source/model.py

from __future__ import annotations

import torch
from torch import nn


class ConditionalVAE(nn.Module):
    def __init__(self, latent_dimensions: int = 8) -> None:
        super().__init__()
        self.latent_dimensions = latent_dimensions
        self.label_embedding = nn.Embedding(10, 8)
        self.encoder = nn.Sequential(
            nn.Linear(64, 64),
            nn.GELU(),
            nn.Linear(64, 32),
            nn.GELU(),
        )
        self.mean = nn.Linear(32, latent_dimensions)
        self.log_variance = nn.Linear(32, latent_dimensions)
        self.decoder = nn.Sequential(
            nn.Linear(latent_dimensions + 8, 32),
            nn.GELU(),
            nn.Linear(32, 64),
            nn.Sigmoid(),
        )

    def encode(self, pixels: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
        hidden = self.encoder(pixels)
        return self.mean(hidden), self.log_variance(hidden)

    def reparameterize(
        self,
        mean: torch.Tensor,
        log_variance: torch.Tensor,
    ) -> torch.Tensor:
        if not self.training:
            return mean
        standard_deviation = torch.exp(0.5 * log_variance)
        return mean + torch.randn_like(standard_deviation) * standard_deviation

    def decode(self, latent: torch.Tensor, labels: torch.Tensor) -> torch.Tensor:
        condition = self.label_embedding(labels)
        return self.decoder(torch.cat([latent, condition], dim=1))

    def forward(
        self,
        pixels: torch.Tensor,
        labels: torch.Tensor,
    ) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
        mean, log_variance = self.encode(pixels)
        latent = self.reparameterize(mean, log_variance)
        return self.decode(latent, labels), mean, log_variance


class TinyVisionJudge(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.MaxPool2d(2),
        )
        self.classifier = nn.Sequential(
            nn.Flatten(),
            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())