HuggingFace - 简明教程

HuggingFace 官网：https://huggingface.co HuggingFace 它是一个的开源社区 NLP 模型、数据集等方便工具。

提供的主要模型：

1、自回归 GPT2 Transformer-XL XLNet

2、自编码 BERT ALBERT RoBERTa ELECTRA

3、Seg2Seq BART Pegasus T5

主要用于这里 bert-base-chinese，即bert中文模型。

基于以下安装 python 3.6，pytorch 1.10

# 1、安装 transformers (HuggingFace 提供的模型) ## 方式一：pip 安装 $ pip install transformers ## 方式二：conda 安装 $ conda install -c huggingface transformers  # 2、安装 datasets ## 方式一：pip 安装 $ pip install datasets ## 方式二：conda 安装 $ conda install -c huggingface -c conda-forge datasets   

 
from datasets import load_dataset

# 加载数据
dataset = load_dataset(path='seamew/ChnSentiCorp', split='train')
dataset
''' Using custom data configuration default Reusing dataset chn_senti_corp (/Users/lee/.cache/huggingface/datasets/seamew___chn_senti_corp/default/0.0.0/1f242195a37831906957a11a2985a4329167e60657c07dc95ebe266c03fdfb85) Dataset({ features: ['text', 'label'], num_rows: 9600 }) '''

 
# 查看一个数据
dataset[0]
''' {'text': '选择珠江花园的原因就是方便，有电动扶梯直接到达海边，周围餐馆、食廊、商场、超市、摊位一应俱全。酒店装修一般，但还算整洁。 泳池在大堂的屋顶，因此很小，不过女儿倒是喜欢。 包的早餐是西式的，还算丰富。 服务吗，一般', 'label': 1} '''

 
# sort

# 未排序的label是乱序的
print(dataset['label'][:10])

# 排序之后label有序了
sorted_dataset = dataset.sort('label')
print(sorted_dataset['label'][:10])
print(sorted_dataset['label'][-10:])

 
# shuffle

# 打乱顺序
shuffled_dataset = sorted_dataset.shuffle(seed=42)
shuffled_dataset['label'][:10]

 
# select
dataset.select([0, 10, 20, 30, 40, 50])

# filter
def f(data):
    return data['text'].startswith('选择')

start_with_ar = dataset.filter(f)

len(start_with_ar), start_with_ar['text']

# train_test_split, 切分训练集和测试集
dataset.train_test_split(test_size=0.1)

# shard 把数据切分到4个桶中,均匀分配
dataset.shard(num_shards=4, index=0)

# 列操作和类型转换
# rename_column 列重命名
dataset.rename_column('text', 'textA')

# remove_columns 列移除
dataset.remove_columns(['text'])

# map 把每一个数据输入到 lamda 函数，操作返回
def f(data):
    data['text'] = 'My sentence: ' + data['text']
    return data

datatset_map = dataset.map(f)
datatset_map['text'][:5]
 
# set_format
dataset.set_format(type='torch', columns=['label'])
dataset[0]# {'label': tensor(1)}

# 保存和加载
from datasets import load_from_disk

dataset = load_dataset(path='seamew/ChnSentiCorp', split='train')

# 保存和加载
dataset.save_to_disk("./")
dataset = load_from_disk("./")
dataset

# 导出为其他格式
# dataset.to_csv('./datasets.csv')
# dataset.to_json('./datasets.json')

from datasets import list_metrics

# 列出评价指标
metrics_list = list_metrics()
len(metrics_list), metrics_list
''' (34, ['accuracy', 'bertscore', 'bleu', 'bleurt', 'cer', 'chrf', 'code_eval', 'comet', 'competition_math', 'coval', 'cuad', 'f1', 'gleu', 'glue', 'google_bleu', '''
 
#加载一个评价指标
metric = load_metric('glue', 'mrpc')

print(metric.inputs_description)  # 打印使用文档 


#计算一个评价指标
predictions = [0, 1, 0]
references = [0, 1, 1]

final_score = metric.compute(predictions=predictions, references=references)
final_score  # {'accuracy': 0.6666666666666666, 'f1': 0.6666666666666666}

实用价值不高

from transformers import pipeline

# 文本分类
classifier = pipeline("sentiment-analysis")

result = classifier("I hate you")[0] # {'label': 'NEGATIVE', 'score': 0.9991129040718079} 

result = classifier("I love you")[0] # {'label': 'POSITIVE', 'score': 0.9998656511306763} 
 
# 阅读理解
question_answerer = pipeline("question-answering")

context = r""" Extractive Question Answering is the task of extracting an answer from a text given a question. An example of a question answering dataset is the SQuAD dataset, which is entirely based on that task. If you would like to fine-tune a model on a SQuAD task, you may leverage the examples/pytorch/question-answering/run_squad.py script. """

result = question_answerer(question="What is extractive question answering?", context=context) # {'score': 0.6177279353141785, 'start': 34, 'end': 95, 'answer': 'the task of extracting an answer from a text given a question'} 

# 答案需要在文本当中
result = question_answerer( question="What is a good example of a question answering dataset?", context=context) # {'score': 0.5152313113212585, 'start': 148, 'end': 161, 'answer': 'SQuAD dataset'}
  
# 完形填空
unmasker = pipeline("fill-mask")
from pprint import pprint
sentence = 'HuggingFace is creating a <mask> that the community uses to solve NLP tasks.'
unmasker(sentence)

 
# 文本生成
text_generator = pipeline("text-generation")
text_generator("As far as I am concerned, I will", max_length=50, do_sample=False)
from transformers import pipeline

# 命名实体识别
ner_pipe = pipeline("ner")

sequence = """Hugging Face Inc. is a company based in New York City. Its headquarters are in DUMBO, therefore very close to the Manhattan Bridge which is visible from the window."""

for entity in ner_pipe(sequence):
    print(entity)

 
# 文本总结
summarizer = pipeline("summarization")

ARTICLE = """ New York (CNN)When Liana Barrientos was 23 years old, she got married in Westchester County, New York. A year later, she got married again in Westchester County, but to a different man and without divorcing her first husband. Only 18 days after that marriage, she got hitched yet again. Then, Barrientos declared "I do" five more times, sometimes only within two weeks of each other. In 2010, she married once more, this time in the Bronx. In an application for a marriage license, she stated it was her "first and only" marriage. Barrientos, now 39, is facing two criminal counts of "offering a false instrument for filing in the first degree," referring to her false statements on the 2010 marriage license application, according to court documents. Prosecutors said the marriages were part of an immigration scam. On Friday, she pleaded not guilty at State Supreme Court in the Bronx, according to her attorney, Christopher Wright, who declined to comment further. After leaving court, Barrientos was arrested and charged with theft of service and criminal trespass for allegedly sneaking into the New York subway through an emergency exit, said Detective Annette Markowski, a police spokeswoman. In total, Barrientos has been married 10 times, with nine of her marriages occurring between 1999 and 2002. All occurred either in Westchester County, Long Island, New Jersey or the Bronx. She is believed to still be married to four men, and at one time, she was married to eight men at once, prosecutors say. Prosecutors said the immigration scam involved some of her husbands, who filed for permanent residence status shortly after the marriages. Any divorces happened only after such filings were approved. It was unclear whether any of the men will be prosecuted. The case was referred to the Bronx District Attorney\'s Office by Immigration and Customs Enforcement and the Department of Homeland Security\'s Investigation Division. Seven of the men are from so-called "red-flagged" countries, including Egypt, Turkey, Georgia, Pakistan and Mali. Her eighth husband, Rashid Rajput, was deported in 2006 to his native Pakistan after an investigation by the Joint Terrorism Task Force. If convicted, Barrientos faces up to four years in prison. Her next court appearance is scheduled for May 18. """

summarizer(ARTICLE, max_length=130, min_length=30, do_sample=False)


# 翻译
translator = pipeline("translation_en_to_de")
sentence = "Hugging Face is a technology company based in New York and Paris"
translator(sentence, max_length=40)

import torch
from datasets import load_dataset

# 定义数据集
class Dataset(torch.utils.data.Dataset):
    def __init__(self, split):
        self.dataset = load_dataset(path='seamew/ChnSentiCorp', split=split)

    def __len__(self):
        return len(self.dataset)

    def __getitem__(self, i):
        text = self.dataset[i]['text']
        label = self.dataset[i]['label']

        return text, label


dataset = Dataset('train')

len(dataset), dataset[0]


from transformers import BertTokenizer # 加载分词工具（需要和与预训练模型相匹配）

# 加载字典和分词工具
token = BertTokenizer.from_pretrained('bert-base-chinese')
# PreTrainedTokenizer(name_or_path='bert-base-chinese', vocab_size=21128, model_max_len=512, is_fast=False, padding_side='right', special_tokens={'unk_token': '[UNK]', 'sep_token': '[SEP]', 'pad_token': '[PAD]', 'cls_token': '[CLS]', 'mask_token': '[MASK]'})


def collate_fn(data):
    sents = [i[0] for i in data]
    labels = [i[1] for i in data]

    # 编码
    data = token.batch_encode_plus(batch_text_or_text_pairs=sents,
                                   truncation=True,
                                   padding='max_length',
                                   max_length=500,
                                   return_tensors='pt',
                                   return_length=True)

    # input_ids:编码之后的数字
    # attention_mask:是补零的位置是0,其他位置是1
    input_ids = data['input_ids']
    attention_mask = data['attention_mask']
    token_type_ids = data['token_type_ids']
    labels = torch.LongTensor(labels)

    # print(data['length'], data['length'].max())

    return input_ids, attention_mask, token_type_ids, labels


# 数据加载器
loader = torch.utils.data.DataLoader(dataset=dataset,
                                     batch_size=16,
                                     collate_fn=collate_fn,
                                     shuffle=True,
                                     drop_last=True)

for i, (input_ids, attention_mask, token_type_ids,
        labels) in enumerate(loader):
    break

print(len(loader)) # 600
input_ids.shape, attention_mask.shape, token_type_ids.shape, labels
''' (torch.Size([16, 500]), torch.Size([16, 500]), torch.Size([16, 500]), tensor([1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 0, 1])) '''


from transformers import BertModel

# 加载预训练模型
pretrained = BertModel.from_pretrained('bert-base-chinese')

# 不训练,不需要计算梯度
for param in pretrained.parameters():
    param.requires_grad_(False)

# 模型试算
out = pretrained(input_ids=input_ids,
           attention_mask=attention_mask,
           token_type_ids=token_type_ids)

out.last_hidden_state.shape # torch.Size([16, 500, 768]) 16是一个 batch-size，对应数据中的16句话；指定每一句话编码成 500个词的长度， 768 是词编码的维度；


# 定义下游任务模型
class Model(torch.nn.Module):
    def __init__(self):
        super().__init__()
        self.fc = torch.nn.Linear(768, 2) # 此处做的是2分类，想做n分类，改成n即可。

    def forward(self, input_ids, attention_mask, token_type_ids):
        with torch.no_grad():
            out = pretrained(input_ids=input_ids,
                       attention_mask=attention_mask,
                       token_type_ids=token_type_ids)

        out = self.fc(out.last_hidden_state[:, 0]) # 取第0个特征（这个和bert 的设计思路有关）
        out = out.softmax(dim=1)
        return out

model = Model()

model(input_ids=input_ids,
      attention_mask=attention_mask,
      token_type_ids=token_type_ids).shape


from transformers import AdamW

# 训练
optimizer = AdamW(model.parameters(), lr=5e-4)
criterion = torch.nn.CrossEntropyLoss()

model.train()
for i, (input_ids, attention_mask