34 lines
1.2 KiB
Python
34 lines
1.2 KiB
Python
import torch
|
|
import torch.nn as nn
|
|
import torch.nn.functional as F
|
|
import math
|
|
|
|
class Embeddings(nn.Module):
|
|
def __init__(self, d_model, vocab):
|
|
super().__init__()
|
|
self.lut = nn.Embedding(vocab, d_model)
|
|
self.d_model = d_model
|
|
def forward(self, x):
|
|
return self.lut(x) * math.sqrt(self.d_model)
|
|
|
|
class PositionalEncoding(nn.Module):
|
|
def __init__(self, d_model, dropout, max_len=4000):
|
|
super().__init__()
|
|
self.dropout = nn.Dropout(dropout)
|
|
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)
|
|
self.register_buffer("pe", pe.unsqueeze(0))
|
|
def forward(self, x):
|
|
x = x + self.pe[:, :x.size(1)].detach()
|
|
return self.dropout(x)
|
|
|
|
class Generator(nn.Module):
|
|
def __init__(self, d_model, vocab):
|
|
super().__init__()
|
|
self.proj = nn.Linear(d_model, vocab)
|
|
def forward(self, x):
|
|
return F.log_softmax(self.proj(x), dim=-1)
|