transformer/config.json

{
  "_class_name": "WorldModel",
  "_diffusers_version": "0.36.0.dev0",
  "auto_map": {
    "AutoModel": "model.WorldModel"
  },
  "d_model": 2560,
  "n_heads": 40,
  "n_kv_heads": 20,
  "n_layers": 22,
  "mlp_ratio": 5,
  "channels": 16,
  "height": 16,
  "width": 16,
  "patch": [
    2,
    2
  ],
  "tokens_per_frame": 256,
  "n_frames": 4096,
  "local_window": 16,
  "global_window": 128,
  "global_attn_period": 4,
  "global_pinned_dilation": 8,
  "global_attn_offset": 0,
  "value_residual": false,
  "gated_attn": true,
  "n_buttons": 256,
  "ctrl_conditioning": "mlp_fusion",
  "ctrl_conditioning_period": 3,
  "ctrl_cond_dropout": 0.0,
  "prompt_conditioning": "cross_attention",
  "prompt_conditioning_period": 3,
  "prompt_embedding_dim": 2048,
  "prompt_cond_dropout": 0.0,
  "noise_conditioning": "wan",
  "base_fps": 60,
  "causal": true,
  "mlp_gradient_checkpointing": true,
  "block_gradient_checkpointing": true,
  "rope_impl": "ortho",
  "scheduler_sigmas": [
    1.0,
    0.8609585762023926,
    0.729332447052002,
    0.3205108940601349,
    0.0
  ]
}