ntkuhn/anime-diffusion

Anime Face Diffusion Model

A diffusion model trained to generate 64x64 anime-style faces.

Model Details

• Model type: Denoising Diffusion Probabilistic Model (DDPM)
• Training data: Anime faces dataset
• Image size: 64x64 RGB
• Timesteps: 1000
• Architecture: U-Net with sinusoidal time embeddings

Usage

import torch
from diffusion_model import SimpleUnet, Diffuser  # Your model code

# Load the model
checkpoint = torch.load("model.pt")
model = SimpleUnet(in_channels=3)
model.load_state_dict(checkpoint['ema_model_state_dict'])  # Use EMA weights
model.eval()

# Generate samples
# (Add your sampling code here)

Training Details

• Trained for X epochs
• Batch size: 64
• Learning rate: 1e-4
• EMA decay: 0.9999
• Noise schedule: Cosine

Samples

(Add generated sample images here)

Citation

If you use this model, please cite:

@misc{your-anime-diffusion,
  author = {Your Name},
  title = {Anime Face Diffusion Model},
  year = {2024},
  publisher = {HuggingFace},
  url = {https://huggingface.co/your-username/anime-diffusion}
}