如何截断 Huggingface 管道中的输入?

How to truncate input in the Huggingface pipeline?

我目前使用 huggingface 管道进行情绪分析,如下所示:

from transformers import pipeline
classifier = pipeline('sentiment-analysis', device=0)

问题是当我传递大于 512 个标记的文本时,它会崩溃并提示输入太长。有什么方法可以将 max_length 和截断参数从分词器直接传递到管道吗?

我的解决方法是:

从转换器导入 AutoTokenizer、AutoModelForSequenceClassification

model_name = "nlptown/bert-base-multilingual-uncased-sentiment"
model = AutoModelForSequenceClassification.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)
classifier = pipeline('sentiment-analysis', model=model, tokenizer=tokenizer, device=0)

然后当我调用分词器时:

pt_batch = tokenizer(text, padding=True, truncation=True, max_length=512, return_tensors="pt")

但是像这样直接调用管道会更好:

classifier(text, padding=True, truncation=True, max_length=512)

这种方式应该可行:

classifier(text, padding=True, truncation=True)

如果它不尝试将分词器加载为:

tokenizer = AutoTokenizer.from_pretrained(model_name, model_max_len=512)

您可以在推理时使用 tokenizer_kwargs :

model_pipline = pipeline("text-classification",model=model,tokenizer=tokenizer,device=0, return_all_scores=True)

tokenizer_kwargs = {'padding':True,'truncation':True,'max_length':512,'return_tensors':'pt'}

prediction = model_pipeline('sample text to predict',**tokenizer_kwargs)

有关详细信息,您可以查看此 link