Failed to build TensorRT engine

I have been using the Audio2Face app on Omniverse and it is working smoothly, except for one issue. I get the error message “Failed to build TensorRT engine” when I try to animate the face model with an audio track. The face model loads fine, but there is no movement when the audio plays.

RTX 3050 laptop gpu
Audio2Face 2023.1.1

@pingligamers2828 if you could provide your latest Audio2Face log file, that’d be great. the log file can be found here:

  • Windows - C:\Users\<USER>\.nvidia-omniverse\logs\Kit\Audio2Face (2023.1 subfolder)
  • Linux - ~/.nvidia-omniverse/logs/Kit/Audio2Face

in the meantime, can you check to see if you are using the recommended version of graphics driver specified on this page?

1 Like

kit_20230813_205353.log (1.4 MB)
studio driver version 528.24

i am sure the dev would appreciate the upload! as a layman, it may be related to VRAM bottleneck happening:

2023-08-13 18:01:49 [475,868ms] [Info] [omni.audio2face.utility.scripts.utils] [TRTEXEC] [08/13/2023-21:01:49] [E] Error[1]: [virtualMemoryBuffer.cpp::nvinfer1::StdVirtualMemoryBufferImpl::resizePhysical::132] Error Code 1: Cuda Driver (invalid argument)
2023-08-13 18:01:49 [475,868ms] [Info] [omni.audio2face.utility.scripts.utils] [TRTEXEC] [08/13/2023-21:01:49] [W] [TRT] Requested amount of GPU memory (1024 bytes) could not be allocated. There may not be enough free memory for allocation to succeed.
2023-08-13 18:01:49 [475,886ms] [Info] [omni.audio2face.utility.scripts.utils] [TRTEXEC] [08/13/2023-21:01:49] [E] Error[2]: [globWriter.cpp::nvinfer1::builder::HybridGlobWriter::HybridGlobWriter::397] Error Code 2: OutOfMemory (no further information)
2023-08-13 18:01:49 [475,886ms] [Info] [omni.audio2face.utility.scripts.utils] [TRTEXEC] [08/13/2023-21:01:49] [E] Error[2]: [builder.cpp::nvinfer1::builder::Builder::buildSerializedNetwork::751] Error Code 2: Internal Error (Assertion engine != nullptr failed. )
2023-08-13 18:01:49 [475,886ms] [Info] [omni.audio2face.utility.scripts.utils] [TRTEXEC] [08/13/2023-21:01:49] [E] Engine could not be created from network
2023-08-13 18:01:49 [475,886ms] [Info] [omni.audio2face.utility.scripts.utils] [TRTEXEC] [08/13/2023-21:01:49] [E] Building engine failed
2023-08-13 18:01:49 [475,886ms] [Info] [omni.audio2face.utility.scripts.utils] [TRTEXEC] [08/13/2023-21:01:49] [E] Failed to create engine from model or file.
2023-08-13 18:01:49 [475,886ms] [Info] [omni.audio2face.utility.scripts.utils] [TRTEXEC] [08/13/2023-21:01:49] [E] Engine set up failed
2023-08-13 18:01:49 [475,886ms] [Info] [omni.audio2face.utility.scripts.utils] [TRTEXEC] &&&& FAILED TensorRT.trtexec [TensorRT v8501] # c:\users\delll\appdata\local\ov\pkg\audio2face-2023.1.1\exts\omni.audio2face.inference_deps\deps\audio2face-inference-deps\bin\trtexec.exe --onnx=c:\users\delll\appdata\local\ov\pkg\audio2face-2023.1.1\exts\omni.audio2face.core_deps\deps\audio2face-nets\nets_fullface\mark_v2\network.onnx --saveEngine=C:\Users\Delll\AppData\Local\ov\cache\Kit\105.0\ed961c5c\audio2face-nets\mark_v2\network_device0_fp32.trt --device=0 --tacticSources=-CUDNN --explicitBatch

but i’ll have to do some more digging to see whether there are other reasons that could’ve caused this. in the interim, here’s a breakdown of the recommended spec for both mobile and desktop workstation. some of the examples in A2F would require 5GB+ of VRAM from the getgo, so i could see the challenges you are experiencing on a 3050 laptop GPU.

1 Like

Hi @pingligamers2828 You seem to have 2 GPUs. Could you please disable the 2nd one from Device Manager and try again?

I disabled the second GPU through Device Manager and tried again, but my issue still persists.
kit_20230821_065305.log (1.1 MB)

Thank you, @Simplychenable , for helping me.

Oh it seems you don’t have enough VRAM:
2023-08-21 03:54:58 [113,707ms] [Info] [] Insufficient VRAM, skipping TextureLoader post-processing

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.