如何使用Hugging Face Transformers Pipeline调用Cross-Encoder模型？

问题背景 #

Hugging Face Hub上有一系列来自sentence_transformers库的Cross-Encoder模型，例如cross-encoder/mmarco-mMiniLMv2-L12-H384-v1。常规调用方式有两种：

1. 通过sentence_transformers库的CrossEncoder类调用

# 使用sentence_transformers库
from sentence_transformers import CrossEncoder

model_name = 'cross-encoder/mmarco-mMiniLMv2-L12-H384-v1'
model = CrossEncoder(model_name)
scores = model.predict([
  ['How many people live in Berlin?', 'How many people live in Berlin?'], 
  ['Berlin has a population of 3,520,031 registered inhabitants in an area of 891.82 square kilometers.', 'New York City is famous for the Metropolitan Museum of Art.']
])
print(scores)

输出：

array([ 0.36782095, -4.2674575 ], dtype=float32)

2. 通过transformers库直接调用模型与分词器

# 使用transformers库直接调用
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

model = AutoModelForSequenceClassification.from_pretrained('cross-encoder/mmarco-mMiniLMv2-L12-H384-v1')
tokenizer = AutoTokenizer.from_pretrained('cross-encoder/mmarco-mMiniLMv2-L12-H384-v1')

features = tokenizer(
    ['How many people live in Berlin?', 'How many people live in Berlin?'], 
    ['Berlin has a population of 3,520,031 registered inhabitants in an area of 891.82 square kilometers.', 'New York City is famous for the Metropolitan Museum of Art.'],  
    padding=True, truncation=True, return_tensors="pt"
)

model.eval()
with torch.no_grad():
    scores = model(**features).logits
    print(scores)

输出：

tensor([[10.7615],
        [-8.1277]])

但直接传入已加载的模型和分词器到pipeline时会报错：

from transformers import AutoTokenizer, AutoModelForSequenceClassification
from transformers import pipeline
import torch

model = AutoModelForSequenceClassification.from_pretrained('cross-encoder/mmarco-mMiniLMv2-L12-H384-v1')
tokenizer = AutoTokenizer.from_pretrained('cross-encoder/mmarco-mMiniLMv2-L12-H384-v1')

pipe = pipeline(model=model, tokenizer=tokenizer)

报错信息：

RuntimeError: Inferring the task automatically requires to check the hub with a model_id defined as a `str`.

问题解答 #

1. 如何使用Hugging Face Transformers Pipeline调用Cross-Encoder模型？ #

报错核心原因是：传入已加载的模型对象而非字符串形式的model_id时，pipeline无法自动推断任务类型。解决方法是明确指定task为text-classification，同时注意Cross-Encoder的输入是文本对，调用时需用元组或列表形式传入成对文本。

方法一：直接传入model_id字符串（推荐，自动处理任务推断）

from transformers import pipeline

# 直接指定model_id，pipeline会自动加载模型、分词器并推断任务
pipe = pipeline("text-classification", model="cross-encoder/mmarco-mMiniLMv2-L12-H384-v1")

# 输入文本对（用元组或列表包裹）
results = pipe([
    ("How many people live in Berlin?", "How many people live in Berlin?"),
    ("Berlin has a population of 3,520,031 registered inhabitants in an area of 891.82 square kilometers.", "New York City is famous for the Metropolitan Museum of Art.")
])
print(results)

方法二：传入已加载的模型和分词器，指定task

from transformers import AutoTokenizer, AutoModelForSequenceClassification
from transformers import pipeline
import torch

model = AutoModelForSequenceClassification.from_pretrained('cross-encoder/mmarco-mMiniLMv2-L12-H384-v1')
tokenizer = AutoTokenizer.from_pretrained('cross-encoder/mmarco-mMiniLMv2-L12-H384-v1')

# 明确指定task为text-classification
pipe = pipeline("text-classification", model=model, tokenizer=tokenizer)

# 调用方式同上
results = pipe([
    ("How many people live in Berlin?", "How many people live in Berlin?"),
    ("Berlin has a population of 3,520,031 registered inhabitants in an area of 891.82 square kilometers.", "New York City is famous for the Metropolitan Museum of Art.")
])
print(results)

2. 若需要model_id，能否将其作为args或kwargs传入pipeline？ #

可以，有两种传入方式：

• 作为model参数的字符串值传入
• 在指定task后，直接将model_id作为位置参数传入

示例1：通过model参数传入

from transformers import pipeline

pipe = pipeline(task="text-classification", model="cross-encoder/mmarco-mMiniLMv2-L12-H384-v1")

示例2：作为位置参数传入（task之后）

from transformers import pipeline

pipe = pipeline("text-classification", "cross-encoder/mmarco-mMiniLMv2-L12-H384-v1")

如果需要单独指定分词器的model_id，也可以通过tokenizer参数传入：

from transformers import pipeline

pipe = pipeline(
    "text-classification",
    model="cross-encoder/mmarco-mMiniLMv2-L12-H384-v1",
    tokenizer="cross-encoder/mmarco-mMiniLMv2-L12-H384-v1"
)

内容的提问来源于stack exchange，提问作者alvas