将一层的权重从一个 Huggingface BERT 模型复制到另一个

Question

我有一个预训练的模型，我可以这样加载：

from transformers import BertForSequenceClassification, AdamW, BertConfig, BertModel
model = BertForSequenceClassification.from_pretrained(
    "bert-base-uncased", # Use the 12-layer BERT model, with an uncased vocab.
    num_labels = 2, # The number of output labels--2 for binary classification.
                    # You can increase this for multi-class tasks.   
    output_attentions = False, # Whether the model returns attentions weights.
    output_hidden_states = False, # Whether the model returns all hidden-states.
)

我想创建一个具有相同架构和随机初始权重的新模型，除了嵌入层：

==== Embedding Layer ====

bert.embeddings.word_embeddings.weight                  (30522, 768)
bert.embeddings.position_embeddings.weight                (512, 768)
bert.embeddings.token_type_embeddings.weight                (2, 768)
bert.embeddings.LayerNorm.weight                              (768,)
bert.embeddings.LayerNorm.bias                                (768,)

看来我可以这样做来创建一个具有相同架构的新模型，但是所有权重都是随机的：

configuration   = model.config
untrained_model = BertForSequenceClassification(configuration)

那么如何将model 的嵌入层权重复制到新的untrained_model？

Answer 1

权重和偏差只是张量，您可以使用copy_ 简单地复制它们：

from transformers import BertForSequenceClassification, BertConfig
jetfire = BertForSequenceClassification.from_pretrained('bert-base-cased')
config = BertConfig.from_pretrained('bert-base-cased')

optimus = BertForSequenceClassification(config)

parts = ['bert.embeddings.word_embeddings.weight'
,'bert.embeddings.position_embeddings.weight'              
,'bert.embeddings.token_type_embeddings.weight'    
,'bert.embeddings.LayerNorm.weight'
,'bert.embeddings.LayerNorm.bias']

def joltElectrify (jetfire, optimus, parts):
  target = dict(optimus.named_parameters())
  source = dict(jetfire.named_parameters())

  for part in parts:
    target[part].data.copy_(source[part].data)  

joltElectrify(jetfire, optimus, parts)