Invalid argument when calling tensorrt inference

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

Description

getting cuda error when calling enqueueV2

E0630 18:57:11.882186 32283 TrtLogger.h:26] …/rtSafe/cuda/genericReformat.cu (1294) - Cuda Error in executeMemcpy: 1 (invalid argument)
E0630 18:57:11.882323 32283 TrtLogger.h:26] FAILED_EXECUTION: std::exception

happens also when using dynamic shapes and static shape engines

the buffers are both in the gpu.

can you please let me know what can cause this error? is it related to the input and output tensors not compatible with the network?

Thanks

6

Hi @alon.pukach,

Following similar issue may help you. If you still face this issue, please share us issue repro script, for better debugging.

Thank you.