diff --git a/torchvision/csrc/cpu/PSROIAlign_cpu.cpp b/torchvision/csrc/cpu/PSROIAlign_cpu.cpp
index be5c2c4966f..e5eb051cb91 100644
--- a/torchvision/csrc/cpu/PSROIAlign_cpu.cpp
+++ b/torchvision/csrc/cpu/PSROIAlign_cpu.cpp
@@ -194,8 +194,6 @@ void bilinear_interpolate_gradient(
   // T val = (w1 * v1 + w2 * v2 + w3 * v3 + w4 * v4);
 
   w1 = hy * hx, w2 = hy * lx, w3 = ly * hx, w4 = ly * lx;
-
-  return;
 }
 
 template <class T>
diff --git a/torchvision/csrc/cpu/ROIAlign_cpu.cpp b/torchvision/csrc/cpu/ROIAlign_cpu.cpp
index 79460c54637..bbffc8189b9 100644
--- a/torchvision/csrc/cpu/ROIAlign_cpu.cpp
+++ b/torchvision/csrc/cpu/ROIAlign_cpu.cpp
@@ -269,8 +269,6 @@ void bilinear_interpolate_gradient(
   // T val = (w1 * v1 + w2 * v2 + w3 * v3 + w4 * v4);
 
   w1 = hy * hx, w2 = hy * lx, w3 = ly * hx, w4 = ly * lx;
-
-  return;
 }
 
 template <class T>
diff --git a/torchvision/csrc/cuda/PSROIAlign_cuda.cu b/torchvision/csrc/cuda/PSROIAlign_cuda.cu
index 6e92add3599..709b0bda208 100644
--- a/torchvision/csrc/cuda/PSROIAlign_cuda.cu
+++ b/torchvision/csrc/cuda/PSROIAlign_cuda.cu
@@ -192,8 +192,6 @@ __device__ void bilinear_interpolate_gradient(
   // T val = (w1 * v1 + w2 * v2 + w3 * v3 + w4 * v4);
 
   w1 = hy * hx, w2 = hy * lx, w3 = ly * hx, w4 = ly * lx;
-
-  return;
 }
 
 template <typename T>
diff --git a/torchvision/csrc/cuda/ROIAlign_cuda.cu b/torchvision/csrc/cuda/ROIAlign_cuda.cu
index c6ea208e458..59b94a780b1 100644
--- a/torchvision/csrc/cuda/ROIAlign_cuda.cu
+++ b/torchvision/csrc/cuda/ROIAlign_cuda.cu
@@ -194,8 +194,6 @@ __device__ void bilinear_interpolate_gradient(
   // T val = (w1 * v1 + w2 * v2 + w3 * v3 + w4 * v4);
 
   w1 = hy * hx, w2 = hy * lx, w3 = ly * hx, w4 = ly * lx;
-
-  return;
 }
 
 template <typename T>