PR 636 miss fix for FusedAddRMSNorm function

[Atream]

In #636, you write const uint32_t smem_size = (ceil_div(num_warps, 4) * 4 + d) * sizeof(float); in GemmaFusedAddRMSNorm but not in FusedAddRMSNorm function.

[yzh119]

Hi @Atream , yes I think you are correct, would you mind creating a pull request to fix it?

[Atream]

Yes, I've created a pull request to address the issue. You can find it here #646 . Please review and let me know if there's anything else that needs to be adjusted. Thanks for your response!