{
    "model_type": "diffusion_cond",
    "sample_size": 4194304,
    "sample_rate": 44100,
    "audio_channels": 2,
    "model": {
        "pretransform": {
            "type": "autoencoder",
            "iterate_batch": true,
            "config": {
                "encoder": {
                    "type": "dac",
                    "config": {
                        "in_channels": 2,
                        "latent_dim": 128,
                        "d_model": 128,
                        "strides": [4, 4, 8, 8]
                    }
                },
                "decoder": {
                    "type": "dac",
                    "config": {
                        "out_channels": 2,
                        "latent_dim": 64,
                        "channels": 1536,
                        "rates": [8, 8, 4, 4]
                    }
                },
                "bottleneck": {
                    "type": "vae"
                },
                "latent_dim": 64,
                "downsampling_ratio": 1024,
                "io_channels": 2
            }
        },
        "conditioning": {
            "configs": [
                {
                    "id": "prompt",
                    "type": "clap_text",
                    "config": {
                        "audio_model_type": "HTSAT-base",
                        "enable_fusion": true,
                        "clap_ckpt_path": "/path/to/clap.ckpt",
                        "use_text_features": true,
                        "feature_layer_ix": -2
                    }
                },
                {
                    "id": "seconds_start",
                    "type": "int",
                    "config": {
                        "min_val": 0,
                        "max_val": 512
                    }
                },
                {
                    "id": "seconds_total",
                    "type": "int",
                    "config": {
                        "min_val": 0,
                        "max_val": 512
                    }
                }
            ],
            "cond_dim": 768
        },
        "diffusion": {
            "type": "adp_cfg_1d",
            "cross_attention_cond_ids": ["prompt", "seconds_start", "seconds_total"],
            "config": {
                "in_channels": 64,
                "context_embedding_features": 768,
                "context_embedding_max_length": 79,
                "channels": 256,
                "resnet_groups": 16,
                "kernel_multiplier_downsample": 2,
                "multipliers": [4, 4, 4, 5, 5],
                "factors": [1, 2, 2, 4],
                "num_blocks": [2, 2, 2, 2],
                "attentions": [1, 3, 3, 3, 3],
                "attention_heads": 16,
                "attention_multiplier": 4,
                "use_nearest_upsample": false,
                "use_skip_scale": true,
                "use_context_time": true
            }
        },
        "io_channels": 64
    },
    "training": {
        "learning_rate": 4e-5,
        "demo": {
            "demo_every": 2000,
            "demo_steps": 250,
            "num_demos": 4,
            "demo_cond": [
                {"prompt": "A beautiful piano arpeggio", "seconds_start": 0, "seconds_total": 95},
                {"prompt": "A tropical house track with upbeat melodies, a driving bassline, and cheery vibes", "seconds_start": 0, "seconds_total": 90},
                {"prompt": "A cool 80s glam rock song with driving drums and distorted guitars", "seconds_start": 0, "seconds_total": 180},
                {"prompt": "A grand orchestral arrangement with thunderous percussion, epic brass fanfares, and soaring strings, creating a cinematic atmosphere fit for a heroic battle.", "seconds_start": 0, "seconds_total": 60}
            ],
            "demo_cfg_scales": [3, 6, 9]
        }
    }
}