From 1529608b7f2cfba8b71853bed3da0e465f8745dd Mon Sep 17 00:00:00 2001 From: PanZezhong Date: Mon, 18 Aug 2025 14:23:24 +0800 Subject: [PATCH] issue/6/fix type convertion --- src/infiniop/ops/rms_norm/cuda/kernel.cuh | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/infiniop/ops/rms_norm/cuda/kernel.cuh b/src/infiniop/ops/rms_norm/cuda/kernel.cuh index 53c4a5587..86bd519e4 100644 --- a/src/infiniop/ops/rms_norm/cuda/kernel.cuh +++ b/src/infiniop/ops/rms_norm/cuda/kernel.cuh @@ -22,7 +22,7 @@ __device__ void rmsnormBlock( // Thread_0 computes RMS=1/sqrt(ss/dim+epsilon) and stores in shared memory __shared__ Tcompute rms; if (threadIdx.x == 0) { - rms = Tdata(rsqrtf(ss / Tcompute(dim) + epsilon)); + rms = Tcompute(rsqrtf(ss / Tcompute(dim) + epsilon)); } __syncthreads();