fluid.io.load_inference_model 载入多个模型的时候会报错Error Occurs, info:enforce version == 0U failed, 1015534361 != 0 Only version 0 is supported

[zyfo2]

我们有多个模型，需要做通用inference，需要载入多个模型
如果只load任意一个模型都是ok的
但如果load一个模型后，load另一个模型，如果其中一个模型用到lod_tensor，就会报错。
函数代码如下：
place = fluid.CPUPlace()
exe = fluid.Executor(place)
[inference_program, _, fetch_targets] = (
fluid.io.load_inference_model(dirname=model_path[0], executor=exe,
model_filename=model_path[1],
params_filename=params_path[1]))

错误如下：
Error Occurs, info:enforce version == 0U failed, 1015534361 != 0
Only version 0 is supported at [/paddle/paddle/fluid/framework/lod_tensor.cc:276]
PaddlePaddle Call Stacks:
0 0x7f3e15df5376p paddle::platform::EnforceNotMet::EnforceNotMet(std::exception_ptr::exception_ptr, char const*, int) + 486
1 0x7f3e16652872p paddle::framework::DeserializeFromStream(std::istream&, paddle::framework::LoDTensor*, paddle::platform::DeviceContext const&) + 1330
2 0x7f3e1645fe7ap paddle::operators::LoadCombineOp::RunImpl(paddle::framework::Scope const&, boost::variant<paddle::platform::CUDAPlace, paddle::platform::CPUPlace, paddle::platform::CUDAPinnedPlace, boost::detail::variant::void, boost::detail::variant::void, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_> const&) const + 778
3 0x7f3e1660e450p paddle::framework::OperatorBase::Run(paddle::framework::Scope const&, boost::variant<paddle::platform::CUDAPlace, paddle::platform::CPUPlace, paddle::platform::CUDAPinnedPlace, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_, boost::detail::variant::void_> const&) + 208
4 0x7f3e15e88cdfp paddle::framework::Executor::RunPreparedContext(paddle::framework::ExecutorPrepareContext*, paddle::framework::Scope*, bool, bool, bool) + 255
5 0x7f3e15e89d30p paddle::framework::Executor::Run(paddle::framework::ProgramDesc const&, paddle::framework::Scope*, int, bool, bool) + 128
6 0x7f3e15e0bfabp void pybind11::cpp_function::initialize<pybind11::cpp_function::initialize<void, paddle::framework::Executor, paddle::framework::ProgramDesc const&, paddle::framework::Scope*, int, bool, bool, pybind11::name, pybind11::is_method, pybind11::sibling>(void (paddle::framework::Executor::)(paddle::framework::ProgramDesc const&, paddle::framework::Scope, int, bool, bool), pybind11::name const&, pybind11::is_method const&, pybind11::sibling const&)::{lambda(paddle::framework::Executor*, paddle::framework::ProgramDesc const&, paddle::framework::Scope*, int, bool, bool)#1}, void, paddle::framework::Executor*, paddle::framework::ProgramDesc const&, paddle::framework::Scope*, int, bool, bool, pybind11::name, pybind11::is_method, pybind11::sibling>(pybind11::cpp_function::initialize<void, paddle::framework::Executor, paddle::framework::ProgramDesc const&, paddle::framework::Scope*, int, bool, bool, pybind11::name, pybind11::is_method, pybind11::sibling>(void (paddle::framework::Executor::)(paddle::framework::ProgramDesc const&, paddle::framework::Scope, int, bool, bool), pybind11::name const&, pybind11::is_method const&, pybind11::sibling const&)::{lambda(paddle::framework::Executor*, paddle::framework::ProgramDesc const&, paddle::framework::Scope*, int, bool, bool)#1}&&, void ()(paddle::framework::Executor, paddle::framework::ProgramDesc const&, paddle::framework::Scope*, int, bool, bool), pybind11::name const&, pybind11::is_method const&, pybind11::sibling const&)::{lambda(pybind11::detail::function_call&)#3}::_FUN(pybind11::detail::function_call) + 555
7 0x7f3e15e0446cp pybind11::cpp_function::dispatcher(_object*, _object*, _object*) + 2540
8 0x7f3efe641fc7p PyEval_EvalFrameEx + 28695
9 0x7f3efe6444e9p PyEval_EvalCodeEx + 2025
10 0x7f3efe6419b8p PyEval_EvalFrameEx + 27144
11 0x7f3efe6444e9p PyEval_EvalCodeEx + 2025
12 0x7f3efe6419b8p PyEval_EvalFrameEx + 27144
13 0x7f3efe6444e9p PyEval_EvalCodeEx + 2025
14 0x7f3efe6419b8p PyEval_EvalFrameEx + 27144
15 0x7f3efe6444e9p PyEval_EvalCodeEx + 2025
16 0x7f3efe6419b8p PyEval_EvalFrameEx + 27144
17 0x7f3efe6444e9p PyEval_EvalCodeEx + 2025
18 0x7f3efe6419b8p PyEval_EvalFrameEx + 27144
19 0x7f3efe642f9ep PyEval_EvalFrameEx + 32750
20 0x7f3efe642f9ep PyEval_EvalFrameEx + 32750
21 0x7f3efe642f9ep PyEval_EvalFrameEx + 32750
22 0x7f3efe6444e9p PyEval_EvalCodeEx + 2025
23 0x7f3efe5cd377p
24 0x7f3efe5a87a3p PyObject_Call + 67
25 0x7f3efe63d4bep PyEval_EvalFrameEx + 9486
26 0x7f3efe642f9ep PyEval_EvalFrameEx + 32750
27 0x7f3efe642f9ep PyEval_EvalFrameEx + 32750
28 0x7f3efe6444e9p PyEval_EvalCodeEx + 2025
29 0x7f3efe5cd28ap
30 0x7f3efe5a87a3p PyObject_Call + 67
31 0x7f3efe5b763dp
32 0x7f3efe5a87a3p PyObject_Call + 67
33 0x7f3efe63aa58p PyEval_CallObjectWithKeywords + 72
34 0x7f3efe673f36p
35 0x7f3efe34be25p
36 0x7f3efd96cbadp clone + 109

[zyfo2]

solved by myself. for those who need it:
use a new scope for every model

        scope = fluid.Scope()
        with fluid.scope_guard(scope):
            place = fluid.CPUPlace()
            exe = fluid.Executor(place)
            [inference_program, _, fetch_targets] = (
                fluid.io.load_inference_model(dirname=model_path[0], executor=exe,
                                          model_filename=model_path[1],
                                          params_filename=params_path[1]))

and for prediction:

       with fluid.scope_guard(scope):
            results = exe.run(inference_program,
                          feed=inputs,
                          fetch_list=fetch_targets)

[1138886114]

When I load two different models in a service, I will report an error: ` precondition notmeterror: tensor not initialized yet when tensor:: type() is called

[Hint: holder_ should not be null.] (at /paddle/paddle/fluid/framework/tensor.h:202)

[operator < conv2d > error]`

scope = fluid.Scope()
with fluid.scope_guard(scope):
    exe = fluid.Executor(fluid.CPUPlace())
    [inference_program, feed_target_names, fetch_targets] = (
        fluid.io.load_inference_model(dirname=path,
                                      executor=exe,
                                      model_filename='res152_model',
                                      params_filename='res152_params'))


with fluid.scope_guard(scope):
    res = exe.run(inference_program, feed={feed_target_names[0]: im_pre}, fetch_list=fetch_targets)


scope_label = fluid.Scope()
with fluid.scope_guard(scope_label):
    exe_label = fluid.Executor(fluid.CPUPlace())
    [inference_label, feed_target_label, fetch_label] = (
        fluid.io.load_inference_model(dirname=path,
                                      executor=exe_label,
                                      model_filename='res152_model2',
                                      params_filename='res152_params2'))

with fluid.scope_guard(scope_label):
    res_classlabel = exe_label.run(inference_label,feed={feed_target_label[0]: im_pre},fetch_list=fetch_label)