pub fn hc_expand(
gpu: &dyn GpuBackend,
kernel: KernelHandle,
hidden: DevicePtr,
streams: DevicePtr,
num_tokens: u32,
hidden_size: u32,
hc_mult: u32,
stream: u64,
) -> Result<()>Expand description
2026-09-25: Broadcast the BF16 hidden state into hc_mult identical FP32 streams:
streams[t, i, d] = hidden[t, d].