加载句子转换模型时出错

Question

我正在尝试从SentenceTransformer加载转换器模型。下面是代码

# Now we create a SentenceTransformer model from scratch

word_emb = models.Transformer('paraphrase-mpnet-base-v2')
pooling = models.Pooling(word_emb.get_word_embedding_dimension())
model = SentenceTransformer(modules=[word_emb, pooling])

下面是错误

---------------------------------------------------------------------------
OSError                                   Traceback (most recent call last)
~\AppData\Local\Temp\ipykernel_2948\3254427654.py in <module>
      1 # Now we create a SentenceTransformer model from scratch
----> 2 word_emb = models.Transformer('paraphrase-mpnet-base-v2')
      3 pooling = models.Pooling(word_emb.get_word_embedding_dimension())
      4 model = SentenceTransformer(modules=[word_emb, pooling])

~\miniconda3\envs\atoti\lib\site-packages\sentence_transformers\models\Transformer.py in __init__(self, model_name_or_path, max_seq_length, model_args, cache_dir, tokenizer_args, do_lower_case, tokenizer_name_or_path)
     27 
     28         config = AutoConfig.from_pretrained(model_name_or_path, **model_args, cache_dir=cache_dir)
---> 29         self._load_model(model_name_or_path, config, cache_dir)
     30 
     31         self.tokenizer = AutoTokenizer.from_pretrained(tokenizer_name_or_path if tokenizer_name_or_path is not None else model_name_or_path, cache_dir=cache_dir, **tokenizer_args)

~\miniconda3\envs\atoti\lib\site-packages\sentence_transformers\models\Transformer.py in _load_model(self, model_name_or_path, config, cache_dir)
     47             self._load_t5_model(model_name_or_path, config, cache_dir)
     48         else:
---> 49             self.auto_model = AutoModel.from_pretrained(model_name_or_path, config=config, cache_dir=cache_dir)
     50 
     51     def _load_t5_model(self, model_name_or_path, config, cache_dir):

~\miniconda3\envs\atoti\lib\site-packages\transformers\models\auto\auto_factory.py in from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs)
    445         elif type(config) in cls._model_mapping.keys():
    446             model_class = _get_model_class(config, cls._model_mapping)
--> 447             return model_class.from_pretrained(pretrained_model_name_or_path, *model_args, config=config, **kwargs)
    448         raise ValueError(
    449             f"Unrecognized configuration class {config.__class__} for this kind of AutoModel: {cls.__name__}.\n"

~\miniconda3\envs\atoti\lib\site-packages\transformers\modeling_utils.py in from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs)
   1310                 elif os.path.join(pretrained_model_name_or_path, FLAX_WEIGHTS_NAME):
   1311                     raise EnvironmentError(
-> 1312                         f"Error no file named {WEIGHTS_NAME} found in directory {pretrained_model_name_or_path} but "
   1313                         "there is a file for Flax weights. Use `from_flax=True` to load this model from those "
   1314                         "weights."

OSError: Error no file named pytorch_model.bin found in directory paraphrase-mpnet-base-v2 but there is a file for Flax weights. Use `from_flax=True` to load this model from those weights.

我正在使用下面的版本

transformers==4.16.2
torch==1.11.0+cu113
torchaudio==0.11.0+cu113
torchvision==0.12.0+cu113
sentence-transformers==2.2.0
faiss-cpu==1.7.2
sentencepiece==0.1.96

我已经做了两个月了。突然间，它返回了一个错误。我也在用FAISS-CPU。

Answer 1

错误是告诉你“我找不到你要加载的模型的权重。”

基于错误跟踪，我猜您正在使用语句-变形金刚库中的models对象(如果我错了，请纠正我)。需要注意的一点是，句子转换器的预训练模型只有以下转述模型

• 释义-多语种-mpnet-base-v2
• 释义-艾伯特-小-v2
• 释义-多语种-MiniLM-L12-v2
• 翻译-MiniLM-L3-v2，因此，你想要载入的不是一个句子-变形金刚预先训练的模型。

这让我想到你正在尝试从你的本地机器加载一个模型。我建议你建立这样一个句子变形金刚模型：

from sentence_transformers import SentenceTransformer

model_path_or_name = "path/to/model" # A folder that contains model config files, including pytorch_model.bin
model = SentenceTransformer(model_path_or_name)

pytorch_model.bin文件也有可能是用另一个文件名下载的，就像所以线在这里中提到的那样。

如果这能解决你的问题请告诉我。干杯。