source/model.py

from __future__ import annotations

import torch
from torch import nn


class ProbabilisticTinyCNN(nn.Module):
    def __init__(self) -> None:
        super().__init__()
        self.features = nn.Sequential(
            nn.Conv2d(1, 8, kernel_size=3, padding=1),
            nn.GELU(),
            nn.Conv2d(8, 8, kernel_size=3, padding=1, groups=8),
            nn.GELU(),
            nn.Conv2d(8, 12, kernel_size=1),
            nn.GELU(),
            nn.Dropout2d(0.08),
            nn.MaxPool2d(2),
        )
        self.classifier = nn.Sequential(
            nn.Flatten(),
            nn.Dropout(0.12),
            nn.Linear(12 * 4 * 4, 10),
        )

    def forward(self, pixels: torch.Tensor) -> torch.Tensor:
        return self.classifier(self.features(pixels))


def parameter_count(model: nn.Module) -> int:
    return sum(parameter.numel() for parameter in model.parameters())