|
|
import torch |
|
|
import torch.nn as nn |
|
|
import torch.nn.functional as F |
|
|
|
|
|
|
|
|
class LambdaLayer(nn.Module): |
|
|
def __init__(self, lambd): |
|
|
super(LambdaLayer, self).__init__() |
|
|
self.lambd = lambd |
|
|
|
|
|
def forward(self, x): |
|
|
return self.lambd(x) |
|
|
|
|
|
|
|
|
class ResidualBlock(nn.Module): |
|
|
expansion = 1 |
|
|
|
|
|
def __init__(self, in_planes, planes, conv_layer, stride=1): |
|
|
super(ResidualBlock, self).__init__() |
|
|
self.conv1 = conv_layer(in_planes, planes, kernel_size=3, stride=stride, padding=1, bias=False) |
|
|
self.bn1 = nn.BatchNorm1d(planes) |
|
|
self.conv2 = conv_layer(planes, planes, kernel_size=3, stride=1, padding=1, bias=False) |
|
|
self.bn2 = nn.BatchNorm1d(planes) |
|
|
|
|
|
self.shortcut = nn.Sequential() |
|
|
if stride != 1 or in_planes != planes: |
|
|
diff = planes - in_planes |
|
|
self.shortcut = LambdaLayer( |
|
|
lambda x: F.pad(x[:, :, ::2], (0, 0, int(diff * 0.5), int((diff + 1) * 0.5)), "constant", 0)) |
|
|
|
|
|
def forward(self, x): |
|
|
out = F.relu(self.bn1(self.conv1(x))) |
|
|
out = self.bn2(self.conv2(out)) |
|
|
out += self.shortcut(x) |
|
|
out = F.relu(out) |
|
|
return out |
|
|
|
|
|
|
|
|
class GateFunction(nn.Module): |
|
|
def __init__(self, input_size, output_size): |
|
|
super(GateFunction, self).__init__() |
|
|
self.fc = nn.Linear(input_size, output_size) |
|
|
|
|
|
def forward(self, x): |
|
|
return F.softmax(self.fc(x), dim=-1) |
|
|
|
|
|
|
|
|
class NLPRecommendationRouter(nn.Module): |
|
|
def __init__(self, block, num_blocks, input_size=64, num_experts=4): |
|
|
super(NLPRecommendationRouter, self).__init__() |
|
|
self.in_planes = 16 |
|
|
self.conv_layer = nn.Conv1d |
|
|
|
|
|
self.conv1 = nn.Conv1d(1, self.in_planes, kernel_size=3, stride=1, padding=1, bias=False) |
|
|
self.bn1 = nn.BatchNorm1d(self.in_planes) |
|
|
self.layer1 = self._make_layer(block, 16, num_blocks[0], stride=1) |
|
|
self.layer2 = self._make_layer(block, 32, num_blocks[1], stride=2) |
|
|
self.layer3 = self._make_layer(block, 64, num_blocks[2], stride=2) |
|
|
self.avgpool = nn.AdaptiveAvgPool1d(1) |
|
|
|
|
|
|
|
|
self.gate = GateFunction(input_size, num_experts) |
|
|
|
|
|
def _make_layer(self, block, planes, num_blocks, stride): |
|
|
strides = [stride] + [1] * (num_blocks - 1) |
|
|
layers = [] |
|
|
for stride in strides: |
|
|
layers.append(block(self.in_planes, planes, self.conv_layer, stride)) |
|
|
self.in_planes = planes * block.expansion |
|
|
return nn.Sequential(*layers) |
|
|
|
|
|
def forward(self, x): |
|
|
out = F.relu(self.bn1(self.conv1(x))) |
|
|
out = self.layer1(out) |
|
|
out = self.layer2(out) |
|
|
out = self.layer3(out) |
|
|
out = self.avgpool(out) |
|
|
out = out.view(out.size(0), -1) |
|
|
out = self.gate(out) |
|
|
return out.unsqueeze(1) |
|
|
|
|
|
|
|
|
def build_router(**kwargs): |
|
|
return NLPRecommendationRouter(ResidualBlock, [3, 3, 3], input_size=64, num_experts=4) |
