File size: 1,407 Bytes
2bd9f2a 85560b0 2bd9f2a 85560b0 2bd9f2a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 |
import torch
from torch import nn
class ActivatorGatingUnit(nn.Module):
def __init__(self,dim, hidden_dim):
super().__init__()
self.proj_1 = nn.Linear(dim, hidden_dim)
self.proj_2 = nn.Linear(dim, hidden_dim)
self.proj_3 = nn.Linear(hidden_dim , dim)
self.gelu = nn.GELU()
self.norm = nn.LayerNorm(hidden_dim)
def forward(self, x):
u, v = x, x
u = self.proj_1(u)
u = self.gelu(u)
u = self.norm(u)
v = self.proj_2(v)
v = self.norm(v)
g = u * v
out = self.proj_3(g)
return out
class ActivatorBlock(nn.Module):
def __init__(self, d_model, d_ffn,dropout):
super().__init__()
self.norm = nn.LayerNorm(d_model)
self.actgu = ActivatorGatingUnit(d_model, d_ffn)
def forward(self, x):
residual = x
x = self.norm(x)
x = self.actgu(x)
x = x + residual
out = x
return out
class ACTIVATOR(nn.Module):
def __init__(self, d_model, d_ffn, num_layers,dropout):
super().__init__()
self.model = nn.Sequential(
*[ActivatorBlock(d_model,d_ffn,dropout) for _ in range(num_layers)]
)
def forward(self, x):
return self.model(x)
|