diff --git a/tests/operators/test_conv2d.py b/tests/operators/test_conv2d.py
index b7a88f8bf..fa0ec1897 100644
--- a/tests/operators/test_conv2d.py
+++ b/tests/operators/test_conv2d.py
@@ -52,6 +52,7 @@ def torch_conv2d(
         [1, 32, 32, 32, 64, 1, 1],  # kernel 1,
     ],
 )
+@pytest.mark.parametrize("padding", [[0, 0, 0], [1, 1, 0]])
 @pytest.mark.parametrize("groups", [1, 2, 4])
 @pytest.mark.parametrize("stride", [[1, 1], [2, 3]])
 @pytest.mark.parametrize("dilations", [[1, 1], [2, 3]])
@@ -59,16 +60,18 @@ def torch_conv2d(
 @pytest.mark.parametrize(
     "device", ["cuda"]
 )  # we don't test for cpu because its quite imprecise in fp16 for larger kernel sizes
-def test_conv2d_gemm_fp16(n, c, h, w, oc, kx, ky, groups, stride, dilations, parallel_k, device):
+def test_conv2d_gemm_fp16(n, c, h, w, oc, kx, ky, padding, groups, stride, dilations, parallel_k, device):
     tol = 0.8
+    padh, padw, padc = padding
     check_binary(
         a_shape=[n, c, h, w],
         b_shape=[oc, c // groups, kx, ky],
-        numpy_op=lambda data, weight: torch_conv2d(data, weight, [0, 0], stride, dilations, groups),
+        numpy_op=lambda data, weight: torch_conv2d(data, weight, [padh, padw], stride, dilations, groups),
         hidet_op=lambda data, weight: ops.transpose(
             ops.conv2d_gemm_fp16_channel_last(
                 ops.transpose(data, [0, 2, 3, 1]),
                 weight,
+                padding=padding,
                 stride=stride,
                 dilations=dilations,
                 groups=groups,