Circumvent findFastestTactic

Is there a way to have TensorRT not try to findFastestTactic? This is useful for benchmarking a custom layer in isolation… The findFastestLayer returns 700 (an illegal memory access was encountered)…

Thanks.

I don’t think currently there is such support to have TensorRT not try to findFastestTactic.
But in TRT 7.1 EA (NVIDIA Jetson Linux for Tegra users) we have algorithm selection.
Algorithm selection provides a mechanism to select and report algorithms for different layers in a network.
Please refer to below link for more details:
https://docs.nvidia.com/deeplearning/tensorrt/developer-guide/index.html#algorithm-select
https://docs.nvidia.com/deeplearning/tensorrt/archives/tensorrt-710-ea/release-notes/tensorrt-7.html#rel_7-1-0-EA

Thanks