[SIGGRAPH Asia 2026] 4DAnyone: Create Anyone in 4D from a Casual Monocular Video
"""Shared denoising math for RCP, single-GPU, and multi-GPU execution."""
from __future__ import annotations
from typing import TYPE_CHECKING
if TYPE_CHECKING:
from torch import Tensor
from fdanyone.model.loader import Denoiser
def denoise_group(
denoiser: Denoiser,
latents: Tensor,
source: Tensor,
context: Tensor,
pose_features: Tensor,
null_pose_feature: Tensor,
step_index: int,
) -> Tensor:
"""Advance one camera group by one scheduler step."""
import torch
timestep = denoiser.scheduler.timesteps[step_index]
batched_timestep = timestep.unsqueeze(0).to(dtype=denoiser.dtype, device=latents.device)
batched_timestep = torch.cat([batched_timestep] * latents.shape[0], dim=0)
prediction = denoiser.model(
x=latents,
x_src=source,
timestep=batched_timestep,
pose_features=pose_features,
null_pose_feature=null_pose_feature,
context=context,
)
return denoiser.scheduler.step(prediction, timestep, latents)