Created by: chenwhql

In some cases, paddle will throw errors like holder_ should not be null. Tensor not initialized yet when Tensor::type() is called.. This error is caused by the direct use of the tensor in the Variable without initialization, but this error message is not user friendly.

By analyzing issues, I found that GetExceptedKernelType is an important source of this problem. Our developers get the tensor type of input variable but don't check if the tensor was initialized. So I think we can add a funciton to deal with this problem.

If Op doesn't overwrite the GetExceptedKernelType function, use the OperatorWithKernel::GetExpectedKernelType function, it wil check if tensor was initialized, so I add the IndicateVarDataType which is just did what we should have done.

Then the interface GetDataTypeOfVar can be removed, and we should add a rule that developers must use this IndicateVarDataType interface in Ops' GetExceptedKernelType.

The error cases:

https://github.com/PaddlePaddle/Paddle/issues/19522

https://github.com/PaddlePaddle/Paddle/issues/18432

This PR's results:

• original error message:

holder_ should not be null
Tensor not initialized yet when Tensor::type() is called. at [/paddle/paddle/fluid/framework/tensor.h:139]

• new error message:

The Tensor in the save Op's Input Variable X(mean_0.tmp_0) is not initialized. at [/work/paddle/paddle/fluid/framework/operator.cc:1162]
  [operator < save > error]

Use Cases

1. GetDataTypeOfVar(ctx.InputVar("X")) -> IndicateVarDataType(ctx, "X")

2. ctx.Input<Tensor>("X")->type() -> IndicateVarDataType(ctx, "X")

Test Case

Code

import paddle.fluid as fluid

place = fluid.CPUPlace()
exe = fluid.Executor(place)

x = fluid.layers.data(name='X', shape=[13], dtype='float32')
y = fluid.layers.data(name='Y', shape=[1], dtype='float32')
y_ = fluid.layers.fc(input=x, size=1, act=None)
loss = fluid.layers.square_error_cost(input=y_, label=y)
avg_loss = fluid.layers.mean(loss)
avg_loss.persistable = True

fluid.optimizer.SGD(learning_rate=0.01).minimize(avg_loss)

exe.run(fluid.default_startup_program())
fluid.io.save_inference_model(dirname="./save_op_test",
                              feeded_var_names=['X', 'Y'],
                              target_vars=[avg_loss],
                              executor=exe)

Result

--------------------------------------------
C++ Call Stacks (More useful to developers):
--------------------------------------------
0   std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > paddle::platform::GetTraceBackString<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&>(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, char const*, int)
1   paddle::platform::EnforceNotMet::EnforceNotMet(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, char const*, int)
2   paddle::framework::OperatorWithKernel::ParseInputDataType(paddle::framework::ExecutionContext const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, paddle::framework::proto::VarType_Type*) const
3   paddle::framework::OperatorWithKernel::IndicateVarDataType(paddle::framework::ExecutionContext const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) const
4   paddle::operators::SaveOp::GetExpectedKernelType(paddle::framework::ExecutionContext const&) const
5   paddle::framework::OperatorWithKernel::ChooseKernel(paddle::framework::RuntimeContext const&, 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
6   paddle::framework::OperatorWithKernel::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&, paddle::framework::RuntimeContext*) const
7   paddle::framework::OperatorWithKernel::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
8   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&)
9   paddle::framework::Executor::RunPreparedContext(paddle::framework::ExecutorPrepareContext*, paddle::framework::Scope*, bool, bool, bool)
10  paddle::framework::Executor::Run(paddle::framework::ProgramDesc const&, paddle::framework::Scope*, int, bool, bool, std::vector<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >, std::allocator<std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > > > const&, bool)

------------------------------------------
Python Call Stacks (More useful to users):
------------------------------------------
  File "/usr/local/lib/python3.5/dist-packages/paddle/fluid/framework.py", line 2423, in append_op
    attrs=kwargs.get("attrs", None))
  File "/usr/local/lib/python3.5/dist-packages/paddle/fluid/io.py", line 241, in save_vars
    attrs={'file_path': save_file_path})
  File "/usr/local/lib/python3.5/dist-packages/paddle/fluid/io.py", line 216, in save_vars
    filename=filename)
  File "/usr/local/lib/python3.5/dist-packages/paddle/fluid/io.py", line 571, in save_persistables
    filename=filename)
  File "/usr/local/lib/python3.5/dist-packages/paddle/fluid/io.py", line 1197, in save_inference_model
    save_persistables(executor, save_dirname, main_program, params_filename)
  File "save_op.py", line 22, in <module>
    executor=exe)

----------------------
Error Message Summary:
----------------------
PaddleCheckError: Expected t->IsInitialized() == true, but received t->IsInitialized():0 != true:1.
The Tensor in the save Op's Input Variable X(mean_0.tmp_0) is not initialized. at [/work/paddle/paddle/fluid/framework/operator.cc:1162]
  [operator < save > error]