如何将BERT权重加载到HuggingFace DPRQuestionEncoder用于RAG微调

希望将BERT（或其他Transformer模型）的权重加载到DPRQuestionEncoder架构中，以便使用HuggingFace的save_pretrained方法，将保存后的模型接入RAG架构完成端到端微调。

运行的测试代码如下：

from transformers import DPRQuestionEncoder
model = DPRQuestionEncoder.from_pretrained('bert-base-uncased')

触发报错如下：

You are using a model of type bert to instantiate a model of type dpr. This is not supported for all configurations of models and can yield errors.

NotImplementedErrorTraceback (most recent call last)
<ipython-input-27-1f1b990b906b> in <module>
----> 1 model = DPRQuestionEncoder.from_pretrained(model_name)
      2 # https://github.com/huggingface/transformers/blob/41cd52a768a222a13da0c6aaae877a92fc6c783c/src/transformers/models/dpr/modeling_dpr.py#L520

/opt/conda/lib/python3.8/site-packages/transformers/modeling_utils.py in from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs)
   1211                     )
   1212 
-> 1213             model, missing_keys, unexpected_keys, error_msgs = cls._load_state_dict_into_model(
   1214                 model, state_dict, pretrained_model_name_or_path, _fast_init=_fast_init
   1215             )

/opt/conda/lib/python3.8/site-packages/transformers/modeling_utils.py in _load_state_dict_into_model(cls, model, state_dict, pretrained_model_name_or_path, _fast_init)
   1286             )
   1287             for module in unintialized_modules:
-> 1288                 model._init_weights(module)
   1289 
   1290         # copy state_dict so _load_from_state_dict can modify it

/opt/conda/lib/python3.8/site-packages/transformers/modeling_utils.py in _init_weights(self, module)
    515         Initialize the weights. This method should be overridden by derived class.
    516         """
--> 517         raise NotImplementedError(f"Make sure `_init_weigths` is implemented for {self.__class__}")
    518 
    519     def tie_weights(self):

NotImplementedError: Make sure `_init_weigths` is implemented for <class 'transformers.models.dpr.modeling_dpr.DPRQuestionEncoder'>

1. DPRQuestionEncoder和原生BERT的权重命名空间不一致，直接调用from_pretrained加载BERT权重会出现大量key不匹配，未匹配到的权重会触发模型初始化逻辑
2. DPRQuestionEncoder本身没有实现_init_weights方法，初始化未匹配模块时就会抛出对应的NotImplementedError

采用手动权重映射的方式加载，代码如下：

from transformers import BertModel, DPRQuestionEncoder, DPRConfig

# 加载原生BERT模型
bert_model = BertModel.from_pretrained('bert-base-uncased')
# 初始化DPRQuestionEncoder，配置参数和BERT对齐
dpr_config = DPRConfig(
    hidden_size=bert_model.config.hidden_size,
    num_hidden_layers=bert_model.config.num_hidden_layers,
    num_attention_heads=bert_model.config.num_attention_heads,
    intermediate_size=bert_model.config.intermediate_size,
    hidden_act=bert_model.config.hidden_act,
    layer_norm_eps=bert_model.config.layer_norm_eps
)
dpr_model = DPRQuestionEncoder(dpr_config)

# 权重映射：BERT主干权重对应DPR的question_encoder子模块
bert_state_dict = bert_model.state_dict()
dpr_state_dict = dpr_model.state_dict()

for k, v in bert_state_dict.items():
    dpr_key = f"question_encoder.{k}"
    if dpr_key in dpr_state_dict:
        dpr_state_dict[dpr_key] = v

# 加载映射后的权重
dpr_model.load_state_dict(dpr_state_dict)

# 保存后即可接入RAG架构使用
dpr_model.save_pretrained("./dpr-question-encoder-from-bert-base")

• 需保证DPR和BERT的模型配置参数完全对齐，否则会出现维度不匹配的加载错误
• 如果替换为其他Transformer模型作为底座，只需调整对应的权重前缀映射规则即可
• 加载完成后可先输入测试样本验证输出逻辑正常，再进行后续微调操作

内容的提问来源于stack exchange，提问作者albero