CUDA error 719: unspecified launch failure


I used TensorRT to package ResNet into a DLL that works fine when called by other software, but may crash after a day or two. CUDA Error 719: Unspecified launch failure, showing an error on line 164 of the source code.
The software would load five or six models, and then parallel inference was made on one device. Therefore, I used the writing method of structure to distinguish engine, context and stream of each model.Does anyone know what I wrote wrong?


TensorRT Version:
GPU Type: 1080Ti
Nvidia Driver Version: 441.22
CUDA Version: 10.2
CUDNN Version: 7.6.5
Operating System + Version: win10

Relevant Files

this is source code link:
issue/main.cpp at main · YadongLau/issue (

Please check the below links, as they might answer your concerns.


Hello, I am now sure that my network construction part is OK, because now the output results of running are correct, and the results of multi-model simultaneous inference are also correct. Now the problem is that the multi-model runs for a period of time (a few hours, It will crash after one day, or two days), the time of the crash is not fixed, can you take a look at the code I provided


Looks like you’re using an old version of the TensorRT.
Could you please try on the latest TensorRT version 8.4 GA and if you still face this issue, please share with us complete logs for better debugging.

Thank you.