diff --git a/paddle/operators/iou_similarity_op.cu b/paddle/operators/iou_similarity_op.cu
index c270167e117defc4131a2b40a7cf17b30b583ca5..526edaefeea98b0760832c6b24fcb2d38b3aa00e 100755
--- a/paddle/operators/iou_similarity_op.cu
+++ b/paddle/operators/iou_similarity_op.cu
@@ -18,4 +18,5 @@ limitations under the License. */
 namespace ops = paddle::operators;
 REGISTER_OP_CUDA_KERNEL(
     iou_similarity,
-    ops::IOUSimilarityKernel<paddle::platform::CUDADeviceContext, float>);
+    ops::IOUSimilarityKernel<paddle::platform::CUDADeviceContext, float>,
+    ops::IOUSimilarityKernel<paddle::platform::CUDADeviceContext, double>);
diff --git a/paddle/operators/iou_similarity_op.h b/paddle/operators/iou_similarity_op.h
index 45bef2f7eece369fcc2e755c386dafd2504222f8..fa1b60ae34eeee2cf534607606568f527f1bd67b 100644
--- a/paddle/operators/iou_similarity_op.h
+++ b/paddle/operators/iou_similarity_op.h
@@ -46,7 +46,7 @@ struct IOUSimilarityFunctor {
     T y_min1 = x_[row_id * 4 + 1];
     T x_max1 = x_[row_id * 4 + 2];
     T y_max1 = x_[row_id * 4 + 3];
-    for (int i = 0; i < cols_; ++i) {
+    for (size_t i = 0; i < cols_; ++i) {
       T x_min2 = y_[i * 4];
       T y_min2 = y_[i * 4 + 1];
       T x_max2 = y_[i * 4 + 2];