使用 TorchText 进行语言翻译

使用 TorchText 进行语言翻译

本教程说明如何使用torchtext的几个便捷类来预处理包含英语和德语句子的著名数据集的数据，并使用它来训练序列到序列模型，并注意将德语句子翻译成英语 。

它基于 PyTorch 社区成员 Ben Trevett 的本教程，并由 Seth Weidman 在 Ben 的允许下创建。

在本教程结束时，您将能够：

• Preprocess sentences into a commonly-used format for NLP modeling using the following torchtext convenience classes:
  

  • TranslationDataset
  • 字段
  • BucketIterator

<cite>字段</cite>和 <cite>TranslationDataset</cite>

torchtext具有用于创建数据集的实用程序，可以轻松地对其进行迭代，以创建语言翻译模型。 一个关键类是字段，它指定应该对每个句子进行预处理的方式，另一个关键类是 <cite>TranslationDataset</cite> ； torchtext有几个这样的数据集； 在本教程中，我们将使用 Multi30k 数据集，其中包含约 30,000 个英语和德语句子(平均长度约为 13 个单词）。

注意：本教程中的标记化需要 Spacy 我们使用 Spacy，因为它为英语以外的其他语言的标记化提供了强大的支持。 torchtext提供了basic_english标记器，并支持其他英语标记器(例如摩西），但对于语言翻译(需要多种语言），Spacy 是您的最佳选择。

要运行本教程，请先使用pip或conda安装spacy。 接下来，下载英语和德语 Spacy 分词器的原始数据：

python -m spacy download en
python -m spacy download de

安装 Spacy 后，以下代码将根据Field中定义的标记器，标记TranslationDataset中的每个句子。

from torchtext.datasets import Multi30k
from torchtext.data import Field, BucketIteratorSRC = Field(tokenize = "spacy",tokenizer_language="de",init_token = '<sos>',eos_token = '<eos>',lower = True)TRG = Field(tokenize = "spacy",tokenizer_language="en",init_token = '<sos>',eos_token = '<eos>',lower = True)train_data, valid_data, test_data = Multi30k.splits(exts = ('.de', '.en'),fields = (SRC, TRG))

出：

downloading training.tar.gz
downloading validation.tar.gz
downloading mmt_task1_test2016.tar.gz

现在我们已经定义了train_data，我们可以看到torchtext的Field的一个非常有用的功能：build_vocab方法现在允许我们创建与每种语言相关的词汇

SRC.build_vocab(train_data, min_freq = 2)
TRG.build_vocab(train_data, min_freq = 2)

一旦运行了这些代码行，SRC.vocab.stoi将是一个词典，其词汇表中的标记作为键，而其对应的索引作为值； SRC.vocab.itos将是相同的字典，其中的键和值被交换。 在本教程中，我们不会广泛使用此事实，但这在您将遇到的其他 NLP 任务中可能很有用。

BucketIterator

我们将使用的最后torchtext个特定功能是BucketIterator，它很容易使用，因为它以TranslationDataset作为第一个参数。 具体来说，正如文档所说：定义一个迭代器，该迭代器将相似长度的示例批处理在一起。 在为每个新纪元生产新鲜改组的批次时，最大程度地减少所需的填充量。 有关使用的存储过程，请参阅池。

import torchdevice = torch.device('cuda' if torch.cuda.is_available() else 'cpu')BATCH_SIZE = 128train_iterator, valid_iterator, test_iterator = BucketIterator.splits((train_data, valid_data, test_data),batch_size = BATCH_SIZE,device = device)

可以像DataLoader``s; below, in the ``train和evaluate函数一样调用这些迭代器，只需使用以下命令即可调用它们：

for i, batch in enumerate(iterator):

每个batch然后具有src和trg属性：

src = batch.src
trg = batch.trg

定义我们的nn.Module和Optimizer

这大部分是从torchtext角度出发的：构建了数据集并定义了迭代器，本教程的其余部分仅将模型定义为nn.Module以及Optimizer，然后对其进行训练。

具体来说，我们的模型遵循在此处中描述的架构(您可以在此处找到更多注释的版本）。

注意：此模型只是可用于语言翻译的示例模型； 我们选择它是因为它是任务的标准模型，而不是因为它是用于翻译的推荐模型。 如您所知，目前最先进的模型基于“变形金刚”； 您可以在此处看到 PyTorch 的实现 Transformer 层的功能； 特别是，以下模型中使用的“注意”与变压器模型中存在的多头自我注意不同。

import random
from typing import Tupleimport torch.nn as nn
import torch.optim as optim
import torch.nn.functional as F
from torch import Tensorclass Encoder(nn.Module):def __init__(self,input_dim: int,emb_dim: int,enc_hid_dim: int,dec_hid_dim: int,dropout: float):super().__init__()self.input_dim = input_dimself.emb_dim = emb_dimself.enc_hid_dim = enc_hid_dimself.dec_hid_dim = dec_hid_dimself.dropout = dropoutself.embedding = nn.Embedding(input_dim, emb_dim)self.rnn = nn.GRU(emb_dim, enc_hid_dim, bidirectional = True)self.fc = nn.Linear(enc_hid_dim * 2, dec_hid_dim)self.dropout = nn.Dropout(dropout)def forward(self,src: Tensor) -> Tuple[Tensor]:embedded = self.dropout(self.embedding(src))outputs, hidden = self.rnn(embedded)hidden = torch.tanh(self.fc(torch.cat((hidden[-2,:,:], hidden[-1,:,:]), dim = 1)))return outputs, hiddenclass Attention(nn.Module):def __init__(self,enc_hid_dim: int,dec_hid_dim: int,attn_dim: int):super().__init__()self.enc_hid_dim = enc_hid_dimself.dec_hid_dim = dec_hid_dimself.attn_in = (enc_hid_dim * 2) + dec_hid_dimself.attn = nn.Linear(self.attn_in, attn_dim)def forward(self,decoder_hidden: Tensor,encoder_outputs: Tensor) -> Tensor:src_len = encoder_outputs.shape[0]repeated_decoder_hidden = decoder_hidden.unsqueeze(1).repeat(1, src_len, 1)encoder_outputs = encoder_outputs.permute(1, 0, 2)energy = torch.tanh(self.attn(torch.cat((repeated_decoder_hidden,encoder_outputs),dim = 2)))attention = torch.sum(energy, dim=2)return F.softmax(attention, dim=1)class Decoder(nn.Module):def __init__(self,output_dim: int,emb_dim: int,enc_hid_dim: int,dec_hid_dim: int,dropout: int,attention: nn.Module):super().__init__()self.emb_dim = emb_dimself.enc_hid_dim = enc_hid_dimself.dec_hid_dim = dec_hid_dimself.output_dim = output_dimself.dropout = dropoutself.attention = attentionself.embedding = nn.Embedding(output_dim, emb_dim)self.rnn = nn.GRU((enc_hid_dim * 2) + emb_dim, dec_hid_dim)self.out = nn.Linear(self.attention.attn_in + emb_dim, output_dim)self.dropout = nn.Dropout(dropout)def _weighted_encoder_rep(self,decoder_hidden: Tensor,encoder_outputs: Tensor) -> Tensor:a = self.attention(decoder_hidden, encoder_outputs)a = a.unsqueeze(1)encoder_outputs = encoder_outputs.permute(1, 0, 2)weighted_encoder_rep = torch.bmm(a, encoder_outputs)weighted_encoder_rep = weighted_encoder_rep.permute(1, 0, 2)return weighted_encoder_repdef forward(self,input: Tensor,decoder_hidden: Tensor,encoder_outputs: Tensor) -> Tuple[Tensor]:input = input.unsqueeze(0)embedded = self.dropout(self.embedding(input))weighted_encoder_rep = self._weighted_encoder_rep(decoder_hidden,encoder_outputs)rnn_input = torch.cat((embedded, weighted_encoder_rep), dim = 2)output, decoder_hidden = self.rnn(rnn_input, decoder_hidden.unsqueeze(0))embedded = embedded.squeeze(0)output = output.squeeze(0)weighted_encoder_rep = weighted_encoder_rep.squeeze(0)output = self.out(torch.cat((output,weighted_encoder_rep,embedded), dim = 1))return output, decoder_hidden.squeeze(0)class Seq2Seq(nn.Module):def __init__(self,encoder: nn.Module,decoder: nn.Module,device: torch.device):super().__init__()self.encoder = encoderself.decoder = decoderself.device = devicedef forward(self,src: Tensor,trg: Tensor,teacher_forcing_ratio: float = 0.5) -> Tensor:batch_size = src.shape[1]max_len = trg.shape[0]trg_vocab_size = self.decoder.output_dimoutputs = torch.zeros(max_len, batch_size, trg_vocab_size).to(self.device)encoder_outputs, hidden = self.encoder(src)# first input to the decoder is the <sos> tokenoutput = trg[0,:]for t in range(1, max_len):output, hidden = self.decoder(output, hidden, encoder_outputs)outputs[t] = outputteacher_force = random.random() < teacher_forcing_ratiotop1 = output.max(1)[1]output = (trg[t] if teacher_force else top1)return outputsINPUT_DIM = len(SRC.vocab)
OUTPUT_DIM = len(TRG.vocab)
# ENC_EMB_DIM = 256
# DEC_EMB_DIM = 256
# ENC_HID_DIM = 512
# DEC_HID_DIM = 512
# ATTN_DIM = 64
# ENC_DROPOUT = 0.5
# DEC_DROPOUT = 0.5ENC_EMB_DIM = 32
DEC_EMB_DIM = 32
ENC_HID_DIM = 64
DEC_HID_DIM = 64
ATTN_DIM = 8
ENC_DROPOUT = 0.5
DEC_DROPOUT = 0.5enc = Encoder(INPUT_DIM, ENC_EMB_DIM, ENC_HID_DIM, DEC_HID_DIM, ENC_DROPOUT)attn = Attention(ENC_HID_DIM, DEC_HID_DIM, ATTN_DIM)dec = Decoder(OUTPUT_DIM, DEC_EMB_DIM, ENC_HID_DIM, DEC_HID_DIM, DEC_DROPOUT, attn)model = Seq2Seq(enc, dec, device).to(device)def init_weights(m: nn.Module):for name, param in m.named_parameters():if 'weight' in name:nn.init.normal_(param.data, mean=0, std=0.01)else:nn.init.constant_(param.data, 0)model.apply(init_weights)optimizer = optim.Adam(model.parameters())def count_parameters(model: nn.Module):return sum(p.numel() for p in model.parameters() if p.requires_grad)print(f'The model has {count_parameters(model):,} trainable parameters')

Out:

The model has 1,856,685 trainable parameters

注意：特别是在对语言翻译模型的性能进行评分时，我们必须告诉nn.CrossEntropyLoss函数忽略仅填充目标的索引。

PAD_IDX = TRG.vocab.stoi['<pad>']criterion = nn.CrossEntropyLoss(ignore_index=PAD_IDX)

最后，我们可以训练和评估该模型：

import math
import timedef train(model: nn.Module,iterator: BucketIterator,optimizer: optim.Optimizer,criterion: nn.Module,clip: float):model.train()epoch_loss = 0for _, batch in enumerate(iterator):src = batch.srctrg = batch.trgoptimizer.zero_grad()output = model(src, trg)output = output[1:].view(-1, output.shape[-1])trg = trg[1:].view(-1)loss = criterion(output, trg)loss.backward()torch.nn.utils.clip_grad_norm_(model.parameters(), clip)optimizer.step()epoch_loss += loss.item()return epoch_loss / len(iterator)def evaluate(model: nn.Module,iterator: BucketIterator,criterion: nn.Module):model.eval()epoch_loss = 0with torch.no_grad():for _, batch in enumerate(iterator):src = batch.srctrg = batch.trgoutput = model(src, trg, 0) #turn off teacher forcingoutput = output[1:].view(-1, output.shape[-1])trg = trg[1:].view(-1)loss = criterion(output, trg)epoch_loss += loss.item()return epoch_loss / len(iterator)def epoch_time(start_time: int,end_time: int):elapsed_time = end_time - start_timeelapsed_mins = int(elapsed_time / 60)elapsed_secs = int(elapsed_time - (elapsed_mins * 60))return elapsed_mins, elapsed_secsN_EPOCHS = 10
CLIP = 1best_valid_loss = float('inf')for epoch in range(N_EPOCHS):start_time = time.time()train_loss = train(model, train_iterator, optimizer, criterion, CLIP)valid_loss = evaluate(model, valid_iterator, criterion)end_time = time.time()epoch_mins, epoch_secs = epoch_time(start_time, end_time)print(f'Epoch: {epoch+1:02} | Time: {epoch_mins}m {epoch_secs}s')print(f'\tTrain Loss: {train_loss:.3f} | Train PPL: {math.exp(train_loss):7.3f}')print(f'\t Val. Loss: {valid_loss:.3f} |  Val. PPL: {math.exp(valid_loss):7.3f}')test_loss = evaluate(model, test_iterator, criterion)print(f'| Test Loss: {test_loss:.3f} | Test PPL: {math.exp(test_loss):7.3f} |')

Out:

Epoch: 01 | Time: 0m 35sTrain Loss: 5.667 | Train PPL: 289.080Val. Loss: 5.201 |  Val. PPL: 181.371
Epoch: 02 | Time: 0m 35sTrain Loss: 4.968 | Train PPL: 143.728Val. Loss: 5.096 |  Val. PPL: 163.375
Epoch: 03 | Time: 0m 35sTrain Loss: 4.720 | Train PPL: 112.221Val. Loss: 4.989 |  Val. PPL: 146.781
Epoch: 04 | Time: 0m 35sTrain Loss: 4.586 | Train PPL:  98.094Val. Loss: 4.841 |  Val. PPL: 126.612
Epoch: 05 | Time: 0m 35sTrain Loss: 4.430 | Train PPL:  83.897Val. Loss: 4.809 |  Val. PPL: 122.637
Epoch: 06 | Time: 0m 35sTrain Loss: 4.331 | Train PPL:  75.997Val. Loss: 4.797 |  Val. PPL: 121.168
Epoch: 07 | Time: 0m 35sTrain Loss: 4.240 | Train PPL:  69.434Val. Loss: 4.694 |  Val. PPL: 109.337
Epoch: 08 | Time: 0m 35sTrain Loss: 4.116 | Train PPL:  61.326Val. Loss: 4.714 |  Val. PPL: 111.452
Epoch: 09 | Time: 0m 35sTrain Loss: 4.004 | Train PPL:  54.815Val. Loss: 4.563 |  Val. PPL:  95.835
Epoch: 10 | Time: 0m 36sTrain Loss: 3.922 | Train PPL:  50.519Val. Loss: 4.452 |  Val. PPL:  85.761
| Test Loss: 4.456 | Test PPL:  86.155 |