Function sigmoid_gate_mul_batched 

pub fn sigmoid_gate_mul_batched(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    input: DevicePtr,
    gate: DevicePtr,
    output: DevicePtr,
    dim: u32,
    gate_stride: u32,
    num_tokens: u32,
    stream: u64,
) -> Result<()>

2026-09-25: sigmoid_gate_mul over num_tokens rows in one launch: output[t, d] = input[t, d] * sigmoid(gate[t * gate_stride + d]). input and output are contiguous [num_tokens, dim].