I have a python program in which I am using PyCUDA to pre-process some data using the GPU before the results are then fed into a CNN implemented using Tensorflow. The pre-processing function works fine on it’s own. But when I add the line to load the CNN model… self.model = load_model(model_path) …

PyCUDA 'invalid resource handle' when Tensorflow model is loaded

Robert_Crovella May 21, 2022, 1:50am 3

this may be of interest

Topic		Replies	Views
Issue with calling PyCuda function (LogicError: cuFuncSetBlockShape failed: invalid resource handle) CUDA Programming and Performance	3	2171	August 28, 2022
Failed call to cuInit: CUDA_ERROR_OUT_OF_MEMORY: out of memory Frameworks cuda , tensorflow	1	2925	April 22, 2021
pycuda._driver.LogicError: explicit_context_dependent failed: invalid device context - no currently active context? TensorRT	7	3933	October 22, 2018
Combine pycuda and cupy GPU-Accelerated Libraries pycuda , cupy	8	3995	September 20, 2021
"Cuda Error in NCHWTONCHHW2: 33 (invalid resource handle) "，How to solve it? Jetson Nano cuda	30	6030	October 18, 2021
Tensoflow error: Could not create cudnn handle: CUDNN_STATUS_NOT_INITIALIZED CUDA Developer Tools	0	1266	January 25, 2021
[ubuntu1404][GTX-1080] Cublas handle: not initialized in driver version 384.111 Linux	6	5131	October 14, 2021
pycuda._driver.LogicError: cuMemcpyDtoH failed: an illegal memory access was encountered CUDA Programming and Performance	7	6292	May 17, 2023
Cudnn PoolForward launch failed CUDA NVCC Compiler cuda , tensorflow , ubuntu , python	0	992	June 1, 2022
Invalid resource handle when doing inference with TensorRT TensorRT	3	2312	August 10, 2022