小黑醒着鼻涕:搭建了初步LE-Roberta主框架，还待进一步测试

from transformers.modeling_roberta import RobertaPreTrainedModel, RobertaEmbeddings, RobertaPooler, \
    ROBERTA_INPUTS_DOCSTRING, _TOKENIZER_FOR_DOC, BaseModelOutputWithPooling, _CONFIG_FOR_DOC, RobertaLayer
from transformers import RobertaConfig,AutoConfig
import torch
import torch.nn as nn
import torch.nn.functional as F
from transformers.file_utils import (
    add_code_sample_docstrings,
    add_start_docstrings,
    add_start_docstrings_to_callable,
    replace_return_docstrings,
)

class WordEmbeddingAdapter(nn.Module):
    
    def __init__(self,config):
        super(WordEmbeddingAdapter,self).__init__()
        self.dropout = nn.Dropout(config.hidden_dropout_prob)
        self.tanh = nn.Tanh()
        
        self.linear1 = nn.Linear(config.word_embed_dim,config.hidden_size)
        self.linear2 = nn.Linear(config.hidden_size,config.hidden_size)
        
        attn_W = torch.zeros(config.hidden_size,config.hidden_size)
        self.attn_W = nn.Parameter(attn_W)
        self.attn_W.data.normal_(mean = 0.0,std = config.initializer_range)
        self.layer_norm = nn.LayerNorm(config.hidden_size,eps = config.layer_norm_eps)
        
    def forward(self,layer_output,word_embeddings,word_mask):
        # layer_output:[batch_size,max_len,bert_dim]
        # word_embeddings:[batch_size,max_len,num_words,word_dim]
        # word_mask:[batch_size,max_len,num_words]
        
        # 对齐词向量与字符向量
        # word_outputs:[batch_size,max_len,num_words,bert_dim]
        word_outputs = self.linear1(word_embeddings)
        word_outputs = self.tanh(word_outputs)
        # word_outputs:[batch_size,max_len,num_words,bert_dim]
        word_outputs = self.linear2(word_outputs)
        word_outputs = self.dropout(word_outputs)
        # 计算每个字符向量,与其对应的所有词向量的注意力权重,然后加权求和。采用双线性映射计算注意力权重
        # scores:[batch_size,max_len,1,bert_dim]
        scores = torch.matmul(layer_output.unsqueeze(2),self.attn_W)
        # scores:[batch_size,max_len,1,num_words]
        scores = torch.matmul(scores,torch.transpose(word_outputs,2,3))
        # scores:[batch_size,max_len,num_words]
        scores = scores.squeeze(2)
        scores.masked_fill_(word_mask,-1e9)
        scores = F.softmax(scores,dim = -1)
        # attn:[batch_size,max_len,num_words,1]
        attn = scores.unsqueeze(-1)
        # weighted_word_embedding:[batch_size,max_len,bert_dim]
        weighted_word_embedding = torch.sum(word_outputs * attn,dim = 2)
        # layer_output:[batch_size,max_len,bert_dim]
        layer_output = layer_output + weighted_word_embedding
        layer_output = self.dropout(layer_output)
        layer_output = self.layer_norm(layer_output)
        
        return layer_output

class RobertaEncoder(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.config = config
        self.layer = nn.ModuleList([RobertaLayer(config) for _ in range(config.num_hidden_layers)])
        self.word_embedding_adapter = WordEmbeddingAdapter(config)

    def forward(
            self,
            hidden_states,
            word_embeddings,
            word_mask,
            attention_mask=None,
            head_mask=None,
            encoder_hidden_states=None,
            encoder_attention_mask=None,
            output_attentions=False,
            output_hidden_states=False,
            return_dict=False,
    ):
        all_hidden_states = () if output_hidden_states else None
        all_attentions = () if output_attentions else None
        for i, layer_module in enumerate(self.layer):
            if output_hidden_states:
                all_hidden_states = all_hidden_states + (hidden_states,)

            layer_head_mask = head_mask[i] if head_mask is not None else None

            if getattr(self.config, "gradient_checkpointing", False):

                def create_custom_forward(module):
                    def custom_forward(*inputs):
                        return module(*inputs, output_attentions)

                    return custom_forward

                layer_outputs = torch.utils.checkpoint.checkpoint(
                    create_custom_forward(layer_module),
                    hidden_states,
                    attention_mask,
                    layer_head_mask,
                    encoder_hidden_states,
                    encoder_attention_mask,
                )
            else:
                layer_outputs = layer_module(
                    hidden_states,
                    attention_mask,
                    layer_head_mask,
                    encoder_hidden_states,
                    encoder_attention_mask,
                    output_attentions,
                )
            hidden_states = layer_outputs[0]
            if output_attentions:
                all_attentions = all_attentions + (layer_outputs[1],)
            if i == self.config.add_layer:
                # hidden_states:[batch_size,max_len,bert_dim]
                hidden_states = self.word_embedding_adapter(hidden_states,word_embeddings,word_mask.byte())
        if output_hidden_states:
            all_hidden_states = all_hidden_states + (hidden_states,)

        if not return_dict:
            return tuple(v for v in [hidden_states, all_hidden_states, all_attentions] if v is not None)
        return BaseModelOutput(
            last_hidden_state=hidden_states, hidden_states=all_hidden_states, attentions=all_attentions
        )



class LEGuwenRobertaModel(RobertaPreTrainedModel):
    """

    The model can behave as an encoder (with only self-attention) as well
    as a decoder, in which case a layer of cross-attention is added between
    the self-attention layers, following the architecture described in `Attention is all you need`_ by Ashish Vaswani,
    Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin.

    To behave as an decoder the model needs to be initialized with the
    :obj:`is_decoder` argument of the configuration set to :obj:`True`.
    To be used in a Seq2Seq model, the model needs to initialized with both :obj:`is_decoder`
    argument and :obj:`add_cross_attention` set to :obj:`True`; an
    :obj:`encoder_hidden_states` is then expected as an input to the forward pass.

    .. _`Attention is all you need`:
        https://arxiv.org/abs/1706.03762

    """

    authorized_missing_keys = [r"position_ids"]

    # Copied from transformers.modeling_bert.BertModel.__init__ with Bert->Roberta
    def __init__(self, config, add_pooling_layer=True):
        super().__init__(config)
        self.config = config

        self.embeddings = RobertaEmbeddings(config)
        self.encoder = RobertaEncoder(config)

        self.pooler = RobertaPooler(config) if add_pooling_layer else None

        self.init_weights()

    def get_input_embeddings(self):
        return self.embeddings.word_embeddings

    def set_input_embeddings(self, value):
        self.embeddings.word_embeddings = value

    def _prune_heads(self, heads_to_prune):
        """Prunes heads of the model.
        heads_to_prune: dict of {layer_num: list of heads to prune in this layer}
        See base class PreTrainedModel
        """
        for layer, heads in heads_to_prune.items():
            self.encoder.layer[layer].attention.prune_heads(heads)

    @add_start_docstrings_to_callable(ROBERTA_INPUTS_DOCSTRING.format("(batch_size, sequence_length)"))
    @add_code_sample_docstrings(
        tokenizer_class=_TOKENIZER_FOR_DOC,
        checkpoint="roberta-base",
        output_type=BaseModelOutputWithPooling,
        config_class=_CONFIG_FOR_DOC,
    )
    # Copied from transformers.modeling_bert.BertModel.forward
    def forward(
            self,
            input_ids=None,
            attention_mask=None,
            token_type_ids=None,
            word_embeddings = None,
            word_mask = None,
            position_ids=None,
            head_mask=None,
            inputs_embeds=None,
            encoder_hidden_states=None,
            encoder_attention_mask=None,
            output_attentions=None,
            output_hidden_states=None,
            return_dict=None,
    ):
        r"""
        encoder_hidden_states  (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`):
            Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention
            if the model is configured as a decoder.
        encoder_attention_mask (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
            Mask to avoid performing attention on the padding token indices of the encoder input. This mask
            is used in the cross-attention if the model is configured as a decoder.
            Mask values selected in ``[0, 1]``:
            ``1`` for tokens that are NOT MASKED, ``0`` for MASKED tokens.
        """
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        if input_ids is not None and inputs_embeds is not None:
            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
        elif input_ids is not None:
            input_shape = input_ids.size()
        elif inputs_embeds is not None:
            input_shape = inputs_embeds.size()[:-1]
        else:
            raise ValueError("You have to specify either input_ids or inputs_embeds")

        device = input_ids.device if input_ids is not None else inputs_embeds.device

        if attention_mask is None:
            attention_mask = torch.ones(input_shape, device=device)
        if token_type_ids is None:
            token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device)

        # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
        # ourselves in which case we just need to make it broadcastable to all heads.
        extended_attention_mask: torch.Tensor = self.get_extended_attention_mask(attention_mask, input_shape, device)

        # If a 2D or 3D attention mask is provided for the cross-attention
        # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
        if self.config.is_decoder and encoder_hidden_states is not None:
            encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size()
            encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
            if encoder_attention_mask is None:
                encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device)
            encoder_extended_attention_mask = self.invert_attention_mask(encoder_attention_mask)
        else:
            encoder_extended_attention_mask = None

        # Prepare head mask if needed
        # 1.0 in head_mask indicate we keep the head
        # attention_probs has shape bsz x n_heads x N x N
        # input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
        # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
        head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)

        embedding_output = self.embeddings(
            input_ids=input_ids, position_ids=position_ids, token_type_ids=token_type_ids, inputs_embeds=inputs_embeds
        )
        encoder_outputs = self.encoder(
            embedding_output,
            word_embeddings = word_embeddings,
            word_mask = word_mask,
            attention_mask=extended_attention_mask,
            head_mask=head_mask,
            encoder_hidden_states=encoder_hidden_states,
            encoder_attention_mask=encoder_extended_attention_mask,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        sequence_output = encoder_outputs[0]
        pooled_output = self.pooler(sequence_output) if self.pooler is not None else None

        if not return_dict:
            return (sequence_output, pooled_output) + encoder_outputs[1:]

        return BaseModelOutputWithPooling(
            last_hidden_state=sequence_output,
            pooler_output=pooled_output,
            hidden_states=encoder_outputs.hidden_states,
            attentions=encoder_outputs.attentions,
        )
pretrain_model_path = '../guwenbert-base/'
config = AutoConfig.from_pretrained(pretrain_model_path)
config.word_embed_dim = 200
config.add_layer = 0
model = LEGuwenRobertaModel(config)
num_words = 3
max_len = 10
bert_dim = 768
batch_size = 4
input_ids = torch.randint(0,100,[batch_size,max_len])
word_embeddings = torch.randn([batch_size,max_len,num_words,config.word_embed_dim])
word_mask = torch.ones([batch_size,max_len,num_words]).long()
attention_mask = torch.ones([batch_size,max_len]).byte()
token_type_ids = torch.zeros([batch_size,max_len]).long()


outputs = model(
                input_ids = input_ids,    
                attention_mask = attention_mask,
                token_type_ids = token_type_ids,
                word_embeddings = word_embeddings,
                word_mask = word_mask,
                position_ids = None,
                head_mask = None,
                inputs_embeds = None,
                encoder_hidden_states = None,
                encoder_attention_mask = None,
                output_attentions = None,
                output_hidden_states = None,
                return_dict = None
            )
print('outputs[0].shape',outputs[0].shape)
print('outputs[1].shape',outputs[1].shape)

输出:

outputs[0].shape torch.Size([4, 10, 768])
outputs[1].shape torch.Size([4, 768])