| 461 | input_path: str = field(default_factory=str) |
| 462 | save_result_path: str = field(default_factory=str) # output root dir |
| 463 | strict_output_path: str = field(default_factory=lambda: None) |
| 464 | return_result_tensor: bool = field(default_factory=lambda: False) |
| 465 | # Optional priors |
| 466 | prior_cam_path: str = field(default_factory=lambda: None) |
| 467 | prior_depth_path: str = field(default_factory=lambda: None) |
| 468 | |
| 469 | |
| 470 | @dataclass |
| 471 | class WorldPlayT2VInputInfo: |
| 472 | """Input info for WorldPlay model (text-to-video with action/pose conditioning).""" |
| 473 | |
| 474 | seed: int = field(default_factory=int) |
| 475 | prompt: str = field(default_factory=str) |
| 476 | prompt_enhanced: str = field(default_factory=str) |
| 477 | negative_prompt: str = field(default_factory=str) |
| 478 | save_result_path: str = field(default_factory=str) |
| 479 | return_result_tensor: bool = field(default_factory=lambda: False) |
| 480 | # shape related |
| 481 | resize_mode: str = field(default_factory=str) |
| 482 | latent_shape: list = field(default_factory=list) |
| 483 | target_shape: list = field(default_factory=list) |
| 484 | # WorldPlay-specific: pose/action conditioning |
| 485 | pose: str = field(default_factory=str) # Pose string (e.g., "w-3, right-0.5") or JSON path |
| 486 | model_type: str = field(default_factory=lambda: "ar") # "ar" (autoregressive) or "bi" (bidirectional) |
| 487 | chunk_latent_frames: int = field(default_factory=lambda: 4) |
| 488 | # Computed pose tensors (set during processing) |
| 489 | viewmats: torch.Tensor = field(default_factory=lambda: None) |
| 490 | Ks: torch.Tensor = field(default_factory=lambda: None) |
| 491 | action: torch.Tensor = field(default_factory=lambda: None) |
| 492 | |
| 493 | |
| 494 | @dataclass |
| 495 | class SenseNovaVisionInputInfo: |
| 496 | seed: int = field(default_factory=lambda: 42) |
| 497 | prompt: str = field(default_factory=str) |
| 498 | image_path: str = field(default_factory=str) |
| 499 | save_result_path: str = field(default_factory=str) |
| 500 | return_result_tensor: bool = field(default_factory=lambda: False) |
| 501 | target_shape: list = field(default_factory=list) |
| 502 | omni_vision_subtask: str = field(default_factory=str) |