Cublas initialization error while making engine

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

**• Hardware Platform (Jetson / GPU) 2080Ti
**• DeepStream Version 5.0.1
**• TensorRT Version 7.2.3
**• NVIDIA GPU Driver Version 450.102.04

I am getting the following error while making the engine file from onnx file. I have Cuda 10.2 with cudnn 8.1.1.33-1.

ERROR: …/nvdsinfer/nvdsinfer_func_utils.cpp:33 [TRT]: …/rtSafe/cublas/cublasWrapper.cpp (71) - cuBLAS Error in CublasWrapper: 14 (Could not initialize cublas. Please check CUDA installation.)
ERROR: …/nvdsinfer/nvdsinfer_func_utils.cpp:33 [TRT]: …/rtSafe/cublas/cublasWrapper.cpp (71) - cuBLAS Error in CublasWrapper: 14 (Could not initialize cublas. Please check CUDA installation.)

Even if I make the engine file manually by trtexec command , same error shows while reading the engine file.

3

Hey, the issue seems related to your setup, can you follow Quickstart Guide — DeepStream 6.1.1 Release documentation

5

I get a very similar error message when trying to run deepstream-segmentation-app from the docker image nvcr.io/nvidia/deepstream:5.1-21.02-devel . Shouldn`t this image have a correct setup with all dependencies?

ERROR: …/nvdsinfer/nvdsinfer_func_utils.cpp:33 [TRT]: safeContext.cpp (124) - Cudnn Error in initializeCommonContext: 4 (Could not initialize cudnn, please check cudnn installation.)

EDIT: I see now that Deepstream 5.1 requires an Ampere based GPU. That can explain why it throws this error.