pub fn hc_post(
gpu: &dyn GpuBackend,
kernel: KernelHandle,
block_out: DevicePtr,
residual: DevicePtr,
post: DevicePtr,
comb: DevicePtr,
out: DevicePtr,
num_tokens: u32,
hidden_size: u32,
hc_mult: u32,
stream: u64,
) -> Result<()>Expand description
Expand the sublayer output back into hc_mult streams, mixing the saved
residual streams through the doubly-stochastic comb. out may alias
residual. One block per token.