| { | |
| "model_type": "diffusion_cond", | |
| "sample_size": 485100, | |
| "sample_rate": 44100, | |
| "video_fps": 5, | |
| "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", | |
| "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": "video_prompt", | |
| "type": "clip-with-sync-w-empty-feat", | |
| "config": { | |
| "clip_model_name": "clip-vit-base-patch32" | |
| } | |
| }, | |
| { | |
| "id": "text_prompt", | |
| "type": "t5", | |
| "config": { | |
| "t5_model_name": "t5-base", | |
| "max_length": 128 | |
| } | |
| }, | |
| { | |
| "id": "audio_prompt", | |
| "type": "audio_autoencoder_v2", | |
| "config": { | |
| "sample_rate": 44100, | |
| "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", | |
| "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 | |
| } | |
| }, | |
| "pretransform_ckpt_path": "./model/VAE.ckpt", | |
| "latent_seq_len": 215, | |
| "mask_ratio_start": 0, | |
| "mask_ratio_end": 0 | |
| } | |
| } | |
| ], | |
| "cond_dim": 768 | |
| }, | |
| "diffusion": { | |
| "cross_attention_cond_ids": ["video_prompt", "text_prompt", "audio_prompt"], | |
| "global_cond_ids": [], | |
| "type": "mmdit", | |
| "gate": true, | |
| "gate_type": "MAF", | |
| "gate_type_config": { | |
| "num_experts_per_modality": 64, | |
| "num_heads": 24, | |
| "num_fusion_layers": 8 | |
| }, | |
| "config": { | |
| "io_channels": 64, | |
| "embed_dim": 1536, | |
| "depth": 24, | |
| "num_heads": 24, | |
| "cond_token_dim": 768, | |
| "global_cond_dim": 768, | |
| "project_cond_tokens": false, | |
| "transformer_type": "continuous_transformer", | |
| "video_fps": 5 | |
| } | |
| }, | |
| "io_channels": 64 | |
| } | |
| } |