Merge branch 'eddiey-main-patch-12774' into 'main'

fix race in `cuApplyLayerNorm`

See merge request ADLR/megatron-lm!336

megatron/fused_kernels/layer_norm_cuda_kernel.cu

@@ -329,6 +329,7 @@ void cuApplyLayerNorm(
       mean[i1] = mu;
       invvar[i1] = c_invvar;
     }
+    __syncthreads();
   }
 }