diff --git a/imperative/python/megengine/core/tensor/megbrain_graph.py b/imperative/python/megengine/core/tensor/megbrain_graph.py
index 37a863690bdd6cdb3e575cfefcae2576833a1054..5ad3391e64154cfcdfd038dca4bbc2f33b2c8fe2 100644
--- a/imperative/python/megengine/core/tensor/megbrain_graph.py
+++ b/imperative/python/megengine/core/tensor/megbrain_graph.py
@@ -12,6 +12,7 @@ import weakref
 from concurrent.futures import Future, ThreadPoolExecutor
 
 from .. import _imperative_rt
+from .._imperative_rt.ops import BackwardGraph
 from .._wrap import device as as_device
 from ..ops.builtin import OpDef
 from .core import OpBase, TensorBase, apply
@@ -131,6 +132,13 @@ def _(op: OpDef, *args: VarNode):
     return _wrap(outputs)
 
 
+@apply.register()
+def _(op: BackwardGraph, *args: VarNode):
+    assert args
+    graph = args[0].graph
+    return op.interpret(lambda op, args: apply(op, *args), graph.make_const, args)
+
+
 def input_callback(callback, *args, device=None, dtype=None, shape=None, graph=None):
     outputs = _imperative_rt.input_callback(
         callback, as_device(device).to_c(), dtype, shape, _unwrap(args), graph=graph
diff --git a/imperative/python/src/ops.cpp b/imperative/python/src/ops.cpp
index abedeeea0ac49f113d0b622a9a82b218ef090b73..c64537959f4675148e2aada54ce3459df28b3fcb 100644
--- a/imperative/python/src/ops.cpp
+++ b/imperative/python/src/ops.cpp
@@ -40,6 +40,18 @@ void init_ops(py::module m) {
                 attr.param.insert(attr.param.end(), s.begin(), s.end());
             });
 
+    py::class_<BackwardGraph, std::shared_ptr<BackwardGraph>, OpDef>(m, "BackwardGraph")
+        .def("interpret", [](BackwardGraph& self, py::object pyf, py::object pyc,
+                             const mgb::SmallVector<py::object>& inputs) {
+                auto f = [pyf](OpDef& op, const mgb::SmallVector<py::object>& inputs) {
+                    return py::cast<mgb::SmallVector<py::object>>(pyf(op.copy(), inputs));
+                };
+                auto c = [pyc](const TensorPtr& tensor) {
+                    return pyc(tensor->dev_tensor());
+                };
+                return self.graph().interpret<py::object>(f, c, inputs);
+            });
+
     py::class_<GetVarShape, std::shared_ptr<GetVarShape>, OpDef>(m, "GetVarShape")
         .def(py::init());
 
@@ -98,7 +110,6 @@ void init_ops(py::module m) {
         .def(py::init<>())
         .def_readwrite("offsets", &ParamPackConcat::offsets);
 
-    py::class_<BackwardGraph, std::shared_ptr<BackwardGraph>, OpDef>(m, "BackwardGraph");
     py::class_<CondTake, std::shared_ptr<CondTake>, OpDef>(m, "CondTake")
         .def(py::init<>());
 
diff --git a/imperative/src/impl/ops/backward_graph.cpp b/imperative/src/impl/ops/backward_graph.cpp
index e844d481ad0b8a2274c460b7d59e7a5ec8f511a7..e452ac5902f04f9679756afb3501eee471fbc2f2 100644
--- a/imperative/src/impl/ops/backward_graph.cpp
+++ b/imperative/src/impl/ops/backward_graph.cpp
@@ -18,34 +18,10 @@ namespace imperative {
 SmallVector<TensorPtr>
 BackwardGraph::InternalGraph::apply(
         const SmallVector<TensorPtr>& inputs) const {
-    ThinHashMap<size_t, TensorPtr> node2tensor;
-    auto&& input_nodes = this->inputs;
-    mgb_assert(inputs.size() == input_nodes.size());
-    for (size_t i = 0; i < inputs.size(); ++ i) {
-        node2tensor[input_nodes[i]] = inputs[i];
-    }
-    for (auto &&i : constants) {
-        node2tensor[i.first] = i.second;
-    }
-    for (size_t i = 0; i < exprs.size(); ++ i) {
-        auto&& expr = exprs[i];
-        SmallVector<TensorPtr> inputs;
-        for (auto &&in : std::get<1>(expr)) {
-            inputs.push_back(node2tensor.at(in));
-        }
-        auto outputs = OpDef::apply_on_physical_tensor(
-                *std::get<0>(expr), inputs);
-        auto output_nodes = std::get<2>(expr);
-        mgb_assert(outputs.size() == output_nodes.size());
-        for (size_t i = 0; i < outputs.size(); ++ i) {
-            node2tensor[output_nodes[i]] = outputs[i];
-        }
-    }
-    SmallVector<TensorPtr> ret;
-    for (auto &&i : outputs) {
-        ret.push_back(node2tensor.at(i));
-    }
-    return ret;
+    return interpret<TensorPtr>(
+        &OpDef::apply_on_physical_tensor,
+        [](const TensorPtr& x) {return x;},
+        inputs);
 }
 
 SmallVector<LogicalTensorDesc>
diff --git a/imperative/src/include/megbrain/imperative/ops/backward_graph.h b/imperative/src/include/megbrain/imperative/ops/backward_graph.h
index 3a8f00bf69863627afdb39dec8de9d954797a18c..e1d8876828f598587e4256b23107a1235ea2df92 100644
--- a/imperative/src/include/megbrain/imperative/ops/backward_graph.h
+++ b/imperative/src/include/megbrain/imperative/ops/backward_graph.h
@@ -40,6 +40,37 @@ public:
 
         SmallVector<LogicalTensorDesc>
         infer_attrs(const SmallVector<LogicalTensorDesc>& inputs) const;
+
+        template <typename T, typename F, typename C>
+        SmallVector<T> interpret(F&& f, C&& c, const SmallVector<T>& inputs) const {
+            ThinHashMap<size_t, T> node2tensor;
+            auto&& input_nodes = this->inputs;
+            mgb_assert(inputs.size() == input_nodes.size());
+            for (size_t i = 0; i < inputs.size(); ++ i) {
+                node2tensor[input_nodes[i]] = inputs[i];
+            }
+            for (auto &&i : constants) {
+                node2tensor[i.first] = c(i.second);
+            }
+            for (size_t i = 0; i < exprs.size(); ++ i) {
+                auto&& expr = exprs[i];
+                SmallVector<T> inputs;
+                for (auto &&in : std::get<1>(expr)) {
+                    inputs.push_back(node2tensor.at(in));
+                }
+                auto&& outputs = f(*std::get<0>(expr), std::move(inputs));
+                auto&& output_nodes = std::get<2>(expr);
+                mgb_assert(outputs.size() == output_nodes.size());
+                for (size_t i = 0; i < outputs.size(); ++ i) {
+                    node2tensor[output_nodes[i]] = std::move(outputs[i]);
+                }
+            }
+            SmallVector<T> ret;
+            for (auto &&i : outputs) {
+                ret.push_back(node2tensor.at(i));
+            }
+            return ret;
+        }
     };
 
     const InternalGraph& graph() const {