import torch
import torch.nn as nn
import torch.nn.functional as F
from layers.Transformer_EncDec import Encoder, EncoderLayer
from layers.SelfAttention_Family import FlashAttention, AttentionLayer
from layers.Embed import DataEmbedding_inverted
import numpy as np


class Model(nn.Module):
    """
    Vanilla Transformer
    with O(L^2) complexity
    Paper link: https://proceedings.neurips.cc/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf
    """

    def __init__(self, configs):
        super(Model, self).__init__()
        self.seq_len = configs.seq_len
        self.pred_len = configs.pred_len
        self.output_attention = configs.output_attention
        # Embedding
        self.enc_embedding = DataEmbedding_inverted(configs.seq_len, configs.d_model, configs.embed, configs.freq,
                                                    configs.dropout)
        # Encoder-only architecture
        self.encoder = Encoder(
            [
                EncoderLayer(
                    AttentionLayer(
                        FlashAttention(False, configs.factor, attention_dropout=configs.dropout,
                                       output_attention=configs.output_attention), configs.d_model, configs.n_heads),
                    configs.d_model,
                    configs.d_ff,
                    dropout=configs.dropout,
                    activation=configs.activation
                ) for l in range(configs.e_layers)
            ],
            norm_layer=torch.nn.LayerNorm(configs.d_model)
        )
        self.projector = nn.Linear(configs.d_model, configs.pred_len, bias=True)



    def forecast(self, x_enc, x_mark_enc, x_dec, x_mark_dec):
        # Normalization from Non-stationary Transformer
        means = x_enc.mean(1, keepdim=True).detach()
        x_enc = x_enc - means
        stdev = torch.sqrt(torch.var(x_enc, dim=1, keepdim=True, unbiased=False) + 1e-5)
        x_enc /= stdev

        _, _, N = x_enc.shape

        # Embedding
        enc_out = self.enc_embedding(x_enc, x_mark_enc)
        enc_out, attns = self.encoder(enc_out, attn_mask=None)

        dec_out = self.projector(enc_out).permute(0, 2, 1)[:, :, :N]
        # De-Normalization from Non-stationary Transformer
        dec_out = dec_out * (stdev[:, 0, :].unsqueeze(1).repeat(1, self.pred_len, 1))
        dec_out = dec_out + (means[:, 0, :].unsqueeze(1).repeat(1, self.pred_len, 1))
        return dec_out

    def forward(self, x_enc, x_mark_enc, x_dec, x_mark_dec, mask=None):
        dec_out = self.forecast(x_enc, x_mark_enc, x_dec, x_mark_dec)
        return dec_out[:, -self.pred_len:, :]  # [B, L, D]