Memory error for kFLOAT optimization.

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX2
1

Hi!
I am getting following error when I try to optimize the network with kFLOAT precision. I have tried increasing workspace size to maximum value of size_t (1<<31) but still got the same error. Can anybody tell me how I can fix this?

ERROR: Internal error: could not find any implementation for node conv1, try increasing the workspace size with IBuilder::setMaxWorkspaceSize()
ERROR: cudnnBuilder2.cpp (452) - OutOfMemory Error in buildSingleLayer
python: ctp.cpp:71: void caffeToGIEModel(const char*, const char*, const char*, uint16_t): Assertion `engine' failed.
Aborted (core dumped)

Thanks.

2

Hi,

Please remember that it is required to leave Half2Mode=false(default) if you want to inference a model with kFLOAT precision.
Thanks.

3

Thanks for your reply. I figured it out by hit and trial. kindly mention such things properly in documentation.