任务背景
利用LSTM(长短期记忆)网络结构训练小样本文本分类任务。
数据集及代码如下:
LSTM文本分类数据集+代码+模型
一、Model/TextRNN.py
# coding: UTF-8
import torch
import torch.nn as nn
import torch.nn.functional as F
import numpy as np
class Config(object):
"""配置参数"""
def __init__(self, dataset, embedding):
# 均为相对路径,无需修改直接使用
self.model_name = 'TextRNN'
self.train_path = dataset + '/data/train.txt' # 训练集——相对路径
self.dev_path = dataset + '/data/dev.txt' # 验证集
self.test_path = dataset + '/data/test.txt' # 测试集
self.class_list = [x.strip() for x in open(
dataset + '/data/class.txt').readlines()] # 读取类别名单
self.vocab_path = dataset + '/data/vocab.pkl' # 词表
self.save_path = dataset + '/saved_dict/' + self.model_name + '.ckpt' # 模型训练结果——保存路径
self.log_path = dataset + '/log/' + self.model_name # 训练日志——保存路径
# 加载预训练词向量,以关键字获取,数据转32位浮点型再转tensor格式,(4762, 300)
# NpzFile 'D:\\咕泡人工智能-配套资料\\配套资料\\4.第四章 深度学习核⼼框架PyTorch\\第七章:LSTM文本分类实战\\text\\THUCNews/data/embedding_SougouNews.npz' with keys: embeddings
self.embedding_pretrained = torch.tensor(
np.load(dataset + '/data/' + embedding)["embeddings"].astype('float32'))\
if embedding != 'random' else None # 预训练词向量
self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') # 设备,cpu还是gpu
self.dropout = 0.5 # 随机失活
self.require_improvement = 1000 # 若超过1000batch效果还没提升,则提前结束训练
self.num_classes = len(self.class_list) # 类别数
self.n_vocab = 0 # 词表大小,在运行时赋值
self.num_epochs = 10 # epoch数
self.batch_size = 128 # mini-batch大小
self.pad_size = 40 # 每句话处理成的长度(短填长切)
self.learning_rate = 1e-3 # 学习率
self.embed = self.embedding_pretrained.size(1)\
if self.embedding_pretrained is not None else 300 # 字向量维度, 若使用了预训练词向量,则维度统一,300
self.hidden_size = 128 # lstm隐藏层
self.num_layers = 3 # lstm层数
'''Recurrent Neural Network for Text Classification with Multi-Task Learning'''
# <bound method Module.parameters of Model(
# (embedding): Embedding(4762, 300)
# (lstm): LSTM(300, 128, num_layers=3, batch_first=True, dropout=0.5, bidirectional=True)
# (fc): Linear(in_features=256, out_features=10, bias=True)
# )>
class Model(nn.Module):
def __init__(self, config):
super(Model, self).__init__()
# 获取预训练模型数据
if config.embedding_pretrained is not None:
# print(config.embedding_pretrained)
# tensor([[0.2983, 0.4106, 0.8946, ..., 0.6417, 0.8806, 0.1683],
# [0.0010, -0.1334, -0.1902, ..., -0.1443, -0.5212, 0.2069],
# [-0.0249, 0.1308, -0.4010, ..., 0.3485, -0.5099, -0.1834],
# ...,
# [-0.2030, 0.1445, -0.0035, ..., -0.2927, -0.1555, 0.0662],
# [0.5052, 0.6515, 0.4099, ..., 0.6391, 0.2736, 0.7934],
# [0.2890, 0.8864, 0.6253, ..., 0.6731, 0.7833, 0.1340]])
self.embedding = nn.Embedding.from_pretrained(config.embedding_pretrained, freeze=False)
# print(self.embedding)
# Embedding(4762, 300)
else:
self.embedding = nn.Embedding(config.n_vocab, config.embed, padding_idx=config.n_vocab - 1)
# (lstm): LSTM(300, 128, num_layers=3, batch_first=True, dropout=0.5, bidirectional=True)
self.lstm = nn.LSTM(config.embed, config.hidden_size, config.num_layers,
bidirectional=True, batch_first=True, dropout=config.dropout)
# bidirectional=True:RNN每层正向传递得到128的向量与反向传递得到128的向量结合,得到256的向量
# batch_first=True:输入张量的形状应为(batch, pad_size, embedding),即batch_size,句子长度, 词向量长度;输出张量的形状是(batch, pad_size, hidden_size)
# 输出层,(256,10)
# (fc): Linear(in_features=256, out_features=10, bias=True)
self.fc = nn.Linear(config.hidden_size * 2, config.num_classes)
def forward(self, x):
# print(x)
# (tensor([[14, 125, 55, ..., 4760, 4760, 4760],
# [135, 80, 33, ..., 4760, 4760, 4760],
# [152, 13, 469, ..., 4760, 4760, 4760],
# ...,
# [160, 1667, 1147, ..., 4760, 4760, 4760],
# [31, 75, 4, ..., 4760, 4760, 4760],
# [321, 566, 130, ..., 4760, 4760, 4760]]),
# tensor([18, 22, 25, 25, 23, 20, 17, 22, 16, 11, 23, 23, 22, 15, 7, 23, 20, 25,
# 15, 9, 17, 15, 24, 20, 17, 17, 13, 20, 19, 20, 22, 22, 21, 22, 23, 19,
# 12, 20, 23, 18, 22, 25, 23, 20, 19, 17, 17, 15, 17, 26, 16, 22, 21, 18,
# 16, 12, 23, 19, 20, 21, 12, 24, 18, 14, 25, 16, 24, 24, 23, 20, 20, 20,
# 18, 16, 23, 14, 23, 21, 19, 17, 24, 21, 23, 23, 19, 15, 12, 22, 25, 14,
# 21, 20, 22, 15, 22, 18, 16, 17, 13, 21, 21, 18, 21, 11, 19, 22, 14, 22,
# 15, 22, 15, 22, 22, 15, 25, 16, 18, 18, 14, 19, 13, 29, 20, 18, 22, 16,
# 18, 22]))
x, _ = x
# print(x)
# tensor([[14, 125, 55, ..., 4760, 4760, 4760],
# [135, 80, 33, ..., 4760, 4760, 4760],
# [152, 13, 469, ..., 4760, 4760, 4760],
# ...,
# [160, 1667, 1147, ..., 4760, 4760, 4760],
# [31, 75, 4, ..., 4760, 4760, 4760],
# [321, 566, 130, ..., 4760, 4760, 4760]])
# 每个ID转换为embedding中的词向量
out = self.embedding(x) # [batch_size, seq_len, embeding]=[128, 40, 300]
# 输入LSTM网络训练
out, _ = self.lstm(out) # out: torch.Size([128, 40, 256])
# 取最后一个词的结果
out = self.fc(out[:, -1, :]) # 句子最后时刻的 hidden state
return out
2、THUCNews/run.py
import time
import torch
import numpy as np
from train_eval import train, init_network
from importlib import import_module
import argparse
# 可视化脚本
from tensorboardX import SummaryWriter
# 参数命令行
parser = argparse.ArgumentParser(description='Chinese Text Classification')
# 选择模型
parser.add_argument('--model', type=str, required=True, help='choose a model: TextCNN, TextRNN, FastText, TextRCNN, TextRNN_Att, DPCNN, Transformer')
# 选择预训练Embedding表
parser.add_argument('--embedding', default='pre_trained', type=str, help='random or pre_trained')
# 词分割还是字分割
parser.add_argument('--word', default=False, type=bool, help='True for word, False for char')
args = parser.parse_args()
# model='TextRNN',embedding='pre_trained', word=False
if __name__ == '__main__':
# 数据集——相对路径
dataset = 'THUCNews' # 数据集
# 搜狗新闻:embedding_SougouNews.npz, 腾讯:embedding_Tencent.npz, 随机初始化:random
embedding = 'embedding_SougouNews.npz'
if args.embedding == 'random':
embedding = 'random'
model_name = args.model #TextCNN, TextRNN,
if model_name == 'FastText':
from utils_fasttext import build_dataset, build_iterator, get_time_dif
embedding = 'random'
else:
from utils import build_dataset, build_iterator, get_time_dif
x = import_module('models.' + model_name)
config = x.Config(dataset, embedding)
# config: <models.TextRNN.Config object at 0x00000243F8850B80>
# 随机数种子,确保每次训练随机结果相同
np.random.seed(1)
torch.manual_seed(1)
torch.cuda.manual_seed_all(1)
torch.backends.cudnn.deterministic = True # 保证每次结果一样
start_time = time.time()
print("Loading data...")
# 将训练集,验证集,测试集中的文本数据转换为词ID形式,格式为(文本词ID, 标签, 文本长度)
# ([51, 112, 19, 31, 439, 726, 296, 1028, 0, 179, 51, 47, 62, 2070, 669, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760], 2, 15)
vocab, train_data, dev_data, test_data = build_dataset(config, args.word)
# 划分数据集
train_iter = build_iterator(train_data, config)
# print(train_iter)
# <utils.DatasetIterater object at 0x0000028AEE21DC40>
dev_iter = build_iterator(dev_data, config)
test_iter = build_iterator(test_data, config)
time_dif = get_time_dif(start_time)
print("Time usage:", time_dif)
# train
# 词表长度Vocab size: 4762
config.n_vocab = len(vocab)
model = x.Model(config).to(config.device)
writer = SummaryWriter(log_dir=config.log_path + '/' + time.strftime('%m-%d_%H.%M', time.localtime()))
if model_name != 'Transformer':
init_network(model)
print(model.parameters)
# <bound method Module.parameters of Model(
# (embedding): Embedding(4762, 300)
# (lstm): LSTM(300, 128, num_layers=3, batch_first=True, dropout=0.5, bidirectional=True)
# (fc): Linear(in_features=256, out_features=10, bias=True)
# )>
train(config, model, train_iter, dev_iter, test_iter, writer)
三、THUCNews/train_eval.py
# coding: UTF-8
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
from sklearn import metrics
import time
from utils import get_time_dif
from tensorboardX import SummaryWriter
# 权重初始化,默认xavier
def init_network(model, method='xavier', exclude='embedding', seed=123):
# print(model.named_parameters)
# <bound method Module.named_parameters of Model(
# (embedding): Embedding(4762, 300)
# (lstm): LSTM(300, 128, num_layers=3, batch_first=True, dropout=0.5, bidirectional=True)
# (fc): Linear(in_features=256, out_features=10, bias=True)
# )>
for name, w in model.named_parameters():
# print(name)
# print(w.shape)
# embedding.weight
# torch.Size([4762, 300])
# lstm.weight_ih_l0
# torch.Size([512, 300])
# lstm.weight_hh_l0
# torch.Size([512, 128])
# lstm.bias_ih_l0
# torch.Size([512])
# lstm.bias_hh_l0
# torch.Size([512])
# lstm.weight_ih_l0_reverse
# torch.Size([512, 300])
# lstm.weight_hh_l0_reverse
# torch.Size([512, 128])
# lstm.bias_ih_l0_reverse
# torch.Size([512])
# lstm.bias_hh_l0_reverse
# torch.Size([512])
# lstm.weight_ih_l1
# torch.Size([512, 256])
# lstm.weight_hh_l1
# torch.Size([512, 128])
# lstm.bias_ih_l1
# torch.Size([512])
# lstm.bias_hh_l1
# torch.Size([512])
# lstm.weight_ih_l1_reverse
# torch.Size([512, 256])
# lstm.weight_hh_l1_reverse
# torch.Size([512, 128])
# lstm.bias_ih_l1_reverse
# torch.Size([512])
# lstm.bias_hh_l1_reverse
# torch.Size([512])
# lstm.weight_ih_l2
# torch.Size([512, 256])
# lstm.weight_hh_l2
# torch.Size([512, 128])
# lstm.bias_ih_l2
# torch.Size([512])
# lstm.bias_hh_l2
# torch.Size([512])
# lstm.weight_ih_l2_reverse
# torch.Size([512, 256])
# lstm.weight_hh_l2_reverse
# torch.Size([512, 128])
# lstm.bias_ih_l2_reverse
# torch.Size([512])
# lstm.bias_hh_l2_reverse
# torch.Size([512])
# fc.weight
# torch.Size([10, 256])
# fc.bias
# torch.Size([10])
if exclude not in name:
if 'weight' in name:
if method == 'xavier':
nn.init.xavier_normal_(w)
elif method == 'kaiming':
nn.init.kaiming_normal_(w)
else:
nn.init.normal_(w)
elif 'bias' in name:
nn.init.constant_(w, 0)
else:
pass
def train(config, model, train_iter, dev_iter, test_iter,writer):
start_time = time.time()
# 训练模式。dropout随机失活,标准化和归一化参数更新
model.train()
optimizer = torch.optim.Adam(model.parameters(), lr=config.learning_rate)
# 学习率指数衰减,每次epoch:学习率 = gamma * 学习率
# scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer, gamma=0.9)
total_batch = 0 # 记录进行到多少batch
dev_best_loss = float('inf')
last_improve = 0 # 记录上次验证集loss下降的batch数
flag = False # 记录是否很久没有效果提升
#writer = SummaryWriter(log_dir=config.log_path + '/' + time.strftime('%m-%d_%H.%M', time.localtime()))
for epoch in range(config.num_epochs):
print('Epoch [{}/{}]'.format(epoch + 1, config.num_epochs))
# scheduler.step() # 学习率衰减
for i, (trains, labels) in enumerate(train_iter):
#print (trains[0].shape)
# train为元组格式,train[0]为词向量,train[1]为词长度
outputs = model(trains)
# 梯度清零
model.zero_grad()
loss = F.cross_entropy(outputs, labels)
loss.backward()
optimizer.step()
if total_batch % 100 == 0:
# 每多少轮输出在训练集和验证集上的效果
true = labels.data.cpu()
predic = torch.max(outputs.data, 1)[1].cpu()
train_acc = metrics.accuracy_score(true, predic)
dev_acc, dev_loss = evaluate(config, model, dev_iter)
if dev_loss < dev_best_loss:
dev_best_loss = dev_loss
torch.save(model.state_dict(), config.save_path)
improve = '*'
last_improve = total_batch
else:
improve = ''
time_dif = get_time_dif(start_time)
msg = 'Iter: {0:>6}, Train Loss: {1:>5.2}, Train Acc: {2:>6.2%}, Val Loss: {3:>5.2}, Val Acc: {4:>6.2%}, Time: {5} {6}'
print(msg.format(total_batch, loss.item(), train_acc, dev_loss, dev_acc, time_dif, improve))
writer.add_scalar("loss/train", loss.item(), total_batch)
writer.add_scalar("loss/dev", dev_loss, total_batch)
writer.add_scalar("acc/train", train_acc, total_batch)
writer.add_scalar("acc/dev", dev_acc, total_batch)
model.train()
total_batch += 1
if total_batch - last_improve > config.require_improvement:
# 验证集loss超过1000batch没下降,结束训练
print("No optimization for a long time, auto-stopping...")
flag = True
break
if flag:
break
writer.close()
test(config, model, test_iter)
def test(config, model, test_iter):
# test
model.load_state_dict(torch.load(config.save_path))
# 验证模式。dropout不会随机失活,标准化和归一化参数使用训练时的参数
model.eval()
start_time = time.time()
test_acc, test_loss, test_report, test_confusion = evaluate(config, model, test_iter, test=True)
msg = 'Test Loss: {0:>5.2}, Test Acc: {1:>6.2%}'
print(msg.format(test_loss, test_acc))
print("Precision, Recall and F1-Score...")
print(test_report)
print("Confusion Matrix...")
print(test_confusion)
time_dif = get_time_dif(start_time)
print("Time usage:", time_dif)
def evaluate(config, model, data_iter, test=False):
model.eval()
loss_total = 0
predict_all = np.array([], dtype=int)
labels_all = np.array([], dtype=int)
with torch.no_grad():
for texts, labels in data_iter:
outputs = model(texts)
loss = F.cross_entropy(outputs, labels)
loss_total += loss
labels = labels.data.cpu().numpy()
predic = torch.max(outputs.data, 1)[1].cpu().numpy()
labels_all = np.append(labels_all, labels)
predict_all = np.append(predict_all, predic)
acc = metrics.accuracy_score(labels_all, predict_all)
if test:
report = metrics.classification_report(labels_all, predict_all, target_names=config.class_list, digits=4)
confusion = metrics.confusion_matrix(labels_all, predict_all)
return acc, loss_total / len(data_iter), report, confusion
return acc, loss_total / len(data_iter)
四、THUCNews/utils.py
# coding: UTF-8
import os
import torch
import numpy as np
import pickle as pkl
from tqdm import tqdm
import time
from datetime import timedelta
MAX_VOCAB_SIZE = 10000 # 词表长度限制
# UNK未知字符,如果文本数据中出现了语料表中没有的字,那么使用UNK填充;如果字符长度不足pad_size,使用PAD填充
# 如果embedding中使用到4759位,数据中出现语料表中不存在的字,那么用4760表示所有语料表中不存在的字的ID,用4761表示PAD填充,如果不存在UNK,那么用4760表示PAD————个人理解,不一定正确
UNK, PAD = '<UNK>', '<PAD>' # 未知字,padding符号
def build_vocab(file_path, tokenizer, max_size, min_freq):
vocab_dic = {}
with open(file_path, 'r', encoding='UTF-8') as f:
for line in tqdm(f):
lin = line.strip()
if not lin:
continue
content = lin.split('\t')[0]
for word in tokenizer(content):
vocab_dic[word] = vocab_dic.get(word, 0) + 1
vocab_list = sorted([_ for _ in vocab_dic.items() if _[1] >= min_freq], key=lambda x: x[1], reverse=True)[:max_size]
vocab_dic = {word_count[0]: idx for idx, word_count in enumerate(vocab_list)}
vocab_dic.update({UNK: len(vocab_dic), PAD: len(vocab_dic) + 1})
return vocab_dic
def build_dataset(config, ues_word):
# 判断以词分割还是字分割
if ues_word:
tokenizer = lambda x: x.split(' ') # 以空格隔开,word-level
else:
tokenizer = lambda x: [y for y in x] # char-level
# 加载语料表
if os.path.exists(config.vocab_path):
vocab = pkl.load(open(config.vocab_path, 'rb'))
else:
vocab = build_vocab(config.train_path, tokenizer=tokenizer, max_size=MAX_VOCAB_SIZE, min_freq=1)
pkl.dump(vocab, open(config.vocab_path, 'wb'))
print(f"Vocab size: {len(vocab)}")
def load_dataset(path, pad_size=32):
contents = []
with open(path, 'r', encoding='UTF-8') as f:
for line in tqdm(f):
# line: '中华女子学院:本科层次仅1专业招男生\t3\n'
# 删去换行符
lin = line.strip()
# lin: '中华女子学院:本科层次仅1专业招男生\t3'
if not lin:
continue
# 以转义字符\t分割,得到句子和标签
# content: '中华女子学院:本科层次仅1专业招男生' label: '3'
content, label = lin.split('\t')
words_line = []
# 以字分割句子,不足pad_size部分补足,此处填充ID为4761
token = tokenizer(content)
# ['中', '华', '女', '子', '学', '院', ':', '本', '科', '层', '次', '仅', '1', '专', '业', '招', '男', '生', 4761, 4761, 4761, 4761, 4761, 4761, 4761, 4761, 4761, 4761, 4761, 4761, 4761, 4761]
seq_len = len(token)
if pad_size:
# 句子长度小于pad_size,不足pad_size部分补足PAD
if len(token) < pad_size:
token.extend([vocab.get(PAD)] * (pad_size - len(token)))
else:
# 句子长度大于pad_size,多余部分舍去
token = token[:pad_size]
seq_len = pad_size
# word to id,文本字转语料表对应ID
for word in token:
# ID不存在的用UNK代替
words_line.append(vocab.get(word, vocab.get(UNK)))
contents.append((words_line, int(label), seq_len))
# 字ID列表,标签,长度 contents: [([14, 125, 55, 45, 35, 307, 4, 81, 161, 941, 258, 494, 2, 175, 48, 145, 97, 17, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760, 4760], 3, 18)]
return contents # [([...], 0), ([...], 1), ...]
train = load_dataset(config.train_path, config.pad_size)
dev = load_dataset(config.dev_path, config.pad_size)
test = load_dataset(config.test_path, config.pad_size)
return vocab, train, dev, test
class DatasetIterater(object):
def __init__(self, batches, batch_size, device):
self.batch_size = batch_size #128
self.batches = batches #训练集180000
self.n_batches = len(batches) // batch_size #划分batch个数
self.residue = False # 记录batch数量是否为整数
# 数据集是否能整除batch数
if len(batches) % self.n_batches != 0:
self.residue = True
self.index = 0
self.device = device
def _to_tensor(self, datas):
x = torch.LongTensor([_[0] for _ in datas]).to(self.device)
y = torch.LongTensor([_[1] for _ in datas]).to(self.device)
# pad前的长度(超过pad_size的设为pad_size)
seq_len = torch.LongTensor([_[2] for _ in datas]).to(self.device)
return (x, seq_len), y
def __next__(self):
if self.residue and self.index == self.n_batches:
batches = self.batches[self.index * self.batch_size: len(self.batches)]
self.index += 1
batches = self._to_tensor(batches)
return batches
elif self.index > self.n_batches:
self.index = 0
raise StopIteration
else:
batches = self.batches[self.index * self.batch_size: (self.index + 1) * self.batch_size]
self.index += 1
batches = self._to_tensor(batches)
return batches
def __iter__(self):
return self
def __len__(self):
if self.residue:
return self.n_batches + 1
else:
return self.n_batches
def build_iterator(dataset, config):
iter = DatasetIterater(dataset, config.batch_size, config.device)
return iter
def get_time_dif(start_time):
"""获取已使用时间"""
end_time = time.time()
time_dif = end_time - start_time
return timedelta(seconds=int(round(time_dif)))
if __name__ == "__main__":
'''提取预训练词向量'''
# 下面的目录、文件名按需更改。
train_dir = "./THUCNews/data/train.txt"
vocab_dir = "./THUCNews/data/vocab.pkl"
pretrain_dir = "./THUCNews/data/sgns.sogou.char"
emb_dim = 300
filename_trimmed_dir = "./THUCNews/data/embedding_SougouNews"
if os.path.exists(vocab_dir):
word_to_id = pkl.load(open(vocab_dir, 'rb'))
else:
# tokenizer = lambda x: x.split(' ') # 以词为单位构建词表(数据集中词之间以空格隔开)
tokenizer = lambda x: [y for y in x] # 以字为单位构建词表
word_to_id = build_vocab(train_dir, tokenizer=tokenizer, max_size=MAX_VOCAB_SIZE, min_freq=1)
pkl.dump(word_to_id, open(vocab_dir, 'wb'))
embeddings = np.random.rand(len(word_to_id), emb_dim)
f = open(pretrain_dir, "r", encoding='UTF-8')
for i, line in enumerate(f.readlines()):
# if i == 0: # 若第一行是标题,则跳过
# continue
lin = line.strip().split(" ")
if lin[0] in word_to_id:
idx = word_to_id[lin[0]]
emb = [float(x) for x in lin[1:301]]
embeddings[idx] = np.asarray(emb, dtype='float32')
f.close()
np.savez_compressed(filename_trimmed_dir, embeddings=embeddings)
分类结果:
第6个epoch开始训练时由于准确率在1000个batch效果没有提升,训练中断,最佳模型出现在第5个epoch中,后续迭代出现过拟合现象。