16, num_layers=2)
output, state = encoder(torch.zeros((4, 7)), encoder.begin_state())

包含一个词嵌入(向量化输入)和一个GRU门控循环单元

词嵌入将输入转换为向量

GRU对输入上下文实现长期依赖和记忆功能,输出隐藏状态

实现注意力机制

def attention_model(input_size, attention_size):
    model = nn.Sequential(nn.Linear(input_size, attention_size, bias=False),
                          nn.Tanh(),
                          nn.Linear(attention_size, 1, bias=False))
    return model

def attention_forward(model, enc_states, dec_state):
    """
    enc_states: (时间步数, 批量大小, 隐藏单元个数)
    dec_state: (批量大小, 隐藏单元个数)
    """
    # 将解码器隐藏状态广播到和编码器隐藏状态形状相同后进行连结
    dec_states = dec_state.unsqueeze(dim=0).expand_as(enc_states)
    enc_and_dec_states = torch.cat((enc_states, dec_states), dim=2)
    e = model(enc_and_dec_states)  # 形状为(时间步数, 批量大小, 1)
    alpha = F.softmax(e, dim=0)  # 在时间步维度做softmax运算
    return (alpha * enc_states).sum(dim=0)  # 返回背景变量

解码器(Decoder)的实现

class Decoder(nn.Module):
    def __init__(self, vocab_size, embed_size, num_hiddens, num_layers,
                 attention_size, drop_prob=0):
        super(Decoder, self).__init__()
        self.embedding = nn.Embedding(vocab_size, embed_size)
        self.attention = attention_model(2*num_hiddens, attention_size)
        # GRU的输入包含attention输出的c和实际输入, 所以尺寸是 num_hiddens+embed_size
        self.rnn = nn.GRU(num_hiddens + embed_size, num_hiddens, 
                          num_layers, dropout=drop_prob)
        self.out = nn.Linear(num_hiddens, vocab_size)

    def forward(self, cur_input, state, enc_states):
        """
        cur_input shape: (batch, )
        state shape: (num_layers, batch, num_hiddens)
        """
        # 使用注意力机制计算背景向量
        c = attention_forward(self.attention, enc_states, state[-1])
        # 将嵌入后的输入和背景向量在特征维连结, (批量大小, num_hiddens+embed_size)
        input_and_c = torch.cat((self.embedding(cur_input), c), dim=1) 
        # 为输入和背景向量的连结增加时间步维,时间步个数为1
        output, state = self.rnn(input_and_c.unsqueeze(0), state)
        # 移除时间步维,输出形状为(批量大小, 输出词典大小)
        output = self.out(output).squeeze(dim=0)
        return output, state

    def begin_state(self, enc_state):
        # 直接将编码器最终时间步的隐藏状态作为解码器的初始隐藏状态
        return enc_state

使用注意力机制和Encoder的输出作为Decoder的输入,将编码器最终时间步的隐藏状态作为解码器的初始隐藏状态

训练模型

定义损失函数

def batch_loss(encoder, decoder, X, Y, loss):
    batch_size = X.shape[0]
    enc_state = encoder.begin_state()
    enc_outputs, enc_state = encoder(X, enc_state)
    # 初始化解码器的隐藏状态
    dec_state = decoder.begin_state(enc_state)
    # 解码器在最初时间步的输入是BOS
    dec_input = torch.tensor([out_vocab.stoi[BOS]] * batch_size)
    # 我们将使用掩码变量mask来忽略掉标签为填充项PAD的损失, 初始全1
    mask, num_not_pad_tokens = torch.ones(batch_size,), 0
    l = torch.tensor([0.0])
    for y in Y.permute(1,0): # Y shape: (batch, seq_len)
        dec_output, dec_state = decoder(dec_input, dec_state, enc_outputs)
        l = l + (mask * loss(dec_output, y)).sum()
        dec_input = y  # 使用强制教学
        num_not_pad_tokens += mask.sum().item()
        # EOS后面全是PAD. 下面一行保证一旦遇到EOS接下来的循环中mask就一直是0
        mask = mask * (y != out_vocab.stoi[EOS]).float()
    return l / num_not_pad_tokens

使用强制学习,将真实的目标序列中的每个时间步的真实输出作为Decoder的输入,而不是将前一个时间步生成的输出作为当前时间步的输入。会使收敛更快,但在训练和推断时的不一致性会导致泛化能力不强

训练

使用交叉熵损失函数,自适应动量的随机优化方法(Adam)

def train(encoder, decoder, dataset, lr, batch_size, num_epochs):
    enc_optimizer = torch.optim.Adam(encoder.parameters(), lr=lr)
    dec_optimizer = torch.optim.Adam(decoder.parameters(), lr=lr)

    loss = nn.CrossEntropyLoss(reduction='none')
    data_iter = Data.DataLoader(dataset, batch_size, shuffle=True)
    for epoch in range(num_epochs):
        l_sum = 0.0
        for X, Y in data_iter:
            enc_optimizer.zero_grad()
            dec_optimizer.zero_grad()
            l = batch_loss(encoder, decoder, X, Y, loss)
            l.backward()
            enc_optimizer.step()
            dec_optimizer.step()
            l_sum += l.item()
        if (epoch + 1) % 10 == 0:
            print("epoch %d, loss %.3f" % (epoch + 1, l_sum / len(data_iter)))
embed_size, num_hiddens, num_layers = 64, 64, 2
attention_size, drop_prob, lr, batch_size, num_epochs = 10, 0.5, 0.01, 2, 50
encoder = Encoder(len(in_vocab), embed_size, num_hiddens, num_layers,
                  drop_prob)
decoder = Decoder(len(out_vocab), embed_size, num_hiddens, num_layers,
                  attention_size, drop_prob)
train(encoder, decoder, dataset, lr, batch_size, num_epochs)

模型性能检测

序列预测输出

使用贪婪搜索

def translate(encoder, decoder, input_seq, max_seq_len):
    in_tokens = input_seq.split(' ')
    in_tokens += [EOS] + [PAD] * (max_seq_len - len(in_tokens) - 1)
    enc_input = torch.tensor([[in_vocab.stoi[tk] for tk in in_tokens]]) # batch=1
    enc_state = encoder.begin_state()
    enc_output, enc_state = encoder(enc_input, enc_state)
    dec_input = torch.tensor([out_vocab.stoi[BOS]])
    dec_state = decoder.begin_state(enc_state)
    output_tokens = []
    for _ in range(max_seq_len):
        dec_output, dec_state = decoder(dec_input, dec_state, enc_output)
        pred = dec_output.argmax(dim=1)
        pred_token = out_vocab.itos[int(pred.item())]
        if pred_token == EOS:  # 当任一时间步搜索出EOS时,输出序列即完成
            break
        else:
            output_tokens.append(pred_token)
            dec_input = pred
    return output_tokens
input_seq = 'ils regardent .'
translate(encoder, decoder, input_seq, max_seq_len)

使用bleu评价翻译结果

评价机器翻译结果通常使用BLEU(Bilingual Evaluation Understudy)[1]。对于模型预测序列中任意的子序列,BLEU考察这个子序列是否出现在标签序列中。

具体来说,设词数为𝑛的子序列的精度为p_n。它是预测序列与标签序列匹配词数为𝑛的子序列的数量与预测序列中词数为𝑛的子序列的数量之比。举个例子,假设标签序列为𝐴、𝐵、𝐶、𝐷、𝐸、𝐹,预测序列为𝐴、𝐵、𝐵、𝐶、𝐷,那么𝑝1=4/5, 𝑝2=3/4, 𝑝3=1/3, 𝑝4=0 。设len_{\text{label}}len_{\text{pred}}分别为标签序列和预测序列的词数,那么,BLEU的定义为

\exp\left(\min\left(0, 1 - \frac{len_{\text{label}}}{len_{\text{pred}}}\right)\right) \prod_{n=1}^k p_n^{1/2^n}

其中𝑘是我们希望匹配的子序列的最大词数。可以看到当预测序列和标签序列完全一致时,BLEU为1。

因为匹配较长子序列比匹配较短子序列更难,BLEU对匹配较长子序列的精度赋予了更大权重。例如,当p_n固定在0.5时,随着𝑛的增大,0.5^{1/2} \approx 0.7, 0.5^{1/4} \approx 0.84, 0.5^{1/8} \approx 0.92, 0.5^{1/16} \approx 0.96

另外,模型预测较短序列往往会得到较高p_n值。因此,上式中连乘项前面的系数是为了惩罚较短的输出而设的。举个例子,当𝑘=2时,假设标签序列为𝐴、𝐵、𝐶、𝐷、𝐸、𝐹,而预测序列为𝐴、𝐵。虽然𝑝1=𝑝2=1,但惩罚系数exp(1−6/2)≈0.14,因此BLEU也接近0.14。

def bleu(pred_tokens, label_tokens, k):
    len_pred, len_label = len(pred_tokens), len(label_tokens)
    score = math.exp(min(0, 1 - len_label / len_pred))
    for n in range(1, k + 1):
        num_matches, label_subs = 0, collections.defaultdict(int)
        for i in range(len_label - n + 1):
            label_subs[''.join(label_tokens[i: i + n])] += 1
        for i in range(len_pred - n + 1):
            if label_subs[''.join(pred_tokens[i: i + n])] > 0:
                num_matches += 1
                label_subs[''.join(pred_tokens[i: i + n])] -= 1
        score *= math.pow(num_matches / (len_pred - n + 1), math.pow(0.5, n))
    return score
def score(input_seq, label_seq, k):
    pred_tokens = translate(encoder, decoder, input_seq, max_seq_len)
    label_tokens = label_seq.split(' ')
    print('bleu %.3f, predict: %s' % (bleu(pred_tokens, label_tokens, k),
                                      ' '.join(pred_tokens)))
score('ils regardent .', 'they are watching .', k=2)

score('ils sont canadienne .', 'they are canadian .', k=2)

使用更大的数据集进行训练

数据来源

Tatoeba Project. Tab-delimited Bilingual Sentence Pairs from the Tatoeba Project (Good for Anki and Similar Flashcard Applications)

Hi.	嗨。	CC-BY 2.0 (France) Attribution: tatoeba.org #538123 (CM) & #891077 (Martha)
Hi.	你好。	CC-BY 2.0 (France) Attribution: tatoeba.org #538123 (CM) & #4857568 (musclegirlxyp)
Run.	你用跑的。	CC-BY 2.0 (France) Attribution: tatoeba.org #4008918 (JSakuragi) & #3748344 (egg0073)
Stop!	住手!	CC-BY 2.0 (France) Attribution: tatoeba.org #448320 (CM) & #448321 (GlossaMatik)
Wait!	等等!	CC-BY 2.0 (France) Attribution: tatoeba.org #1744314 (belgavox) & #4970122 (wzhd)
Wait!	等一下!	CC-BY 2.0 (France) Attribution: tatoeba.org #1744314 (belgavox) & #5092613 (mirrorvan)
Begin.	开始!	CC-BY 2.0 (France) Attribution: tatoeba.org #6102432 (mailohilohi) & #5094852 (Jin_Dehong)
Hello!	你好。	CC-BY 2.0 (France) Attribution: tatoeba.org #373330 (CK) & #4857568 (musclegirlxyp)
I try.	我试试。	CC-BY 2.0 (France) Attribution: tatoeba.org #20776 (CK) & #8870261 (will66)
I won!	我赢了。	CC-BY 2.0 (France) Attribution: tatoeba.org #2005192 (CK) & #5102367 (mirrorvan)
Oh no!	不会吧。	CC-BY 2.0 (France) Attribution: tatoeba.org #1299275 (CK) & #5092475 (mirrorvan)
Cheers!	乾杯!	CC-BY 2.0 (France) Attribution: tatoeba.org #487006 (human600) & #765577 (Martha)
Got it?	知道了没有?	CC-BY 2.0 (France) Attribution: tatoeba.org #455353 (CM) & #455357 (GlossaMatik)
Got it?	懂了吗?	CC-BY 2.0 (France) Attribution: tatoeba.org #455353 (CM) & #2032276 (ydcok)
Got it?	你懂了吗?	CC-BY 2.0 (France) Attribution: tatoeba.org #455353 (CM) & #7768205 (jiangche)
He ran.	他跑了。	CC-BY 2.0 (France) Attribution: tatoeba.org #672229 (CK) & #5092389 (mirrorvan)
Hop in.	跳进来。	CC-BY 2.0 (France) Attribution: tatoeba.org #1111548 (Scott) & #5092444 (mirrorvan)
I know.	我知道。	CC-BY 2.0 (France) Attribution: tatoeba.org #319990 (CK) & #3378015 (GlossaMatik)
I quit.	我退出。	CC-BY 2.0 (France) Attribution: tatoeba.org #731636 (Eldad) & #5102253 (mirrorvan)
I quit.	我不干了。	CC-BY 2.0 (France) Attribution: tatoeba.org #731636 (Eldad) & #9569182 (MiracleQ)
……

代码修正

加入下列代码,读取你下载的数据集并做简单切分

df = pd.read_csv('./en-zh.txt', sep='\\t', engine='python', header=None)
trainen = df[0].values.tolist()#[:10000]
trainzh = df[1].values.tolist()#[:10000]

 更改下列循环项

def read_data(max_seq_len):
    # in和out分别是input和output的缩写
    in_tokens, out_tokens, in_seqs, out_seqs = [], [], [], []
    for i in range(len(df)):
        in_seq, out_seq = trainen[i], trainzh[i]
        in_seq_tokens, out_seq_tokens = in_seq.split(' '), out_seq.split(' ')
        if max(len(in_seq_tokens), len(out_seq_tokens)) > max_seq_len - 1:
            continue  # 如果加上EOS后长于max_seq_len,则忽略掉此样本
        process_one_seq(in_seq_tokens, in_tokens, in_seqs, max_seq_len)
        process_one_seq(out_seq_tokens, out_tokens, out_seqs, max_seq_len)
    in_vocab, in_data = build_data(in_tokens, in_seqs)
    out_vocab, out_data = build_data(out_tokens, out_seqs)
    return in_vocab, out_vocab, Data.TensorDataset(in_data, out_data)
for i in range(len(df)):
        in_seq, out_seq = trainen[i], trainzh[i]

参数设定

embed_size, num_hiddens, num_layers = 128, 128, 2
attention_size, drop_prob, lr, batch_size, num_epochs = 10, 0.5, 0.001, 64, 16
encoder = Encoder(len(in_vocab), embed_size, num_hiddens, num_layers,
                  drop_prob)
decoder = Decoder(len(out_vocab), embed_size, num_hiddens, num_layers,
                  attention_size, drop_prob)
encoder = encoder.to(device)
decoder = decoder.to(device)
train(encoder, decoder, dataset, lr, batch_size, num_epochs)

训练结果

翻译

input_seq = 'I thought Tom and Mary were crazy.'
input_seq = input_seq
translate(encoder, decoder, input_seq, max_seq_len)

评估

score('I thought Tom and Mary were crazy.', '我本以为汤姆和玛丽疯了呢。', k=2)

 

score('I visited my friend Tom yesterday.', '我昨天拜訪了我的朋友湯姆。', k=2)

  • 32
    点赞
  • 19
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
好的,以下是一个简单的encoder-decoder模型的代码,仅使用PyTorch和PyCharm: ``` import torch import torch.nn as nn import torch.optim as optim # 定义 Encoder 模型 class Encoder(nn.Module): def __init__(self, input_size, hidden_size): super(Encoder, self).__init__() self.hidden_size = hidden_size self.embedding = nn.Embedding(input_size, hidden_size) self.gru = nn.GRU(hidden_size, hidden_size) def forward(self, input, hidden): embedded = self.embedding(input).view(1, 1, -1) output = embedded output, hidden = self.gru(output, hidden) return output, hidden def init_hidden(self): return torch.zeros(1, 1, self.hidden_size) # 定义 Decoder 模型 class Decoder(nn.Module): def __init__(self, hidden_size, output_size): super(Decoder, self).__init__() self.hidden_size = hidden_size self.embedding = nn.Embedding(output_size, hidden_size) self.gru = nn.GRU(hidden_size, hidden_size) self.out = nn.Linear(hidden_size, output_size) self.softmax = nn.LogSoftmax(dim=1) def forward(self, input, hidden): output = self.embedding(input).view(1, 1, -1) output = nn.functional.relu(output) output, hidden = self.gru(output, hidden) output = self.softmax(self.out(output[0])) return output, hidden def init_hidden(self): return torch.zeros(1, 1, self.hidden_size) # 定义训练函数 def train(input_tensor, target_tensor, encoder, decoder, encoder_optimizer, decoder_optimizer, criterion, max_length=20): encoder_hidden = encoder.init_hidden() encoder_optimizer.zero_grad() decoder_optimizer.zero_grad() input_length = input_tensor.size(0) target_length = target_tensor.size(0) encoder_outputs = torch.zeros(max_length, encoder.hidden_size, device=device) loss = 0 for ei in range(input_length): encoder_output, encoder_hidden = encoder( input_tensor[ei], encoder_hidden) encoder_outputs[ei] = encoder_output[0, 0] decoder_input = torch.tensor([[SOS_token]], device=device) decoder_hidden = encoder_hidden for di in range(target_length): decoder_output, decoder_hidden = decoder( decoder_input, decoder_hidden) loss += criterion(decoder_output, target_tensor[di]) decoder_input = target_tensor[di] loss.backward() encoder_optimizer.step() decoder_optimizer.step() return loss.item() / target_length # 定义翻译函数 def evaluate(encoder, decoder, sentence, max_length=20): with torch.no_grad(): input_tensor = tensorFromSentence(input_lang, sentence) input_length = input_tensor.size()[0] encoder_hidden = encoder.init_hidden() encoder_outputs = torch.zeros(max_length, encoder.hidden_size, device=device) for ei in range(input_length): encoder_output, encoder_hidden = encoder(input_tensor[ei], encoder_hidden) encoder_outputs[ei] += encoder_output[0, 0] decoder_input = torch.tensor([[SOS_token]], device=device) decoder_hidden = encoder_hidden decoded_words = [] for di in range(max_length): decoder_output, decoder_hidden = decoder( decoder_input, decoder_hidden) topv, topi = decoder_output.data.topk(1) if topi.item() == EOS_token: decoded_words.append('<EOS>') break else: decoded_words.append(output_lang.index2word[topi.item()]) decoder_input = topi.squeeze().detach() return decoded_words # 主程序 device = torch.device("cuda" if torch.cuda.is_available() else "cpu") input_lang, output_lang, pairs = prepareData('eng', 'fra', True) print(random.choice(pairs)) hidden_size = 256 encoder1 = Encoder(input_lang.n_words, hidden_size).to(device) decoder1 = Decoder(hidden_size, output_lang.n_words).to(device) learning_rate = 0.01 encoder_optimizer = optim.SGD(encoder1.parameters(), lr=learning_rate) decoder_optimizer = optim.SGD(decoder1.parameters(), lr=learning_rate) criterion = nn.NLLLoss() n_iters = 10000 print_every = 1000 plot_every = 1000 training_pairs = [tensorsFromPair(random.choice(pairs)) for i in range(n_iters)] losses = [] for iter in range(1, n_iters + 1): training_pair = training_pairs[iter - 1] input_tensor = training_pair[0] target_tensor = training_pair[1] loss = train(input_tensor, target_tensor, encoder1, decoder1, encoder_optimizer, decoder_optimizer, criterion) losses.append(loss) if iter % print_every == 0: print('(%d %d%%) %.4f' % (iter, iter / n_iters * 100, loss)) ``` 这个代码实现了一个简单的英法翻译的encoder-decoder模型,使用了PyTorch和PyCharm。不过需要注意的是,这个代码使用了一些其他的函数和数据,需要自行调整和补充。

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值