vllm_omni.diffusion.layers.indexed_modulation ¶
Fused indexed modulation with FP32 accumulation.
indexed_gate ¶
Return x + gate[indices] * other without indexed temporaries.
indexed_gate_rms_norm_scale_shift ¶
indexed_gate_rms_norm_scale_shift(
residual: Tensor,
gate: Tensor,
branch: Tensor,
weight: Tensor,
shift: Tensor,
scale: Tensor,
indices: Tensor,
eps: float,
) -> tuple[Tensor, Tensor]
Fuse gated residual, the following RMSNorm, and AdaLN affine.
indexed_scale_shift_ ¶
Apply indexed scale/shift in-place to a disposable contiguous input.