| channels: 16 |
| n_layers: 22 |
| n_heads: 40 |
| n_kv_heads: 20 |
| d_model: 2560 |
| mlp_ratio: 5 |
| mlp_gradient_checkpointing: true |
| causal: true |
| n_buttons: 256 |
| tokens_per_frame: 256 |
| height: 16 |
| width: 16 |
| patch: |
| - 2 |
| - 2 |
| base_fps: 60 |
| local_window: 16 |
| global_window: 128 |
| global_pinned_dilation: 8 |
| global_attn_period: 4 |
| global_attn_offset: 0 |
| value_residual: false |
| n_frames: 800 |
| rope_impl: ortho |
| gated_attn: true |
| noise_conditioning: wan |
| ctrl_conditioning: mlp_fusion |
| ctrl_cond_dropout: 0.0 |
| ctrl_conditioning_period: 3 |
| prompt_conditioning: cross_attention |
| prompt_cond_dropout: 0.0 |
| prompt_embedding_dim: 2048 |
| prompt_conditioning_period: 3 |
| block_gradient_checkpointing: true |
| scheduler_sigmas: |
| - 1.0 |
| - 0.8609585762023926 |
| - 0.729332447052002 |
| - 0.3205108940601349 |
| - 0.0 |
| ae_uri: "OpenWorldLabs/owl_vae_f16_c16_distill_v0_nogan" |
