Source code for pyabsa.tasks.AspectPolarityClassification.models.__plm__.memnet_bert

# -*- coding: utf-8 -*-
# file: memnet.py
# author: songyouwei <youwei0314@gmail.com>
# Copyright (C) 2018. All Rights Reserved.

import torch
import torch.nn as nn

from pyabsa.networks.attention import Attention
from pyabsa.networks.squeeze_embedding import SqueezeEmbedding




class MemNet_BERT(nn.Module):


    inputs = ["context_indices", "aspect_indices"]




    def locationed_memory(self, memory, memory_len):
        # here we just simply calculate the location vector in Model2's manner
        batch_size = memory.shape[0]
        seq_len = memory.shape[1]
        memory_len = memory_len.cpu().numpy()
        weight = [[] for i in range(batch_size)]
        for i in range(batch_size):
            for idx in range(memory_len[i]):
                weight[i].append(1 - float(idx + 1) / memory_len[i])
            for idx in range(memory_len[i], seq_len):
                weight[i].append(1)
        weight = torch.tensor(weight).to(self.config.device)
        memory = weight.unsqueeze(2) * memory
        return memory


    def __init__(self, bert, config):
        super(MemNet_BERT, self).__init__()
        self.config = config
        self.embed = bert
        self.squeeze_embedding = SqueezeEmbedding(batch_first=True)
        self.attention = Attention(config.embed_dim, score_function="mlp")
        self.x_linear = nn.Linear(config.embed_dim, config.embed_dim)
        self.dense = nn.Linear(config.embed_dim, config.output_dim)



    def forward(self, inputs):
        text_raw_without_aspect_indices, aspect_indices = (
            inputs["context_indices"],
            inputs["aspect_indices"],
        )
        memory_len = torch.sum(text_raw_without_aspect_indices != 0, dim=-1)
        aspect_len = torch.sum(aspect_indices != 0, dim=-1)
        nonzeros_aspect = torch.tensor(aspect_len, dtype=torch.float).to(
            self.config.device
        )

        memory = self.embed(text_raw_without_aspect_indices)["last_hidden_state"]
        memory = self.squeeze_embedding(memory, memory_len)
        # memory = self.locationed_memory(memory, memory_len)
        aspect = self.embed(aspect_indices)["last_hidden_state"]
        aspect = torch.sum(aspect, dim=1)
        aspect = torch.div(aspect, nonzeros_aspect.view(nonzeros_aspect.size(0), 1))
        x = aspect.unsqueeze(dim=1)
        if "hops" not in self.config.args:
            self.config.hops = 3
        for _ in range(self.config.hops):
            x = self.x_linear(x)
            out_at, _ = self.attention(memory, x)
            x = out_at + x
        x = x.view(x.size(0), -1)
        out = self.dense(x)
        return {"logits": out}