File size: 3,854 Bytes
00d514b | 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 | """
Protected Model — основна модель, яку ми захищаємо від отруєння.
Це невеликий ResNet для CIFAR-розмірів.
Її задача — звичайна класифікація.
Захист реалізований у train.py через зважування loss за trust_weights від Detector'а.
"""
import torch
import torch.nn as nn
import torch.nn.functional as F
class BasicBlock(nn.Module):
"""Стандартний ResNet block."""
expansion = 1
def __init__(self, in_channels: int, out_channels: int, stride: int = 1):
super().__init__()
self.conv1 = nn.Conv2d(in_channels, out_channels, 3, stride, 1, bias=False)
self.bn1 = nn.BatchNorm2d(out_channels)
self.conv2 = nn.Conv2d(out_channels, out_channels, 3, 1, 1, bias=False)
self.bn2 = nn.BatchNorm2d(out_channels)
self.shortcut = nn.Sequential()
if stride != 1 or in_channels != out_channels:
self.shortcut = nn.Sequential(
nn.Conv2d(in_channels, out_channels, 1, stride, bias=False),
nn.BatchNorm2d(out_channels),
)
def forward(self, x: torch.Tensor) -> torch.Tensor:
out = F.relu(self.bn1(self.conv1(x)), inplace=True)
out = self.bn2(self.conv2(out))
out = out + self.shortcut(x)
return F.relu(out, inplace=True)
class ProtectedModel(nn.Module):
"""
Малий ResNet (~ResNet-14) для CIFAR.
"""
def __init__(self, num_classes: int = 10, in_channels: int = 3):
super().__init__()
self.stem = nn.Sequential(
nn.Conv2d(in_channels, 64, 3, 1, 1, bias=False),
nn.BatchNorm2d(64),
nn.ReLU(inplace=True),
)
self.layer1 = self._make_layer(64, 64, num_blocks=2, stride=1)
self.layer2 = self._make_layer(64, 128, num_blocks=2, stride=2)
self.layer3 = self._make_layer(128, 256, num_blocks=2, stride=2)
self.pool = nn.AdaptiveAvgPool2d(1)
self.fc = nn.Linear(256, num_classes)
def _make_layer(self, in_ch: int, out_ch: int, num_blocks: int, stride: int) -> nn.Sequential:
layers = [BasicBlock(in_ch, out_ch, stride)]
for _ in range(num_blocks - 1):
layers.append(BasicBlock(out_ch, out_ch, 1))
return nn.Sequential(*layers)
def forward(self, x: torch.Tensor) -> torch.Tensor:
x = self.stem(x)
x = self.layer1(x)
x = self.layer2(x)
x = self.layer3(x)
x = self.pool(x).flatten(1)
return self.fc(x)
def weighted_cross_entropy(
logits: torch.Tensor, targets: torch.Tensor, weights: torch.Tensor
) -> torch.Tensor:
"""
Cross-entropy з вагою на кожен зразок.
Це ключове місце "імунного захисту":
weights близько 1 → нормальне навчання
weights близько 0 → зразок майже не впливає
"""
per_sample_loss = F.cross_entropy(logits, targets, reduction="none")
# Нормалізуємо за сумою ваг, щоб масштаб loss'у був стабільним
weight_sum = weights.sum().clamp(min=1e-6)
return (per_sample_loss * weights).sum() / weight_sum
if __name__ == "__main__":
model = ProtectedModel(num_classes=10)
x = torch.rand(4, 3, 32, 32)
out = model(x)
print(f"Output shape: {out.shape}")
print(f"Params: {sum(p.numel() for p in model.parameters()) / 1e6:.2f}M")
# Тест weighted loss
targets = torch.randint(0, 10, (4,))
weights = torch.tensor([1.0, 0.1, 1.0, 0.0]) # 2-й трохи підозрілий, 4-й — точно отрута
loss = weighted_cross_entropy(out, targets, weights)
print(f"Weighted loss: {loss.item():.4f}")
|