diff --git a/tests/python/test_ndarray.py b/tests/python/test_ndarray.py
index acf000aeb5449..8579c04b4acf0 100644
--- a/tests/python/test_ndarray.py
+++ b/tests/python/test_ndarray.py
@@ -84,8 +84,11 @@ def test_matrix_ndarray(n, m, dtype, shape):
 
 
 @pytest.mark.parametrize('dtype', [ti.f32, ti.f64])
+@test_utils.test(arch=supported_archs_taichi_ndarray)
 def test_default_fp_ndarray(dtype):
-    ti.init(arch=supported_archs_taichi_ndarray, default_fp=dtype)
+    arch = ti.lang.impl.current_cfg().arch
+    ti.reset()
+    ti.init(arch=arch, default_fp=dtype)
 
     x = ti.Vector.ndarray(2, float, ())
 
@@ -93,8 +96,11 @@ def test_default_fp_ndarray(dtype):
 
 
 @pytest.mark.parametrize('dtype', [ti.i32, ti.i64])
+@test_utils.test(arch=supported_archs_taichi_ndarray)
 def test_default_ip_ndarray(dtype):
-    ti.init(arch=supported_archs_taichi_ndarray, default_ip=dtype)
+    arch = ti.lang.impl.current_cfg().arch
+    ti.reset()
+    ti.init(arch=arch, default_ip=dtype)
 
     x = ti.Vector.ndarray(2, int, ())
 
@@ -254,8 +260,8 @@ def _test_ndarray_deepcopy():
     assert y[4][1, 0] == 9
 
 
+@test_utils.test(arch=[ti.cuda], ndarray_use_cached_allocator=True)
 def test_ndarray_cuda_caching_allocator():
-    ti.init(arch=ti.cuda, ndarray_use_cached_allocator=True)
     n = 8
     a = ti.ndarray(ti.i32, shape=(n))
     a.fill(2)