Failure in conversion for DLA

Hi, i’m seeing a conversion error when trying to convert to DLA. I already eliminated all layers that don’t run on DLA. Now I’m perplexed as to the origin of this error as the diagnostics displayed appear to give no information. I tried increasing WS size to 3000 on AGX Xavier.

/trtexec --verbose --onnx=file.onnx --fp16 --useDLACore=0 --workspace=3128

[07/19/2021-13:26:32] [W] [TRT] DLA Node compilation Failed.
[07/19/2021-13:26:32] [V] [TRT] Tactic: 548706303427 skipped. ProcessNode failure.
[07/19/2021-13:26:32] [V] [TRT] Fastest Tactic: -3360065831133338131 Time: 3.40282e+38
[07/19/2021-13:26:32] [E] [TRT] Try increasing the workspace size with IBuilderConfig::setMaxWorkspaceSize() if using IBuilder::buildEngineWithConfig, or IBuilder::setMaxWorkspaceSize() if using IBuilder::buildCudaEngine.

[07/19/2021-13:26:32] [V] [TRT] Builder timing cache: created 0 entries, 0 hit(s)
[07/19/2021-13:26:32] [E] [TRT] …/builder/tacticOptimizer.cpp (1715) - TRTInternal Error in computeCosts: 0 ()
[07/19/2021-13:26:32] [E] Engine creation failed
[07/19/2021-13:26:32] [E] Engine set up failed

Any suggestions?
Thank you!!!

Hi,
Please check the below links, as they might answer your concerns.
https://docs.nvidia.com/deeplearning/tensorrt/developer-guide/index.html#dla_topic
https://docs.nvidia.com/deeplearning/tensorrt/developer-guide/index.html#dla_layers
https://docs.nvidia.com/deeplearning/tensorrt/developer-guide/#restrictions-with-dla
Thanks!

Thank you for your response, what are you proposing try changing every single parameter in the network until it works?
The diagnostic message has no information or any clue as to what is actually wrong.
I can do a binary search over removing layers but it seems a bit icky.

Could you open source TensorRT so at least it’s possible to debug errors?

Hi @solaris.insight,

We recommend you to please post your concern here to get better help.

Thank you.