3.3.1版本QueuedPaddleOcrAll使用Mkldnn推理时报错

[MooWeii]

Describe the bug

初始化使用的

FullOcrModel OcrModel = LocalFullModels.EnglishV5;
SysExt.OcrEngine = new QueuedPaddleOcrAll(() =>
{
    return new PaddleOcrAll(OcrModel, PaddleDevice.Mkldnn())
    {
        AllowRotateDetection = true, 
        Enable180Classification = true, 
        
    };
});

private async Task<PaddleOcrResult?> OcrCharPredict(Mat rawMat, bool isEnable)
{
    try
    {
        if (SysExt.OcrEngine != null && isEnable)
        {
            var predicted = await SysExt.OcrEngine.Run(rawMat);//只要一Run等待就报错
            return predicted;
        }
        else
        {
            return null;
        }
    }
    catch (Exception ex)
    {
        return null;
    }
}

报错内容如下

引发的异常:“System.InvalidOperationException”(位于 System.Private.CoreLib.dll 中)
Running predictor failed with Paddle status code 2: 

  Compile Traceback (most recent call last):
    File "tools/export_model.py", line 74, in <module>
      main()
    File "tools/export_model.py", line 69, in main
      paddle.jit.save(model, save_path)
    File "<decorator-gen-60>", line 2, in save
      
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/wrapped_decorator.py", line 25, in __impl__
      return wrapped_func(*args, **kwargs)
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/dygraph/base.py", line 38, in __impl__
      return func(*args, **kwargs)
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/dygraph/jit.py", line 680, in save
      concrete_program = static_func.concrete_program
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 461, in concrete_program
      concrete_program, _ = self.get_concrete_program(*input_spec)
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 401, in get_concrete_program
      concrete_program, partial_program_layer = self._program_cache[cache_key]
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 684, in __getitem__
      self._caches[item] = self._build_once(item)
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 675, in _build_once
      class_instance=cache_key.class_instance)
    File "<decorator-gen-58>", line 2, in from_func_spec
      
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/wrapped_decorator.py", line 25, in __impl__
      return wrapped_func(*args, **kwargs)
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/dygraph/base.py", line 38, in __impl__
      return func(*args, **kwargs)
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/dygraph/dygraph_to_static/program_translator.py", line 625, in from_func_spec
      outputs = static_func(*inputs)
    File "/home/zhoujun20/dygraph/PaddleOCR_rc/ppocr/modeling/architectures/base_model.py", line 74, in forward
      x = self.backbone(x)
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/dygraph/layers.py", line 884, in __call__
      outputs = self.forward(*inputs, **kwargs)
    File "/home/zhoujun20/dygraph/PaddleOCR_rc/ppocr/modeling/backbones/rec_mobilenet_v3.py", line 134, in forward
      x = self.conv1(x)
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/dygraph/layers.py", line 884, in __call__
      outputs = self.forward(*inputs, **kwargs)
    File "/home/zhoujun20/dygraph/PaddleOCR_rc/ppocr/modeling/backbones/det_mobilenet_v3.py", line 180, in forward
      x = self.conv(x)
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/dygraph/layers.py", line 884, in __call__
      outputs = self.forward(*inputs, **kwargs)
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/nn/layer/conv.py", line 644, in forward
      use_cudnn=self._use_cudnn)
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/nn/functional/conv.py", line 144, in _conv_nd
      type=op_type, inputs=inputs, outputs=outputs, attrs=attrs)
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/layer_helper.py", line 43, in append_op
      return self.main_program.current_block().append_op(*args, **kwargs)
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/framework.py", line 2849, in append_op
      attrs=kwargs.get("attrs", None))
    File "/usr/local/python3.5.1/lib/python3.5/site-packages/paddle/fluid/framework.py", line 1934, in __init__
      for frame in traceback.extract_stack():

--------------------------------------
C++ Traceback (most recent call last):
--------------------------------------
0   0x00007FF80EB46360
1   PD_PredictorRun2
2   PD_ConfigXpuDeviceId
3   PD_PredictorRun
4   paddle_infer::UpdateDllFlag
5   paddle::framework::NaiveExecutor::Run
6   paddle::framework::ExecutionContext::OutputVar
7   paddle::framework::OperatorWithKernel::RunImpl
8   paddle::framework::OperatorWithKernel::RunImpl
9   phi::StridedCopyKernel<bool,phi::CPUContext>
10  phi::StridedCopyKernel<bool,phi::CPUContext>
11  phi::StridedCopyKernel<bool,phi::CPUContext>
12  phi::OneDNNContext::GetCachedObjectsNumber
13  phi::DeviceManager::DeviceManager
14  common::enforce::GetCurrentTraceBackString

----------------------
Error Message Summary:
----------------------
NotFoundError: OneDnnContext does not have the input Filter.
  [Hint: Expected it != inputs_name_.end(), but received it == inputs_name_.end().] (at D:\a\PaddleSharp\PaddleSharp\paddle-src\paddle\phi\backends\onednn\onednn_context.cc:345)
  [operator < fused_conv2d > error]

Steps to reproduce the bug

如上

Expected behavior

No response

Screenshots

No response

Release version

No response

IDE

No response

OS version

No response

Additional context

No response

[sdcb]

这是已知问题，上游的包也报错，我解决不了
workaround是你可以用GPU的包跑mkldnn，就没问题，很神奇