File size: 4,376 Bytes
9c6594c |
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 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 |
""" MLP module w/ dropout and configurable activation layer
Hacked together by / Copyright 2020 Ross Wightman
"""
from torch import nn as nn
from .helpers import to_2tuple
class Mlp(nn.Module):
""" MLP as used in Vision Transformer, MLP-Mixer and related networks
"""
def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, bias=True, drop=0.):
super().__init__()
out_features = out_features or in_features
hidden_features = hidden_features or in_features
bias = to_2tuple(bias)
drop_probs = to_2tuple(drop)
self.fc1 = nn.Linear(in_features, hidden_features, bias=bias[0])
self.act = act_layer()
self.drop1 = nn.Dropout(drop_probs[0])
self.fc2 = nn.Linear(hidden_features, out_features, bias=bias[1])
self.drop2 = nn.Dropout(drop_probs[1])
def forward(self, x):
x = self.fc1(x)
x = self.act(x)
x = self.drop1(x)
x = self.fc2(x)
x = self.drop2(x)
return x
class GluMlp(nn.Module):
""" MLP w/ GLU style gating
See: https://arxiv.org/abs/1612.08083, https://arxiv.org/abs/2002.05202
"""
def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.Sigmoid, bias=True, drop=0.):
super().__init__()
out_features = out_features or in_features
hidden_features = hidden_features or in_features
assert hidden_features % 2 == 0
bias = to_2tuple(bias)
drop_probs = to_2tuple(drop)
self.fc1 = nn.Linear(in_features, hidden_features, bias=bias[0])
self.act = act_layer()
self.drop1 = nn.Dropout(drop_probs[0])
self.fc2 = nn.Linear(hidden_features // 2, out_features, bias=bias[1])
self.drop2 = nn.Dropout(drop_probs[1])
def init_weights(self):
# override init of fc1 w/ gate portion set to weight near zero, bias=1
fc1_mid = self.fc1.bias.shape[0] // 2
nn.init.ones_(self.fc1.bias[fc1_mid:])
nn.init.normal_(self.fc1.weight[fc1_mid:], std=1e-6)
def forward(self, x):
x = self.fc1(x)
x, gates = x.chunk(2, dim=-1)
x = x * self.act(gates)
x = self.drop1(x)
x = self.fc2(x)
x = self.drop2(x)
return x
class GatedMlp(nn.Module):
""" MLP as used in gMLP
"""
def __init__(
self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU,
gate_layer=None, bias=True, drop=0.):
super().__init__()
out_features = out_features or in_features
hidden_features = hidden_features or in_features
bias = to_2tuple(bias)
drop_probs = to_2tuple(drop)
self.fc1 = nn.Linear(in_features, hidden_features, bias=bias[0])
self.act = act_layer()
self.drop1 = nn.Dropout(drop_probs[0])
if gate_layer is not None:
assert hidden_features % 2 == 0
self.gate = gate_layer(hidden_features)
hidden_features = hidden_features // 2 # FIXME base reduction on gate property?
else:
self.gate = nn.Identity()
self.fc2 = nn.Linear(hidden_features, out_features, bias=bias[1])
self.drop2 = nn.Dropout(drop_probs[1])
def forward(self, x):
x = self.fc1(x)
x = self.act(x)
x = self.drop1(x)
x = self.gate(x)
x = self.fc2(x)
x = self.drop2(x)
return x
class ConvMlp(nn.Module):
""" MLP using 1x1 convs that keeps spatial dims
"""
def __init__(
self, in_features, hidden_features=None, out_features=None, act_layer=nn.ReLU,
norm_layer=None, bias=True, drop=0.):
super().__init__()
out_features = out_features or in_features
hidden_features = hidden_features or in_features
bias = to_2tuple(bias)
self.fc1 = nn.Conv2d(in_features, hidden_features, kernel_size=1, bias=bias[0])
self.norm = norm_layer(hidden_features) if norm_layer else nn.Identity()
self.act = act_layer()
self.drop = nn.Dropout(drop)
self.fc2 = nn.Conv2d(hidden_features, out_features, kernel_size=1, bias=bias[1])
def forward(self, x):
x = self.fc1(x)
x = self.norm(x)
x = self.act(x)
x = self.drop(x)
x = self.fc2(x)
return x
|