From d10c98c1dc4b294726a8d304dfc1c721260567d3 Mon Sep 17 00:00:00 2001
From: rtp-llm <rtp-llm@alibaba>
Date: Sat, 20 Apr 2024 16:04:19 +0800
Subject: [PATCH] fix - fix cuda11 compile

---
 src/fastertransformer/kernels/quantization_tensor.cu | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/fastertransformer/kernels/quantization_tensor.cu b/src/fastertransformer/kernels/quantization_tensor.cu
index 39e3de88..6b29c538 100644
--- a/src/fastertransformer/kernels/quantization_tensor.cu
+++ b/src/fastertransformer/kernels/quantization_tensor.cu
@@ -95,7 +95,7 @@ __global__ void perTokenQuantization(
     T localMax = 1e-6f;
     for (int i = threadIdx.x; i < numCols; i += blockDim.x)
     {
-        localMax = cuda_max(localMax, cuda_abs((srcRow[i])/ cuda_cast<T>(smoother[i])));
+        localMax = cuda_max(localMax, cuda_abs(cuda_cast<T>((srcRow[i])/ cuda_cast<T>(smoother[i]))));
     }
     const float rowMax = blockAllReduceMax(cuda_cast<float>(localMax));