Positional Encoding

import torch
import math

class SinusoidalPE(torch.nn.Module):
    def __init__(self, d_model, max_len=5000):
        super().__init__()
        pe = torch.zeros(max_len, d_model)
        position = torch.arange(0, max_len).unsqueeze(1).float()
        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)  # Chiều chẵn
        pe[:, 1::2] = torch.cos(position * div_term)  # Chiều lẻ
        self.register_buffer('pe', pe.unsqueeze(0))    # (1, max_len, d_model)

    def forward(self, x):
        # x: (batch, seq_len, d_model)
        return x + self.pe[:, :x.size(1)]  # Cộng PE vào embedding

# Sử dụng
pe = SinusoidalPE(d_model=512)
x = word_embedding(tokens)  # (batch, seq_len, 512)
x = pe(x)  # embedding + positional encoding