Pytorch学习(2)——训练词向量的代码

教程：https://www.bilibili.com/video/BV1vz4y1R7Mm?p=2

import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.data as tud

from collections import Counter
import numpy as np
import random
import math

import pandas as pd
import scipy
import sklearn
from sklearn.metrics.pairwise import cosine_similarity

USE_CUDA = torch.cuda.is_available()
random.seed(1)
np.random.seed(1)
torch.manual_seed(1)
if USE_CUDA:
    torch.cuda.manual_seed(1)

C = 3  # content window
K = 100  # number of negtive samples
NUM_EPOCHS = 2
MAX_VOCAB_SIZE = 30000
BATCH_SIZE = 128
LEARNING_RATE = 0.2
EMBEDDING_SIZE = 100 

def word_tokenize(text):
    return text.split()

• 从文本文件中读取所有的文字，通过这些文本创建一个vocabulary
• 由于单词数量可能太大，我们只选取最常见的MAX_VOCAB_SIZE个单词
• 添加一个UNK单词表示所有不常见的单词
• 我们需要记录单词到index的mapping, 以及index到单词的mapping, 单词的count, 单词的normalized frequency, 单词总数

with open('text8/text8.train.txt', 'r') as fin:
    text = fin.read()
    
text = word_tokenize(text)
vocab = dict(Counter(text).most_common(MAX_VOCAB_SIZE-1))
vocab['<unk>'] = len(text) - np.sum(list(vocab.values()))

idx_to_word = [word for word in vocab.keys()]
word_to_idx = {word:i for i, word in enumerate(idx_to_word)}
word_counts = np.array([count for count in vocab.values()], dtype=np.float32)
word_freqs = word_counts / np.sum(word_counts)

word_counts = np.array([count for count in vocab.values()], dtype=np.float32)
word_freqs = word_counts / np.sum(word_counts)
word_freqs = word_freqs ** (0.75)
word_freqs = word_freqs / np.sum(word_freqs)

实现DataLoader

返回batch的数据

class WordEmbeddingDataset(tud.Dataset):
    def __init__(self, text, word_to_idx, idx_to_word, word_freqs, word_counts):
        super()
        self.text_encoded = [word_to_idx.get(word, word_to_idx['<unk>']) for word in text]
        self.text_encoded = torch.LongTensor(self.text_encoded)
        self.word_to_idx = word_to_idx   
        self.word_freqs = torch.Tensor(word_freqs)
        self.word_counts = torch.Tensor(word_counts)
    
    def __len__(self):
        return len(self.text_encoded)
    
    def __getitem__(self, idx):
        center_word = self.text_encoded[idx]
        pos_indices = list(range(idx-C, idx)) + list(range(idx+1, idx+C+1))  # window内的单词（写错了）
        pos_indices = [i % len(self.text_encoded) for i in pos_indices]  # 取余，放置超出长度
        pos_words = self.text_encoded[pos_indices]  # 周围单词
        neg_words = torch.multinomial(self.word_freqs, K * pos_words.shape[0], True)  # 负采样单词
        # Returns a tensor where each row contains num_samples indices sampled from 
        # the multinomial probability distribution located in the corresponding row
        # of tensor input. (多项式概率分布)
        return center_word, pos_words, neg_words

dataset = WordEmbeddingDataset(text, word_to_idx, idx_to_word, word_freqs, word_counts)
dataloader = tud.DataLoader(dataset, batch_size=BATCH_SIZE, shuffle=True, num_workers=0)

dataset.word_freqs.shape

class EmbeddingModel(nn.Module):
    def __init__(self, vocab_size, embed_size):
        super(EmbeddingModel, self).__init__()
        self.vocab_size = vocab_size
        self.embed_size = embed_size
        
        self.in_embed = nn.Embedding(vocab_size, embed_size)
        self.out_embed = nn.Embedding(vocab_size, embed_size)
    
    def forward(self, input_labels, pos_labels, neg_labels):
        # input_label: [batch_size]
        # pos_labels: [batch_size, (window_size * 2)]
        # neg_labels: [batch_size, (window_size * 2 * K)]
        input_embedding = self.in_embed(input_labels)  # [batch_size, embed_size]
        pos_embedding = self.out_embed(pos_labels)  # [batch_size, (window_size*2), embed_size]
        neg_embedding = self.out_embed(neg_labels)  # [batch_size, (window_size*2)*K, embed_size]
        
        input_embedding = input_embedding.unsqueeze(2)  # [batch_size, embed_size, 1]
        pos_dot = torch.bmm(pos_embedding, input_embedding).squeeze(2)
        # documention of torch.bmm: batch matrix-matrix product
        # Performs a batch matrix-matrix product of matrices stored in input and mat2.
        # if A.shape=[batch_size, n, m], B.shape=[batch_size, m, p],
        # then torch.bmm(A, B).shape is [batch_size, n, p]
        neg_dot = torch.bmm(neg_embedding, -input_embedding).squeeze(2)
        
        log_pos = F.logsigmoid(pos_dot).sum(1)
        log_neg = F.logsigmoid(neg_dot).sum(1)
        loss= log_pos+log_neg
        return -loss
    
    def input_embeddings(self):
        return self.in_embed.weight.data.cpu().numpy()

model = EmbeddingModel(MAX_VOCAB_SIZE, EMBEDDING_SIZE)
if USE_CUDA:
    model = model.cuda()

optimizer = torch.optim.SGD(model.parameters(), lr=LEARNING_RATE)
for e in range(NUM_EPOCHS):
    for i, (input_labels, pos_labels, neg_labels) in enumerate(dataloader):
#         print(input_labels, pos_labels, neg_labels)
#         if i > 5: break
        input_labels = input_labels.long()
        pos_labels = pos_labels.long()
        neg_labels = neg_labels.long()
        if USE_CUDA:
            input_labels = input_labels.cuda()
            pos_labels = pos_labels.cuda()
            neg_labels = neg_labels.cuda()
        
        optimizer.zero_grad()
        loss = model(input_labels, pos_labels, neg_labels).mean()
        loss.backward()
        optimizer.step()

        # if i % 100 == 0:
        print("epoch", e, "iteration", i, loss.item())