{
    "latent_size": 256,
    "model_type": "conv_vae"
}