pub fn moe_topk_softmax(
gpu: &dyn GpuBackend,
kernel: KernelHandle,
gate_logits: DevicePtr,
expert_indices: DevicePtr,
expert_weights: DevicePtr,
num_experts: u32,
top_k: u32,
normalize: bool,
stream: u64,
) -> Result<()>Expand description
2026-09-25: Top-k experts of one token’s gate logits, with softmax weights, in one block.
The kernel handle decides the logit type (moe_topk_softmax BF16, moe_topk_softmax_f32).