在 Huggingface BERT 模型之上添加密集层

Question

我想在输出原始隐藏状态的裸 BERT 模型转换器之上添加一个密集层，然后微调生成的模型。具体来说，我使用的是this 基本模型。这是模型应该做的：

1. 对句子进行编码（句子的每个标记包含 768 个元素的向量）
2. 只保留第一个向量（与第一个标记相关）
3. 在该向量顶部添加一个密集层，以获得所需的转换

到目前为止，我已经成功编码了句子：

from sklearn.neural_network import MLPRegressor

import torch

from transformers import AutoModel, AutoTokenizer

# List of strings
sentences = [...]
# List of numbers
labels = [...]

tokenizer = AutoTokenizer.from_pretrained("dbmdz/bert-base-italian-xxl-cased")
model = AutoModel.from_pretrained("dbmdz/bert-base-italian-xxl-cased")

# 2D array, one line per sentence containing the embedding of the first token
encoded_sentences = torch.stack([model(**tokenizer(s, return_tensors='pt'))[0][0][0]
                                 for s in sentences]).detach().numpy()

regr = MLPRegressor()
regr.fit(encoded_sentences, labels)

通过这种方式，我可以通过输入编码句子来训练神经网络。然而，这种方法显然没有微调基础 BERT 模型。有谁能够帮我？如何构建一个可以完全微调的模型（可能在 pytorch 中或使用 Huggingface 库）？

Answer 1

如果要调整 BERT 模型本身，则需要修改模型的参数。为此，您很可能希望使用 PyTorch 进行工作。下面是一些粗略的伪代码来说明：

from torch.optim import SGD

model = ... # whatever model you are using
parameters = model.parameters() # or some more specific set of parameters
optimizer = SGD(parameters,lr=.01) # or whatever optimizer you want
optimizer.zero_grad() # boiler-platy pytorch function

input = ... # whatever the appropriate input for your task is
label = ... # whatever the appropriate label for your task is
loss = model(**input, label) # usuall loss is the first item returned
loss.backward() # calculates gradient
optim.step() # runs optimization algorithm

我省略了所有相关的细节，因为它们非常繁琐且特定于您的具体任务。 Huggingface 有一篇不错的文章，详细介绍了 here，当您使用任何 pytorch 东西时，您肯定会想参考一些 pytorch 文档。我强烈推荐pytorch blitz，然后再尝试使用它。

Answer 2

有两种方法：由于您希望为类似于分类的下游任务微调模型，您可以直接使用：

BertForSequenceClassification 类。对输出维度 768 的逻辑回归层进行微调。

或者，您可以定义一个自定义模块，该模块基于预训练的权重创建一个 bert 模型并在其上添加层。

from transformers import BertModel
class CustomBERTModel(nn.Module):
    def __init__(self):
          super(CustomBERTModel, self).__init__()
          self.bert = BertModel.from_pretrained("dbmdz/bert-base-italian-xxl-cased")
          ### New layers:
          self.linear1 = nn.Linear(768, 256)
          self.linear2 = nn.Linear(256, 3) ## 3 is the number of classes in this example

    def forward(self, ids, mask):
          sequence_output, pooled_output = self.bert(
               ids, 
               attention_mask=mask)

          # sequence_output has the following shape: (batch_size, sequence_length, 768)
          linear1_output = self.linear1(sequence_output[:,0,:].view(-1,768)) ## extract the 1st token's embeddings

          linear2_output = self.linear2(linear2_output)

          return linear2_output

tokenizer = AutoTokenizer.from_pretrained("dbmdz/bert-base-italian-xxl-cased")
model = CustomBERTModel() # You can pass the parameters if required to have more flexible model
model.to(torch.device("cpu")) ## can be gpu
criterion = nn.CrossEntropyLoss() ## If required define your own criterion
optimizer = torch.optim.Adam(filter(lambda p: p.requires_grad, model.parameters()))

for epoch in epochs:
    for batch in data_loader: ## If you have a DataLoader()  object to get the data.

        data = batch[0]
        targets = batch[1] ## assuming that data loader returns a tuple of data and its targets
        
        optimizer.zero_grad()   
        encoding = tokenizer.batch_encode_plus(data, return_tensors='pt', padding=True, truncation=True,max_length=50, add_special_tokens = True)
        outputs = model(input_ids, attention_mask=attention_mask)
        outputs = F.log_softmax(outputs, dim=1)
        input_ids = encoding['input_ids']
        attention_mask = encoding['attention_mask']
        loss = criterion(outputs, targets)
        loss.backward()
        optimizer.step()