无法下载保存HuggingFace模型jinaai/jina-reranker-v2-base-multilingual求助

问题场景 #

尝试从HuggingFace下载并保存jinaai/jina-reranker-v2-base-multilingual模型，原代码如下：

from transformers import AutoModelForSequenceClassification, AutoTokenizer, AutoModelForCausalLM

model_name='jinaai/jina-reranker-v2-base-multilingual'

tokenizer = AutoTokenizer.from_pretrained(model_name)
tokenizer.save_pretrained(model_name)

model = AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=True)
model.save_pretrained(f"model/{model_name}")

报错信息 #

运行后触发以下错误：

ValueError: Unrecognized configuration class <class 'transformers_modules.jina-reranker-v2-base-multilingual.configuration_xlm_roberta.XLMRobertaFlashConfig'> for this kind of AutoModel: AutoModelForCausalLM.
Model type should be one of BartConfig, BertConfig, BertGenerationConfig, BigBirdConfig, BigBirdPegasusConfig.......

项目文件夹结构 #

root/aimodel.py (代码片段所在文件)
root/jinai/jina-reranker-v2-base-multilingual/ < 包含jinaai/jina-reranker-v2-base-multilingual仓库的所有文件 >

解决方案 #

报错核心原因是模型加载类使用错误：jina-reranker-v2-base-multilingual是面向序列分类任务的模型，而非因果语言模型，不能用AutoModelForCausalLM加载，应匹配使用AutoModelForSequenceClassification。

修正后的代码如下：

from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = 'jinaai/jina-reranker-v2-base-multilingual'

# 加载并保存tokenizer
tokenizer = AutoTokenizer.from_pretrained(model_name)
tokenizer.save_pretrained(model_name)

# 使用正确的模型类加载并保存模型
model = AutoModelForSequenceClassification.from_pretrained(model_name, trust_remote_code=True)
model.save_pretrained(f"model/{model_name}")

如果本地已存在root/jinai/jina-reranker-v2-base-multilingual/文件夹，也可以直接指定本地路径加载，避免重复下载：

local_model_path = "./jinai/jina-reranker-v2-base-multilingual"

tokenizer = AutoTokenizer.from_pretrained(local_model_path)
model = AutoModelForSequenceClassification.from_pretrained(local_model_path, trust_remote_code=True)

内容的提问来源于stack exchange，提问作者Abhra Sarkar