Merge pull request #821 from ngc92/out-of-bounds-bugfix

fix out-of-bounds access in rmsnorm kernel

Author: karpathy

llmc/rmsnorm.cuh

@@ -90,7 +90,7 @@ __global__ void fused_residual_rmsnorm_forward_kernel5(floatX* residual, floatX*
     __syncthreads();
 
     int idx = blockIdx.x * blockDim.y + threadIdx.y;
-    if(idx > N) return;
+    if(idx >= N) return;
 
     // adjust pointers to current token
     residual += C * idx;