update neo++ seed

[helloyongyang]

No description provided.

[gemini-code-assist]

Code Review

This pull request updates example scripts with new file paths and null seeds, makes scipy imports optional in the WanMatrixGame3 runner, and adds a conditional check for seeding in the pipeline. Feedback highlights that skipping seed initialization could lead to crashes in runners expecting an integer, suggesting a random seed be generated instead. Furthermore, the use of hardcoded absolute paths in examples is discouraged for portability, and the new optional scipy imports require safety checks in the functions that utilize them to avoid runtime errors.

[gemini-code-assist]

Setting self.seed to None and skipping seed_all will result in input_info.seed being None. This will cause a crash in runners that use torch.manual_seed(input_info.seed) (such as WanMatrixGame3Runner at line 1624). It is better to initialize self.seed with a random integer if it is None to ensure downstream compatibility.

Suggested change

	if self.seed is not None:
	seed_all(self.seed)
	if self.seed is None:
	self.seed = torch.randint(0, 2**32, (1,)).item()
	seed_all(self.seed)

[gemini-code-assist]

Hardcoding absolute paths that include personal user directories (e.g., /yongyang/kkk/) makes the example less portable and harder for others to use. Consider using relative paths or environment variables for output directories.

[gemini-code-assist]

Wrapping these imports in a try-except block without corresponding checks in the functions that use them (e.g., _matrix_game3_interpolate_camera_poses) will lead to a TypeError at runtime if scipy is missing. If scipy is an optional dependency, please add checks to provide a more helpful error message when these features are accessed.