{
    "model_type": "mm_diffusion_cond",
    "sample_size": 397312,
    "sample_rate": 44100,
    "audio_channels": 2,
    "model": {
        "pretransform": {
            "type": "autoencoder",
            "iterate_batch": true,
            "config": {
                "encoder": {
                    "type": "oobleck",
                    "config": {
                        "in_channels": 2,
                        "channels": 128,
                        "c_mults": [1, 2, 4, 8, 16],
                        "strides": [2, 4, 4, 8, 8],
                        "latent_dim": 128,
                        "use_snake": true
                    }
                },
                "decoder": {
                    "type": "oobleck",
                    "config": {
                        "out_channels": 2,
                        "channels": 128,
                        "c_mults": [1, 2, 4, 8, 16],
                        "strides": [2, 4, 4, 8, 8],
                        "latent_dim": 64,
                        "use_snake": true,
                        "final_tanh": false
                    }
                },
                "bottleneck": {
                    "type": "vae"
                },
                "latent_dim": 64,
                "downsampling_ratio": 2048,
                "io_channels": 2
            }
        },
        "conditioning": {
            "configs": [
                {
                    "id": "metaclip_features",
                    "type": "mm_unchang",
                    "config": {
                        "dim": 1024,
                        "output_dim": 1024
                        }
                },
                {
                    "id": "metaclip_text_features",
                    "type": "mm_unchang",
                    "config": {
                        "dim": 1024,
                        "output_dim": 1024
                    }
                },
                {
                    "id": "sync_features",
                    "type": "mm_unchang",
                    "config": {
                        "dim": 768,
                        "output_dim": 768
                    }
                },
                {
                    "id": "t5_features",
                    "type": "mm_unchang",
                    "config": {
                        "dim": 2048,
                        "output_dim": 2048
                    }
                }
            ],
            "cond_dim": 768
        },
        "diffusion": {
            "mm_cond_ids": ["metaclip_features", "sync_features", "metaclip_text_features","t5_features"],
            "type": "mmdit",
            "diffusion_objective": "rectified_flow",
            "config": {
                "latent_dim":64,
                "clip_dim":1024,
                "sync_dim":768,
                "text_dim":2048,
                "hidden_dim":1024,
                "depth":21,
                "fused_depth":14,
                "num_heads":16,
                "latent_seq_len":194,
                "clip_seq_len":72,
                "sync_seq_len":216,
                "v2": true,
                "kernel_size": 3
            }
        },
        "io_channels": 64
    },
    "training": {
        "use_ema": true,
        "log_loss_info": false,
        "cfg_dropout_prob": 0.2,
        "pre_encoded": true,
        "timestep_sampler": "logit_normal",
        "optimizer_configs": {
            "diffusion": {
                "optimizer": {
                    "type": "AdamW",
                    "config": {
                        "lr": 5e-5,
                        "betas": [0.9, 0.95],
                        "weight_decay": 1e-4,
                        "eps": 1e-6
                    }
                },
                "scheduler": {
                    "type": "InverseLR",
                    "config": {
                        "inv_gamma": 1000000,
                        "power": 0.5,
                        "warmup": 0.99
                    }
                }
            }
        },
        "demo": {
            "demo_every": 5000,
            "demo_steps": 24,
            "num_demos": 10,
            "demo_cond": [
                "dataset/vggsound/video_latents_t5_clip_npz/test/0Cu33yBwAPg_000060.npz",
                "dataset/vggsound/video_latents_t5_clip_npz/test/bmKtI808DsU_000009.npz",
                "dataset/vggsound/video_latents_t5_clip_npz/test/VC0c22cJTbM_000424.npz",
                "dataset/vggsound/video_latents_t5_clip_npz/test/F3gsbUTdc2U_000090.npz",
                "dataset/vggsound/video_latents_t5_clip_npz/test/WatvT8A8iug_000100.npz",
                "dataset/vggsound/video_latents_t5_clip_npz/test/0nvBTp-q7tU_000112.npz",
                "dataset/vggsound/video_latents_t5_clip_npz/test/3-PFuDkTM48_000080.npz",
                "dataset/vggsound/video_latents_t5_clip_npz/test/luSAuu-BoPs_000232.npz",
                "dataset/vggsound/video_latents_t5_clip_npz/test/__8UJxW0aOQ_000002.npz",
                "dataset/vggsound/video_latents_t5_clip_npz/test/_0m_YMpQayA_000168.npz"
            ],
            "demo_cfg_scales": [5]
        }
    }
}