文章目录
https://github.com/bentrevett/pytorch-sentiment-analysis/blob/master/5%20-%20Multi-class%20Sentiment%20Analysis.ipynb
Multi-class Sentiment Analysis
之前的notebook中展现了2分类情感分析,negative、positive
当我们只有两个类别时,我们的输出可以是单个标量(没有方向:int,float…),在0-1之间.代表着一个样本的标签是什么
当我们有超过2个样本时,我们的输出必须是C维度向量,C是类的数量
这个笔记里,我们会进行6分类问题,要说明的是这个数据库不是情感分析数据集,该数据集里的样本都是一个个问题,任务是将这些问题分类成6个类别.但不妨碍,这个笔记教我们如何做多分类问题
跟二分类问题相比,我们不需要设置LABEL field中的dtype.当做一个多分类问题时,PyTorch希望标签是数值化的LongTensors类型
第二个区别是,我们使用TREC数据集来代替IMDB,其中的fine_grained参数控制的是选择50类分类还是6分类,这里选false 6分类
1.加载数据
import torch
from torchtext import data
from torchtext import datasets
import random
SEED = 1234
torch.manual_seed(SEED)
torch.backends.cudnn.deterministic = True
TEXT = data.Field(tokenize = 'spacy')
LABEL = data.LabelField()
train_data, test_data = datasets.TREC.splits(TEXT, LABEL, fine_grained=False)
train_data, valid_data = train_data.split(random_state = random.seed(SEED))
查看数据样本
vars(train_data[-1])
{‘text’: [‘What’, ‘is’, ‘a’, ‘Cartesian’, ‘Diver’, ‘?’], ‘label’: ‘DESC’}
2.构建词向量,词表
因为数据集小,只有3800个训练样本,词汇表也很小,只有7500 unique tokens,意味着,我们不需要像以前一样设置max_size.这里还是设置25000是因为,实际的词汇数量比这个小.如果实际词汇很大,需要设置最大词汇量来舍去词频靠后的词
MAX_VOCAB_SIZE = 25_000
TEXT.build_vocab(train_data,
max_size = MAX_VOCAB_SIZE,
vectors = "glove.6B.100d",
unk_init = torch.Tensor.normal_)
LABEL.build_vocab(train_data)
3.查看labels
6个label:
1.HUM for questions about humans 关于人类的问题
2.ENTY for questions about entities 关于实体的问题
3.DESC for questions asking you for a description 关于形容的问题
4.NUM for questions where the answer is numerical 关于数字的问题
5.LOC for questions where the answer is a location 关于地点的问题
6.ABBR for questions asking about abbreviations 关于缩略语的问题
print(LABEL.vocab.stoi)
defaultdict(<function _default_unk_index at 0x7f0a50190d08>, {‘HUM’: 0, ‘ENTY’: 1, ‘DESC’: 2, ‘NUM’: 3, ‘LOC’: 4, ‘ABBR’: 5})
4.设置iterators迭代器
BATCH_SIZE = 64
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
train_iterator, valid_iterator, test_iterator = data.BucketIterator.splits(
(train_data, valid_data, test_data),
batch_size = BATCH_SIZE,
device = device)
5.创建模型
原文用的是CNN卷积神经网络,
import torch.nn as nn
import torch.nn.functional as F
class CNN(nn.Module):
def __init__(self, vocab_size, embedding_dim, n_filters, filter_sizes, output_dim,
dropout, pad_idx):
super().__init__()
self.embedding = nn.Embedding(vocab_size, embedding_dim)
self.convs = nn.ModuleList([
nn.Conv2d(in_channels = 1,
out_channels = n_filters,
kernel_size = (fs, embedding_dim))
for fs in filter_sizes
])
self.fc = nn.Linear(len(filter_sizes) * n_filters, output_dim)
self.dropout = nn.Dropout(dropout)
def forward(self, text):
#text = [sent len, batch size]
text = text.permute(1, 0)
#text = [batch size, sent len]
embedded = self.embedding(text)
#embedded = [batch size, sent len, emb dim]
embedded = embedded.unsqueeze(1)
#embedded = [batch size, 1, sent len, emb dim]
conved = [F.relu(conv(embedded)).squeeze(3) for conv in self.convs]
#conv_n = [batch size, n_filters, sent len - filter_sizes[n]]
pooled = [F.max_pool1d(conv, conv.shape[2]).squeeze(2) for conv in conved]
#pooled_n = [batch size, n_filters]
cat = self.dropout(torch.cat(pooled, dim = 1))
#cat = [batch size, n_filters * len(filter_sizes)]
return self.fc(cat)
6.实例化模型
INPUT_DIM = len(TEXT.vocab)
EMBEDDING_DIM = 100
N_FILTERS = 100
FILTER_SIZES = [2,3,4]
OUTPUT_DIM = len(LABEL.vocab)
DROPOUT = 0.5
PAD_IDX = TEXT.vocab.stoi[TEXT.pad_token]
model = CNN(INPUT_DIM, EMBEDDING_DIM, N_FILTERS, FILTER_SIZES, OUTPUT_DIM, DROPOUT, PAD_IDX)
7.数模型中有多少个参数需要训练
def count_parameters(model):
return sum(p.numel() for p in model.parameters() if p.requires_grad)
print(f'The model has {count_parameters(model):,} trainable parameters')
The model has 3,063,522 trainable parameters
8.加载预训练词向量到模型中
pretrained_embeddings = TEXT.vocab.vectors
model.embedding.weight.data.copy_(pretrained_embeddings)
9.将unknwon,padding tokens的向量进行0初始化
UNK_IDX = TEXT.vocab.stoi[TEXT.unk_token]
model.embedding.weight.data[UNK_IDX] = torch.zeros(EMBEDDING_DIM)
model.embedding.weight.data[PAD_IDX] = torch.zeros(EMBEDDING_DIM)
10.设置损失函数
与前几个notebook比,损失函数(也就是criterion)是不同的.之前用的是BCEWithLogisLoss,现在使用的是CrossEntropyLoss,它使用的是softrmax函数,来计算cross entropy
一般来说:
CrossEntropyLoss :用于多分类问题
BCEWithLogitsLoss :用于2分类问题,(0,1),也用于多标签分类(multilabel classification) 1vs rest
import torch.optim as optim
optimizer = optim.Adam(model.parameters())
criterion = nn.CrossEntropyLoss()
model = model.to(device)
criterion = criterion.to(device)
10.构建精确度函数
之前计算2分类问题的精确度时,我们假设输出超过0.5就是positive,小于0.5时negative.
这里,我们有6分类,输出的时6维的向量,每个元素是其属于某个分类的概率(each element is the beleief that the example belongs to that class.)
举例:‘HUM’ = 0, ‘ENTY’ = 1, ‘DESC’ = 2, ‘NUM’ = 3, ‘LOC’ = 4 and ‘ABBR’ = 5.,
一个句子经过模型预测后得到的输出值是:[5.1, 0.3, 0.1, 2.1, 0.2, 0.6],意味着,模型有很强的信念认为,这个样本属于分类0,也就是有关人类的问题.并且有点认为,这个样本属于分类3,一个数字问题
我们使用argmax获取预测输出中哪个元素最大,将他认为是预测值,跟实际值进行计算准确度
def categorical_accuracy(preds, y):
"""
Returns accuracy per batch, i.e. if you get 8/10 right, this returns 0.8, NOT 8
"""
max_preds = preds.argmax(dim = 1, keepdim = True) # get the index of the max probability
correct = max_preds.squeeze(1).eq(y)
return correct.sum() / torch.FloatTensor([y.shape[0]]).to(device) # 这里要制定.to(device) 设置在GPU、CPU跑要不然之后会报错
11.构建训练函数
训练循环,跟之前几个notebook很像,只是不需要squeeze模型的预测值.因为,CrossEntropyLoss 希望其输入的形状是[batch size, n classes]
label形状是[batch size]
def train(model, iterator, optimizer, criterion):
epoch_loss = 0
epoch_acc = 0
model.train()
for batch in iterator:
optimizer.zero_grad()
predictions = model(batch.text)
loss = criterion(predictions, batch.label)
acc = categorical_accuracy(predictions, batch.label)
loss.backward()
optimizer.step()
epoch_loss += loss.item()
epoch_acc += acc.item()
return epoch_loss / len(iterator), epoch_acc / len(iterator)
12.构建评估函数
def evaluate(model, iterator, criterion):
epoch_loss = 0
epoch_acc = 0
model.eval()
with torch.no_grad():
for batch in iterator:
predictions = model(batch.text)
loss = criterion(predictions, batch.label)
acc = categorical_accuracy(predictions, batch.label)
epoch_loss += loss.item()
epoch_acc += acc.item()
return epoch_loss / len(iterator), epoch_acc / len(iterator)
13.构建计时模块
import time
def epoch_time(start_time, end_time):
elapsed_time = end_time - start_time
elapsed_mins = int(elapsed_time / 60)
elapsed_secs = int(elapsed_time - (elapsed_mins * 60))
return elapsed_mins, elapsed_secs
14. 正式训练
N_EPOCHS = 5
best_valid_loss = float('inf')
for epoch in range(N_EPOCHS):
start_time = time.time()
train_loss, train_acc = train(model, train_iterator, optimizer, criterion)
valid_loss, valid_acc = evaluate(model, valid_iterator, criterion)
end_time = time.time()
epoch_mins, epoch_secs = epoch_time(start_time, end_time)
if valid_loss < best_valid_loss:
best_valid_loss = valid_loss
torch.save(model.state_dict(), 'tut5-model.pt')
print(f'Epoch: {epoch+1:02} | Epoch Time: {epoch_mins}m {epoch_secs}s')
print(f'\tTrain Loss: {train_loss:.3f} | Train Acc: {train_acc*100:.2f}%')
print(f'\t Val. Loss: {valid_loss:.3f} | Val. Acc: {valid_acc*100:.2f}%')
15.评估模型
model.load_state_dict(torch.load('tut5-model.pt'))
test_loss, test_acc = evaluate(model, test_iterator, criterion)
print(f'Test Loss: {test_loss:.3f} | Test Acc: {test_acc*100:.2f}%')
Test Loss: 0.409 | Test Acc: 86.46%
16. 实际预测
import spacy
nlp = spacy.load('en')
def predict_class(model, sentence, min_len = 4):
model.eval()
tokenized = [tok.text for tok in nlp.tokenizer(sentence)]
if len(tokenized) < min_len:
tokenized += ['<pad>'] * (min_len - len(tokenized))
indexed = [TEXT.vocab.stoi[t] for t in tokenized]
tensor = torch.LongTensor(indexed).to(device)
tensor = tensor.unsqueeze(1)
preds = model(tensor)
max_preds = preds.argmax(dim = 1) # 跟之前的区别是,这里用的是argmax取预测值最大的那个元素作为分类结果
return max_preds.item()
pred_class = predict_class(model, "Who is Keyser Söze?")
print(f'Predicted class is: {pred_class} = {LABEL.vocab.itos[pred_class]}')