| { | |
| "model_type": "diffusion_cond", | |
| "sample_size": 2097152, | |
| "sample_rate": 44100, | |
| "audio_channels": 2, | |
| "model": { | |
| "pretransform": { | |
| "type": "autoencoder", | |
| "iterate_batch": true, | |
| "config": { | |
| "encoder": { | |
| "type": "oobleck", | |
| "requires_grad": false, | |
| "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", | |
| "requires_grad": false, | |
| "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": "prompt", | |
| "type": "t5", | |
| "config": { | |
| "t5_model_name": "t5-base", | |
| "max_length": 128 | |
| } | |
| }, | |
| { | |
| "id": "seconds_start", | |
| "type": "number", | |
| "config": { | |
| "min_val": 0, | |
| "max_val": 512 | |
| } | |
| }, | |
| { | |
| "id": "seconds_total", | |
| "type": "number", | |
| "config": { | |
| "min_val": 0, | |
| "max_val": 512 | |
| } | |
| }, | |
| { | |
| "id": "input_audio", | |
| "type": "pretransform", | |
| "config": { | |
| "sample_rate": 44100, | |
| "pretransform_ckpt_path": "vae_model.ckpt", | |
| "pretransform_config": { | |
| "type": "autoencoder", | |
| "iterate_batch": true, | |
| "config": { | |
| "encoder": { | |
| "type": "oobleck", | |
| "requires_grad": false, | |
| "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", | |
| "requires_grad": false, | |
| "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 | |
| } | |
| } | |
| } | |
| } | |
| ], | |
| "cond_dim": 768 | |
| }, | |
| "diffusion": { | |
| "cross_attention_cond_ids": ["prompt", "seconds_start", "seconds_total"], | |
| "global_cond_ids": ["seconds_start", "seconds_total"], | |
| "input_concat_ids": ["input_audio"], | |
| "type": "dit", | |
| "config": { | |
| "io_channels": 64, | |
| "embed_dim": 1536, | |
| "depth": 24, | |
| "num_heads": 24, | |
| "cond_token_dim": 768, | |
| "global_cond_dim": 1536, | |
| "project_cond_tokens": false, | |
| "transformer_type": "continuous_transformer", | |
| "input_concat_dim": 64 | |
| } | |
| }, | |
| "io_channels": 64 | |
| }, | |
| "training": { | |
| "use_ema": true, | |
| "log_loss_info": false, | |
| "optimizer_configs": { | |
| "diffusion": { | |
| "optimizer": { | |
| "type": "AdamW", | |
| "config": { | |
| "lr": 5e-5, | |
| "betas": [0.9, 0.999], | |
| "weight_decay": 1e-3 | |
| } | |
| }, | |
| "scheduler": { | |
| "type": "InverseLR", | |
| "config": { | |
| "inv_gamma": 1000000, | |
| "power": 0.5, | |
| "warmup": 0.99 | |
| } | |
| } | |
| } | |
| }, | |
| "demo": { | |
| "demo_every": 500, | |
| "demo_steps": 50, | |
| "num_demos": 6, | |
| "demo_cond_from_batch": true, | |
| "demo_cfg_scales": [6] | |
| } | |
| } | |
| } |