当前位置：首页 > article >正文

HuggingFace学习笔记--metrics和pipeline的使用

article 2024/11/17 22:41:42

1--metrics的使用

metrics 意为指标，通过 API 可以快速使用内置的评价指标。

代码：

from datasets import list_metrics, load_metric

if __name__ == "__main__":
    # 列出所有评价指标
    metrics_list = list_metrics()
    print(len(metrics_list))
    print(metrics_list)
    
    # 加载一个评价指标
    metric = load_metric('glue', 'mrpc') # glue和mrpc见https://zhuanlan.zhihu.com/p/522017847
    print(metric.inputs_description) 
    
    # 计算一个评价指标
    predictions = [0, 1, 0]
    references = [0, 1, 1]

    final_score = metric.compute(predictions=predictions, references=references)
    print(final_score)
    print("All done!")

通过 list_metrics 查看所有的评价指标，通过 load_metric 选取合适的评价指标；

2--pipeline的使用

使用 pipeline 可以快速使用预训练好的模型，可以直接进行相关的任务，或作为下游任务的预训练模型。

2-1--正负面文本分类任务

代码：

from transformers import pipeline

if __name__ == "__main__":
    # 正面和负面文本分类
    classifier = pipeline("sentiment-analysis")
    result = classifier("I hate you")[0]
    print(result)
    
    result = classifier("I love you")[0]
    print(result)

输出结果：

{'label': 'NEGATIVE', 'score': 0.9991129040718079}
{'label': 'POSITIVE', 'score': 0.9998656511306763}

2-2--阅读理解任务

代码：

from transformers import pipeline

if __name__ == "__main__":
    # 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)
    print(result)

    result = question_answerer(question = "What is a good example of a question answering dataset?", context=context)
    print(result)

输出结果：

{'score': 0.6034508347511292, 'start': 42, 'end': 103, 'answer': 'the task of extracting an answer from a text given a question'}
{'score': 0.4721057713031769, 'start': 165, 'end': 178, 'answer': 'SQuAD dataset'}

2-3--完型填空

代码：

from transformers import pipeline

if __name__ == "__main__":
    # 完形填空
    unmasker = pipeline("fill-mask")
    sentence = 'HuggingFace is creating a <mask> that the community uses to solve NLP tasks.'
    result = unmasker(sentence)
    print(result)

输出结果：

[
    {'score': 0.17927497625350952, 'token': 3944, 'token_str': ' tool', 'sequence': 'HuggingFace is creating a tool that the community uses to solve NLP tasks.'},             
    {'score': 0.11349403858184814, 'token': 7208, 'token_str': ' framework', 'sequence': 'HuggingFace is creating a framework that the community uses to solve NLP tasks.'}, 
    {'score': 0.05243556201457977, 'token': 5560, 'token_str': ' library', 'sequence': 'HuggingFace is creating a library that the community uses to solve NLP tasks.'}, 
    {'score': 0.03493537753820419, 'token': 8503, 'token_str': ' database', 'sequence': 'HuggingFace is creating a database that the community uses to solve NLP tasks.'}, 
    {'score': 0.02860264666378498, 'token': 17715, 'token_str': ' prototype', 'sequence': 'HuggingFace is creating a prototype that the community uses to solve NLP tasks.'}
]

查看全文

http://www.kler.cn/a/147389.html