|
from huggingface_hub import PyTorchModelHubMixin |
|
from torch import nn |
|
|
|
class SurfinBird(nn.Module, PyTorchModelHubMixin): |
|
def __init__(self, config: dict) -> None: |
|
super().__init__() |
|
self.conv1 = nn.Conv2d( |
|
in_channels=config["num_channels"], |
|
out_channels=64, |
|
kernel_size=7, |
|
stride=2, |
|
padding=3) |
|
self.bn1 = nn.BatchNorm2d(64) |
|
self.relu1 = nn.ReLU() |
|
self.mp1 = nn.MaxPool2d(kernel_size=2, |
|
stride=2) |
|
self.conv_block_2 = nn.Sequential( |
|
nn.Conv2d( |
|
in_channels=64, |
|
out_channels=64, |
|
kernel_size=3, |
|
stride=1, |
|
padding=1 |
|
), |
|
nn.BatchNorm2d(64), |
|
nn.ReLU(), |
|
nn.Conv2d( |
|
in_channels=64, |
|
out_channels=64, |
|
kernel_size=3, |
|
stride=1, |
|
padding=1 |
|
), |
|
nn.BatchNorm2d(64), |
|
nn.ReLU(), |
|
nn.Conv2d( |
|
in_channels=64, |
|
out_channels=64, |
|
kernel_size=3, |
|
stride=1, |
|
padding=1 |
|
), |
|
nn.BatchNorm2d(64), |
|
nn.ReLU(), |
|
nn.MaxPool2d(kernel_size=2, |
|
stride=2) |
|
) |
|
self.conv_block_3 = nn.Sequential( |
|
nn.Conv2d( |
|
in_channels=64, |
|
out_channels=128, |
|
kernel_size=3, |
|
stride=1, |
|
padding=1 |
|
), |
|
nn.BatchNorm2d(128), |
|
nn.ReLU(), |
|
nn.Conv2d( |
|
in_channels=128, |
|
out_channels=128, |
|
kernel_size=3, |
|
stride=1, |
|
padding=1 |
|
), |
|
nn.BatchNorm2d(128), |
|
nn.ReLU(), |
|
nn.Conv2d( |
|
in_channels=128, |
|
out_channels=128, |
|
kernel_size=3, |
|
stride=1, |
|
padding=1 |
|
), |
|
nn.BatchNorm2d(128), |
|
nn.ReLU(), |
|
nn.MaxPool2d(kernel_size=2, |
|
stride=2) |
|
) |
|
self.conv_block_4 = nn.Sequential( |
|
nn.Conv2d( |
|
in_channels=128, |
|
out_channels=128, |
|
kernel_size=3, |
|
stride=1, |
|
padding=1 |
|
), |
|
nn.BatchNorm2d(128), |
|
nn.ReLU(), |
|
nn.Conv2d( |
|
in_channels=128, |
|
out_channels=128, |
|
kernel_size=3, |
|
stride=1, |
|
padding=1 |
|
), |
|
nn.BatchNorm2d(128), |
|
nn.ReLU(), |
|
nn.Conv2d( |
|
in_channels=128, |
|
out_channels=128, |
|
kernel_size=3, |
|
stride=1, |
|
padding=1 |
|
), |
|
nn.BatchNorm2d(128), |
|
nn.ReLU(), |
|
nn.MaxPool2d(kernel_size=2, |
|
stride=2) |
|
) |
|
self.conv_block_5 = nn.Sequential( |
|
nn.Conv2d( |
|
in_channels=128, |
|
out_channels=256, |
|
kernel_size=3, |
|
stride=1, |
|
padding=1 |
|
), |
|
nn.BatchNorm2d(256), |
|
nn.ReLU(), |
|
nn.Conv2d( |
|
in_channels=256, |
|
out_channels=256, |
|
kernel_size=3, |
|
stride=1, |
|
padding=1 |
|
), |
|
nn.BatchNorm2d(256), |
|
nn.ReLU(), |
|
nn.Conv2d( |
|
in_channels=256, |
|
out_channels=256, |
|
kernel_size=3, |
|
stride=1, |
|
padding=1 |
|
), |
|
nn.BatchNorm2d(256), |
|
nn.ReLU(), |
|
nn.MaxPool2d(kernel_size=2, |
|
stride=2) |
|
) |
|
self.conv_block_6 = nn.Sequential( |
|
nn.Conv2d( |
|
in_channels=256, |
|
out_channels=256, |
|
kernel_size=3, |
|
stride=1, |
|
padding=1 |
|
), |
|
nn.BatchNorm2d(256), |
|
nn.ReLU(), |
|
nn.Conv2d( |
|
in_channels=256, |
|
out_channels=256, |
|
kernel_size=3, |
|
stride=1, |
|
padding=1 |
|
), |
|
nn.BatchNorm2d(256), |
|
nn.ReLU(), |
|
nn.Conv2d( |
|
in_channels=256, |
|
out_channels=256, |
|
kernel_size=3, |
|
stride=1, |
|
padding=1 |
|
), |
|
nn.BatchNorm2d(256), |
|
nn.ReLU(), |
|
nn.MaxPool2d(kernel_size=2, |
|
stride=2) |
|
) |
|
|
|
self.avgpool = nn.Sequential( |
|
nn.AdaptiveAvgPool2d(output_size=(1, 1)) |
|
) |
|
|
|
self.classifier = nn.Sequential( |
|
nn.Flatten(), |
|
nn.Linear(in_features=config["hidden_units"]*1*1, |
|
out_features=config["num_classes"]) |
|
) |
|
def forward(self, x: torch.Tensor): |
|
return self.classifier(self.avgpool(self.conv_block_6(self.conv_block_5(self.conv_block_4(self.conv_block_3(self.conv_block_2(self.mp1(self.relu1(self.bn1(self.conv1(x))))))))))) |
|
|
|
config = {"num_channels": 3, "hidden_units": 256, "num_classes": 525} |