diff --git a/imperative/python/megengine/core/tensor/megbrain_graph.py b/imperative/python/megengine/core/tensor/megbrain_graph.py
index 9b4707b7..5cddd773 100644
--- a/imperative/python/megengine/core/tensor/megbrain_graph.py
+++ b/imperative/python/megengine/core/tensor/megbrain_graph.py
@@ -274,7 +274,7 @@ def dump_graph(
     keep_var_name: int = 1,
     keep_param_name: bool = False,
     keep_opr_priority: bool = False,
-    strip_info_file=None,
+    strip_info_file=None
 ):
     """serialize the computing graph of `output_vars` and get byte result.
 
diff --git a/imperative/python/megengine/core/tensor/multipledispatch/conflict.py b/imperative/python/megengine/core/tensor/multipledispatch/conflict.py
index 6989755d..ec852aa7 100644
--- a/imperative/python/megengine/core/tensor/multipledispatch/conflict.py
+++ b/imperative/python/megengine/core/tensor/multipledispatch/conflict.py
@@ -40,6 +40,8 @@
 #  All Megvii Modifications are Copyright (C) 2014-2020 Megvii Inc. All rights reserved.
 # --------------------------------------------------------------------------------------
 
+from collections import OrderedDict
+
 from .utils import _toposort, groupby
 from .variadic import isvariadic
 
@@ -159,5 +161,5 @@ def ordering(signatures):
     for s in signatures:
         if s not in edges:
             edges[s] = []
-    edges = dict((k, [b for a, b in v]) for k, v in edges.items())
+    edges = OrderedDict((k, [b for a, b in v]) for k, v in edges.items())
     return _toposort(edges)
diff --git a/imperative/python/megengine/distributed/helper.py b/imperative/python/megengine/distributed/helper.py
index 81cfc77b..72a85f2a 100644
--- a/imperative/python/megengine/distributed/helper.py
+++ b/imperative/python/megengine/distributed/helper.py
@@ -54,14 +54,14 @@ def synchronized(func: Callable):
     return wrapper
 
 
-def get_device_count_by_fork(device_type: str):
-    q = mp.Queue()
+def worker(queue, device_type):
+    num = get_device_count(device_type)
+    queue.put(num)
 
-    def worker(queue):
-        num = get_device_count(device_type)
-        queue.put(num)
 
-    p = mp.Process(target=worker, args=(q,))
+def get_device_count_by_fork(device_type: str):
+    q = mp.Queue()
+    p = mp.Process(target=worker, args=(q, device_type))
     p.start()
     p.join()
     return q.get()
diff --git a/imperative/python/src/dispatcher.cpp b/imperative/python/src/dispatcher.cpp
index 2d2cd844..616e1791 100644
--- a/imperative/python/src/dispatcher.cpp
+++ b/imperative/python/src/dispatcher.cpp
@@ -151,16 +151,19 @@ struct Dispatcher {
 public:
     static constexpr auto tp_name = "Dispatcher";
 
-    PyObject* tp_vectorcall(PyObject*const* args, Py_ssize_t nargs) {
-        if (!prepare_call(args, nargs)) return nullptr;
-        return do_call([=](PyObject* func){return _PyObject_FastCall(func, const_cast<PyObject**>(args), nargs);});
-    }
-
     PyObject* tp_call(PyObject* args, PyObject* kwargs) {
         if (!prepare_call(&PyTuple_GET_ITEM(args, 0), PyTuple_GET_SIZE(args))) return nullptr;
         return do_call([=](PyObject* func){return PyObject_Call(func, args, kwargs);});
     }
 
+#if PY_MINOR_VERSION >= 6
+    PyObject* tp_vectorcall(PyObject*const* args, Py_ssize_t nargs) {
+        if (!prepare_call(args, nargs)) return nullptr;
+        return do_call([=](PyObject* func){return _PyObject_FastCall(func, const_cast<PyObject**>(args), nargs);});
+    }
+#endif
+
+#if PY_MINOR_VERSION >= 6
     PyObject* super(PyObject*const* args, Py_ssize_t nargs) {
         if (stack.empty()) {
             PyErr_SetString(PyExc_RuntimeError, "super called at top level");
@@ -169,6 +172,16 @@ public:
         stack.emplace_back_safely(stack.back()).mro_offset++;
         return do_call([=](PyObject* func){return _PyObject_FastCall(func, const_cast<PyObject**>(args), nargs);});
     }
+#else
+    PyObject* super(PyObject* args, PyObject* kwargs) {
+        if (stack.empty()) {
+            PyErr_SetString(PyExc_RuntimeError, "super called at top level");
+            return nullptr;
+        }
+        stack.emplace_back_safely(stack.back()).mro_offset++;
+        return do_call([=](PyObject* func){return PyObject_Call(func, args, kwargs);});
+    }
+#endif
 
     void enable(PyObject* func) {
         auto obj = py::reinterpret_borrow<py::object>(func);
@@ -204,7 +217,11 @@ void init_dispatcher(py::module m) {
         .def<&Dispatcher::enable>("enable")
         .def<&Dispatcher::disable>("disable")
         .def<&Dispatcher::clear_cache>("clear_cache")
+#if PY_MINOR_VERSION >= 6
         .def<&Dispatcher::tp_vectorcall>("call")
+#else
+        .def<&Dispatcher::tp_call>("call")
+#endif
         .def<&Dispatcher::super>("super")
         .finalize();
     if (!dispatcher_type) throw py::error_already_set();
diff --git a/imperative/python/test/run.sh b/imperative/python/test/run.sh
index 4920e2e9..825c1017 100755
--- a/imperative/python/test/run.sh
+++ b/imperative/python/test/run.sh
@@ -1,6 +1,7 @@
 #!/bin/bash -e
 
 test_dirs="test megengine"
+
 TEST_PLAT=$1
 
 if [[ "$TEST_PLAT" == cpu ]]; then
@@ -13,9 +14,9 @@ else
 fi
 
 pushd $(dirname "${BASH_SOURCE[0]}")/.. >/dev/null
-    PYTHONPATH="." python3 -m pytest $test_dirs -m 'not isolated_distributed'
+    PYTHONPATH="." PY_IGNORE_IMPORTMISMATCH=1 python3 -m pytest $test_dirs -m 'not isolated_distributed'
     if [[ "$TEST_PLAT" == cuda ]]; then
         echo "test GPU pytest now"
-        PYTHONPATH="." python3 -m pytest $test_dirs -m 'isolated_distributed'
+        PYTHONPATH="." PY_IGNORE_IMPORTMISMATCH=1 python3 -m pytest $test_dirs -m 'isolated_distributed'
     fi
 popd >/dev/null