{
    "model_type": "autoencoder",
    "sample_size": 65536,
    "sample_rate": 44100,
    "audio_channels": 1,
    "model": {
        "encoder": {
            "type": "dac",
            "config": {
                "latent_dim": 64,
                "d_model": 128,
                "strides": [4, 8, 8, 8]
            }
        },
        "decoder": {
            "type": "dac",
            "config": {
                "latent_dim": 32,
                "channels": 1536,
                "rates": [8, 8, 8, 4]
            }
        },
        "bottleneck": {
            "type": "vae"
        },
        "latent_dim": 32,
        "downsampling_ratio": 2048,
        "io_channels": 1
    },
    "training": {
        "learning_rate": 1e-4,
        "warmup_steps": 0,
        "use_ema": false,
        "loss_configs": {
            "discriminator": {
                "type": "encodec",
                "config": {
                    "filters": 32,
                    "n_ffts": [2048, 1024, 512, 256, 128, 64, 32],
                    "hop_lengths": [512, 256, 128, 64, 32, 16, 8],
                    "win_lengths": [2048, 1024, 512, 256, 128, 64, 32]
                },
                "weights": {
                    "adversarial": 0.1,
                    "feature_matching": 5.0
                }
            },
            "spectral": {
                "type": "mrstft",
                "config": {
                    "fft_sizes": [2048, 1024, 512, 256, 128, 64, 32],
                    "hop_sizes": [512, 256, 128, 64, 32, 16, 8],
                    "win_lengths": [2048, 1024, 512, 256, 128, 64, 32],
                    "perceptual_weighting": true
                },
                "weights": {
                    "mrstft": 1.0
                }
            },
            "time": {
                "type": "l1",
                "weights": {
                    "l1": 0.0
                }
            }
        },
        "demo": {
            "demo_every": 2000
        }
    }
}