Function moe_transpose_u8_batched 

pub fn moe_transpose_u8_batched(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    src_ptrs: DevicePtr,
    dst_ptrs: DevicePtr,
    rows: u32,
    cols: u32,
    num_experts: u32,
    stream: u64,
) -> Result<()>

2026-09-25: Per-expert uint8 transpose for the MoE down_proj relayout: reads expert e’s [rows, cols] matrix through src_ptrs[e] and writes it as [cols, rows] through dst_ptrs[e]. A NULL entry in either table skips that expert.