Tensorrt: There is a conflict between the simultaneous launch of the tensorrt engine and the pytorch model

[guanshuicheng]

Description
In flask framework, when I started both TRT engine and pytorch model, the first call was normal, and the second call was abnormal, the error message was as follows:
[TensorRT] ERROR: ../rtSafe/safeContext.cpp (133) - Cudnn Error in configure: 7 (CUDNN_STATUS_MAPPING_ERROR)
[TensorRT] ERROR: FAILED_EXECUTION: std::exception
The subsequent calls are exceptions
Environment
TensorRT Version: 7.0.0.11
GPU Type: 1080ti
Nvidia Driver Version: 410.72
CUDA Version: 10.0
CUDNN Version: 7.6.5
Operating System + Version: ubuntu16.04
Python Version (if applicable): python3.5.2
PyTorch Version (if applicable): 1.1.0

Does any body have a solution?

[guanshuicheng]

I've solved this problem by starting a separate thread when executing the pyroch model for reasoning