当前位置：首页 > article >正文

20250226-代码笔记04-class CVRP_Encoder AND class EncoderLayer

article 2025/2/28 17:53:23

文章目录

前言
一、class CVRP_Encoder(nn.Module):__init__(self, **model_params)
- 函数功能
- 函数代码
二、class CVRP_Encoder(nn.Module):forward(self, depot_xy, node_xy_demand)
- 函数功能
- 函数代码
三、class EncoderLayer(nn.Module):__init__(self, **model_params)
- 函数功能
- 函数代码
四、class EncoderLayer(nn.Module):forward(self, input1)
- 函数功能
- 函数代码
附录
- 代码（全）

前言

class CVRP_Encoder
class EncoderLayer(nn.Module)

以上是CVRPModel.py的类

/home/tang/RL_exa/NCO_code-main/single_objective/LCH-Regret/Regret-POMO/CVRP/POMO/CVRPModel.py

一、class CVRP_Encoder(nn.Module):init(self, **model_params)

函数功能

该代码片段是 CVRP_Encoder 类的 init 构造函数。其主要功能是 ==初始化模型的参数 ==和构建神经网络的层，用于解决容量限制的车辆路径问题（CVRP）。

执行流程图链接
在这里插入图片描述

函数代码

    def __init__(self, **model_params):
        super().__init__()
        self.model_params = model_params
        embedding_dim = self.model_params['embedding_dim']
        encoder_layer_num = self.model_params['encoder_layer_num']

        self.embedding_depot = nn.Linear(2, embedding_dim)
        self.embedding_node = nn.Linear(3, embedding_dim)
        self.layers = nn.ModuleList([EncoderLayer(**model_params) for _ in range(encoder_layer_num)])

二、class CVRP_Encoder(nn.Module):forward(self, depot_xy, node_xy_demand)

函数功能

forward 方法是 CVRP_Encoder 类的一部分，主要功能是将仓库（depot）和客户节点（node）的位置和需求信息进行嵌入（embedding）并通过编码器进行处理。
它返回处理后的节点和仓库的嵌入表示，用于后续的模型推理。

执行流程图链接
在这里插入图片描述

函数代码

    def forward(self, depot_xy, node_xy_demand):
        # depot_xy.shape: (batch, 1, 2)
        # node_xy_demand.shape: (batch, problem, 3)

        embedded_depot = self.embedding_depot(depot_xy)
        # shape: (batch, 1, embedding)
        embedded_node = self.embedding_node(node_xy_demand)
        # shape: (batch, problem, embedding)

        out = torch.cat((embedded_depot, embedded_node), dim=1)
        # shape: (batch, problem+1, embedding)

        for layer in self.layers:
            out = layer(out)

        return out
        # shape: (batch, problem+1, embedding)

三、class EncoderLayer(nn.Module):init(self, **model_params)

函数功能

init 方法是 EncoderLayer 类的构造函数，主要功能是初始化编码器层的各个子组件。
设置多头注意力机制所需的权重矩阵和相关的正则化层、前馈网络等，以便在 forward 方法中执行实际的操作。

执行流程图链接
在这里插入图片描述

函数代码

    def __init__(self, **model_params):
        super().__init__()
        self.model_params = model_params
        embedding_dim = self.model_params['embedding_dim']
        head_num = self.model_params['head_num']
        qkv_dim = self.model_params['qkv_dim']

        self.Wq = nn.Linear(embedding_dim, head_num * qkv_dim, bias=False)
        self.Wk = nn.Linear(embedding_dim, head_num * qkv_dim, bias=False)
        self.Wv = nn.Linear(embedding_dim, head_num * qkv_dim, bias=False)
        self.multi_head_combine = nn.Linear(head_num * qkv_dim, embedding_dim)

        self.add_n_normalization_1 = AddAndInstanceNormalization(**model_params)
        self.feed_forward = FeedForward(**model_params)
        self.add_n_normalization_2 = AddAndInstanceNormalization(**model_params)

四、class EncoderLayer(nn.Module):forward(self, input1)

函数功能

forward 方法是 EncoderLayer 类中的前向传播函数，主要功能是执行 多头自注意力机制 和 前馈神经网络，并进行相应的 残差连接和归一化。
执行流程图链接
在这里插入图片描述

函数代码

    def forward(self, input1):
        # input1.shape: (batch, problem+1, embedding)
        head_num = self.model_params['head_num']

        q = reshape_by_heads(self.Wq(input1), head_num=head_num)
        k = reshape_by_heads(self.Wk(input1), head_num=head_num)
        v = reshape_by_heads(self.Wv(input1), head_num=head_num)
        # qkv shape: (batch, head_num, problem, qkv_dim)

        out_concat = multi_head_attention(q, k, v)
        # shape: (batch, problem, head_num*qkv_dim)

        multi_head_out = self.multi_head_combine(out_concat)
        # shape: (batch, problem, embedding)

        out1 = self.add_n_normalization_1(input1, multi_head_out)
        out2 = self.feed_forward(out1)
        out3 = self.add_n_normalization_2(out1, out2)

        return out3
        # shape: (batch, problem, embedding)

附录

代码（全）

########################################
# ENCODER
########################################

class CVRP_Encoder(nn.Module):
    def __init__(self, **model_params):
        super().__init__()
        self.model_params = model_params
        embedding_dim = self.model_params['embedding_dim']
        encoder_layer_num = self.model_params['encoder_layer_num']

        self.embedding_depot = nn.Linear(2, embedding_dim)
        self.embedding_node = nn.Linear(3, embedding_dim)
        self.layers = nn.ModuleList([EncoderLayer(**model_params) for _ in range(encoder_layer_num)])

    def forward(self, depot_xy, node_xy_demand):
        # depot_xy.shape: (batch, 1, 2)
        # node_xy_demand.shape: (batch, problem, 3)

        embedded_depot = self.embedding_depot(depot_xy)
        # shape: (batch, 1, embedding)
        embedded_node = self.embedding_node(node_xy_demand)
        # shape: (batch, problem, embedding)

        out = torch.cat((embedded_depot, embedded_node), dim=1)
        # shape: (batch, problem+1, embedding)

        for layer in self.layers:
            out = layer(out)

        return out
        # shape: (batch, problem+1, embedding)


class EncoderLayer(nn.Module):
    def __init__(self, **model_params):
        super().__init__()
        self.model_params = model_params
        embedding_dim = self.model_params['embedding_dim']
        head_num = self.model_params['head_num']
        qkv_dim = self.model_params['qkv_dim']

        self.Wq = nn.Linear(embedding_dim, head_num * qkv_dim, bias=False)
        self.Wk = nn.Linear(embedding_dim, head_num * qkv_dim, bias=False)
        self.Wv = nn.Linear(embedding_dim, head_num * qkv_dim, bias=False)
        self.multi_head_combine = nn.Linear(head_num * qkv_dim, embedding_dim)

        self.add_n_normalization_1 = AddAndInstanceNormalization(**model_params)
        self.feed_forward = FeedForward(**model_params)
        self.add_n_normalization_2 = AddAndInstanceNormalization(**model_params)

    def forward(self, input1):
        # input1.shape: (batch, problem+1, embedding) 364t
    
        head_num = self.model_params['head_num']

        q = reshape_by_heads(self.Wq(input1), head_num=head_num)
        k = reshape_by_heads(self.Wk(input1), head_num=head_num)
        v = reshape_by_heads(self.Wv(input1), head_num=head_num)
        # qkv shape: (batch, head_num, problem, qkv_dim)

        out_concat = multi_head_attention(q, k, v)
        # shape: (batch, problem, head_num*qkv_dim)

        multi_head_out = self.multi_head_combine(out_concat)
        # shape: (batch, problem, embedding)

        out1 = self.add_n_normalization_1(input1, multi_head_out)
        out2 = self.feed_forward(out1)
        out3 = self.add_n_normalization_2(out1, out2)

        return out3
        # shape: (batch, problem, embedding)