paddleOCR检测识别推理服务在没有显卡的机器上，无法在CPU模式下运行。

[dtiny]

请提供下述完整信息以便快速定位问题/Please provide the following information to quickly locate the problem

问题描述：执行文件是在有显卡的ubuntu16.04上编译的，可以在CPU和GPU两种模式下正常运行，没有问题；如果将程序放到另一台没有显卡的机器上，在CPU模式下会出错，报无法找到显卡。仍然会执行phi::backends::gpu::GetGPUDeviceCount()报错。之前的版本好像还没有这种问题。具体的错误信息见<- 完整报错>。

• 系统环境/System Environment：ubuntu16.04
• 版本号/Version：Paddle： PaddleOCR： 问题相关组件/Related components：推理库(2.4.2 版本)，GPU | 是 | MKL | 5.4 | CUDA11.2/cuDNN8.2/TensorRT8.0 | paddle_inference.tgz
• 运行指令/Command Code：
• 运行语言：c++
• 完整报错/Complete Error Message：
  WARNING: Logging before InitGoogleLogging() is written to STDERR
  W0404 13:53:59.584630 4046 init.cc:179] Compiled with WITH_GPU, but no GPU found in runtime.
  I0404 13:53:59.617218 4046 analysis_predictor.cc:964] MKLDNN is enabled
  --- Running analysis [ir_graph_build_pass]
  terminate called after throwing an instance of 'phi::enforce::EnforceNotMet'
  what():

---

C++ Traceback (most recent call last):

0 paddle_infer::CreatePredictor(paddle::AnalysisConfig const&)
1 paddle_infer::Predictor::Predictor(paddle::AnalysisConfig const&)
2 std::unique_ptr<paddle::PaddlePredictor, std::default_deletepaddle::PaddlePredictor > paddle::CreatePaddlePredictor<paddle::AnalysisConfig, (paddle::PaddleEngineKind)2>(paddle::AnalysisConfig const&)
3 paddle::AnalysisPredictor::Init(std::shared_ptrpaddle::framework::Scope const&, std::shared_ptrpaddle::framework::ProgramDesc const&)
4 paddle::AnalysisPredictor::PrepareProgram(std::shared_ptrpaddle::framework::ProgramDesc const&)
5 paddle::AnalysisPredictor::OptimizeInferenceProgram()
6 paddle::inference::analysis::Analyzer::RunAnalysis(paddle::inference::analysis::Argument*)
7 paddle::inference::analysis::IrGraphBuildPass::RunImpl(paddle::inference::analysis::Argument*)
8 paddle::inference::analysis::IrGraphBuildPass::LoadModel(std::string const&, std::string const&, paddle::framework::Scope*, phi::Place const&, bool)
9 paddle::inference::Load(paddle::framework::Executor*, paddle::framework::Scope*, std::string const&, std::string const&)
10 paddle::inference::LoadPersistables(paddle::framework::Executor*, paddle::framework::Scope*, paddle::framework::ProgramDesc const&, std::string const&, std::string const&, bool)
11 paddle::framework::Executor::Run(paddle::framework::ProgramDesc const&, paddle::framework::Scope*, int, bool, bool, std::vector<std::string, std::allocator<std::string > > const&, bool, bool)
12 paddle::framework::Executor::RunPreparedContext(paddle::framework::ExecutorPrepareContext*, paddle::framework::Scope*, bool, bool, bool)
13 paddle::framework::Executor::RunPartialPreparedContext(paddle::framework::ExecutorPrepareContext*, paddle::framework::Scope*, long, long, bool, bool, bool)
14 paddle::framework::OperatorBase::Run(paddle::framework::Scope const&, phi::Place const&)
15 paddle::framework::OperatorWithKernel::RunImpl(paddle::framework::Scope const&, phi::Place const&) const
16 paddle::framework::OperatorWithKernel::RunImpl(paddle::framework::Scope const&, phi::Place const&, paddle::framework::RuntimeContext*) const
17 paddle::platform::DeviceContextPool::Get(phi::Place const&)
18 std::__future_base::_Deferred_state<std::thread::_Invoker<std::tuple<paddle::platform::EmplaceDeviceContextpaddle::platform::MKLDNNDeviceContext(std::map<phi::Place, std::shared_future<std::unique_ptr<phi::DeviceContext, std::default_deletephi::DeviceContext > >, std::lessphi::Place, std::allocator<std::pair<phi::Place const, std::shared_future<std::unique_ptr<phi::DeviceContext, std::default_deletephi::DeviceContext > > > > >, phi::Place)::{lambda()#1}> >, std::unique_ptr<phi::DeviceContext, std::default_deletephi::DeviceContext > >::_M_complete_async()
19 std::_Function_handler<std::unique_ptr<std::__future_base::_Result_base, std::__future_base::_Result_base::_Deleter> (), std::__future_base::_Task_setter<std::unique_ptr<std::__future_base::_Result<std::unique_ptr<phi::DeviceContext, std::default_deletephi::DeviceContext > >, std::__future_base::_Result_base::_Deleter>, std::thread::_Invoker<std::tuple<paddle::platform::EmplaceDeviceContextpaddle::platform::MKLDNNDeviceContext(std::map<phi::Place, std::shared_future<std::unique_ptr<phi::DeviceContext, std::default_deletephi::DeviceContext > >, std::lessphi::Place, std::allocator<std::pair<phi::Place const, std::shared_future<std::unique_ptr<phi::DeviceContext, std::default_deletephi::DeviceContext > > > > >, phi::Place)::{lambda()#1}> >, std::unique_ptr<phi::DeviceContext, std::default_deletephi::DeviceContext > > >::_M_invoke(std::_Any_data const&)
20 paddle::platform::EmplaceDeviceContextpaddle::platform::MKLDNNDeviceContext(std::map<phi::Place, std::shared_future<std::unique_ptr<phi::DeviceContext, std::default_deletephi::DeviceContext > >, std::lessphi::Place, std::allocator<std::pair<phi::Place const, std::shared_future<std::unique_ptr<phi::DeviceContext, std::default_deletephi::DeviceContext > > > > >, phi::Place)::{lambda()#1}::operator()() const
21 paddle::memory::allocation::AllocatorFacade::Instance()
22 paddle::memory::allocation::AllocatorFacade::AllocatorFacade()
23 paddle::memory::allocation::AllocatorFacadePrivate::AllocatorFacadePrivate(bool)
24 phi::backends::gpu::GetGPUDeviceCount()
25 phi::enforce::EnforceNotMet::EnforceNotMet(phi::ErrorSummary const&, char const, int)
26 phi::enforce::GetCurrentTraceBackStringabi:cxx11

---

Error Message Summary:

ExternalError: CUDA error(999), unknown error.
[Hint: Please search for the error code(999) on website (https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__TYPES.html#group__CUDART__TYPES_1g3f51e3575c2178246db0a94a430e0038) to get Nvidia's official solution and advice about CUDA Error.] (at /paddle/paddle/phi/backends/gpu/cuda/cuda_info.cc:66)
[operator < load_combine > error]
已放弃 (核心已转储)

[andyjiang1116]

可以安装对应的cpu版本paddle

[dtiny]

这个应该是paddle2.4版本推理库的问题，在CPU模式下仍然去调用GPU的相关函数，不合理。
如果切换到2.3版本的GPU库下使用CPU模式就没有这样的问题，希望2.4以后得版本可以修复这个问题。

[andyjiang1116]

因为你是在有gpu的机器上编译的GPU版本，所以在纯cpu机器上可能会有点问题，推荐在cpu机器上安装cpu版本的paddle

[UnstoppableCurry]

要改Cmakelist 代码，编译动态库自动去找GPU版本依赖肯定报错，或者你参数设置错了