Function moe_topk_sigmoid_batched 

pub fn moe_topk_sigmoid_batched(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    gate_logits: DevicePtr,
    bias: DevicePtr,
    expert_indices: DevicePtr,
    expert_weights: DevicePtr,
    num_experts: u32,
    top_k: u32,
    normalize: bool,
    scaling_factor: f32,
    num_tokens: u32,
    stream: u64,
) -> Result<()>

Batched sigmoid + correction-bias top-K MoE routing.

Kernel: moe_topk_sigmoid_batched(gate_logits, bias, expert_indices, expert_weights, num_experts, top_k, normalize, scaling_factor) Grid: (num_tokens, 1, 1) Block: (256, 1, 1)