pub fn hc_head_mean(
gpu: &dyn GpuBackend,
kernel: KernelHandle,
streams: DevicePtr,
y_out: DevicePtr,
num_tokens: u32,
hidden_size: u32,
hc_mult: u32,
stream: u64,
) -> Result<()>Expand description
2026-09-25: Collapse the highway to one BF16 row per token: the unweighted mean of the
hc_mult streams. It takes no weights, unlike DeepSeek-V4’s ops::hc_head.