import math

import torch
from torch import nn
from utils.Arg import Arg

arg = Arg()

class PositionalEncoding(nn.Module):
    def __init__(self, d_model, dropout=0.1, max_len=5000):
        super(PositionalEncoding, self).__init__()
        self.dropout = nn.Dropout(p=dropout)

        pe = torch.zeros(max_len, d_model)
        position = torch.arange(0, max_len, dtype=torch.float).unsqueeze(1)
        div_term = torch.exp(torch.arange(0, d_model, 2).float() * (-math.log(10000.0) / d_model))
        pe[:, 0::2] = torch.sin(position * div_term)
        pe[:, 1::2] = torch.cos(position * div_term)
        pe = pe.unsqueeze(0).transpose(0, 1)
        self.register_buffer('pe', pe)

    def forward(self, x):
        x = x + self.pe[:x.size(0), :]
        return self.dropout(x)


# class TimeSeriesTransformer(nn.Module):
#     def __init__(self, input_dim, output_dim, d_model, nhead, num_layers, dropout=0.1):
#         super().__init__()
#         self.input_dim = input_dim
#         self.output_dim = output_dim
#         self.pos_enc = PositionalEncoding(d_model, dropout)
#         self.lstm = nn.LSTM(input_dim, d_model, batch_first=True) # 添加LSTM层
#         self.transformer = nn.Transformer(d_model, nhead, num_layers, dropout=dropout)
#         self.act = nn.GELU()  # 尝试使用GELU作为激活函数
#         self.hidden_dim = 128  # 定义一个隐藏层的维度
#         self.linear = nn.Linear(d_model, self.hidden_dim)
#         self.output_proj = nn.Linear(self.hidden_dim, output_dim)
#         self.layer_norm = nn.LayerNorm(d_model)
#
#     def forward(self, src, tgt):
#         src = self.pos_enc(src)  # 在输入数据通过LSTM和Transformer之前，先进行位置编码
#         src, _ = self.lstm(src)
#         tgt = self.pos_enc(tgt)  # 在目标数据通过LSTM和Transformer之前，先进行位置编码
#         tgt, _ = self.lstm(tgt)
#
#         output = self.transformer(src, tgt)
#         output = self.layer_norm(output)
#         output = self.act(self.linear(output))
#         output = self.output_proj(output)
#         output = output.squeeze(0)
#         return output

class TimeSeriesTransformer(nn.Module):
    def __init__(self, input_dim, output_dim, d_model, nhead, num_layers, dropout=0.1):
        super().__init__()
        self.input_dim = input_dim
        self.output_dim = output_dim
        self.pos_enc = PositionalEncoding(d_model, dropout)
        self.transformer = nn.Transformer(d_model, nhead, num_layers, dropout=dropout)
        self.act = nn.GELU()  # 尝试使用GELU作为激活函数
        self.hidden_dim = 64  # 降低隐藏层维度
        self.linear = nn.Linear(d_model, self.hidden_dim)
        self.output_proj = nn.Linear(self.hidden_dim, output_dim)
        self.layer_norm = nn.LayerNorm(d_model)
        self.sigmoid = nn.Tanh()

    def forward(self, src, tgt):
        src = self.pos_enc(src)  # 在输入数据通过Transformer之前，先进行位置编码
        tgt = self.pos_enc(tgt)  # 在目标数据通过Transformer之前，先进行位置编码

        output = self.transformer(src, tgt)
        output = self.layer_norm(output)
        output = self.act(self.linear(output))
        output = self.output_proj(output)
        output = self.sigmoid(output.squeeze(0))
        return output