# evo_model.py
import torch
import torch.nn as nn
import math

class PositionalEncoding(nn.Module):
    def __init__(self, d_model, max_len=128):
        super().__init__()
        pe = torch.zeros(max_len, d_model)
        position = torch.arange(0, max_len).unsqueeze(1)
        div_term = torch.exp(torch.arange(0, d_model, 2) * (-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)  # (1, max_len, d_model)
        self.register_buffer('pe', pe)
        self.max_len = max_len

    def forward(self, x):
        seq_len = x.size(1)
        if seq_len > self.max_len:
            raise ValueError(f"Input length {seq_len} exceeds max_len {self.max_len}")
        return x + self.pe[:, :seq_len]

class EvoDecoderModel(nn.Module):
    def __init__(self, vocab_size, d_model=512, nhead=8, num_layers=6, dim_feedforward=2048, dropout=0.1):
        super().__init__()
        self.token_embed = nn.Embedding(vocab_size, d_model)
        self.pos_encoder = PositionalEncoding(d_model)
        decoder_layer = nn.TransformerDecoderLayer(d_model, nhead, dim_feedforward, dropout, batch_first=True)
        self.decoder = nn.TransformerDecoder(decoder_layer, num_layers=num_layers)
        self.lm_head = nn.Linear(d_model, vocab_size)

    def forward(self, input_ids):
        x = self.token_embed(input_ids)
        x = self.pos_encoder(x)
        seq_len = x.size(1)
        mask = torch.triu(torch.ones(seq_len, seq_len, device=x.device), diagonal=1).bool()
        x = self.decoder(x, x, tgt_mask=mask)
        return self.lm_head(x)