AigsEffectsApp fails while loading TRTModel

AI & Data Science Deep Learning (Training & Inference) Maxine
1

Hello,

I’m using NVIDIA_VFX_SDK_Ubuntu18.04_0.6.5.0.tgz on Ubuntu-20.04 with NVIDIA GeForce RTX 3050 Ti Laptop GPU.

I’ve also installed

  • cuda-11-3 (driver is 510)
  • cudnn-11.3-linux-x64-v8.2.1.32.tgz
  • TensorRT-8.0.1.6.Linux.x86_64-gnu.cuda-11.3.cudnn8.2.tar.gz.

I am able to build the VideoFX samples and run them.

nolan@nolan:/usr/local/VideoFX/share$ LD_LIBRARY_PATH=/usr/local/TensorRT-8.0.1.6/lib ./AigsEffectApp --in_file=~/nolan_testing123.mp4 --out_file=~/nolan_testing123.aigs.mp4 --model_dir=/usr/local/VideoFX/lib/models --progress
Processing ~/nolan_testing123.mp4 mode 0 models /usr/local/VideoFX/lib/models
ERROR: Loading TRTModel failed in Aigs
Error loading the model 
Error: There is no model with dependencies that match this system

Please don’t say to go use the container! I’m actually writing another app that needs to use this functionality.

Thank you,
Nolan

2

I should correct the above. I’m on Ubuntu 21.10.

I’ve also tried various other combinations such as cuda-11-6, TensorRT-8.4.0.6.Linux.x86_64-gnu.cuda-11.6.cudnn8.3.tar.gz, cudnn-linux-x86_64-8.4.0.27_cuda11.6-archive.tar.xz.

Any help would be appreciated.

Thank you,
Nolan

3

Hi @user64586,
Did you solve the problem? I am facing the same issue.

Thanks

4

Hello l.miklosko

Unfortunately not. I stuck to Ubuntu 20.04. I was also not able to use Ubuntu 22.04.

It seems that there is a very hard requirement on the exact distribution version to be used, while using the libraries. There are some newer updates that are worth trying. Let me know.

Thanks,
Nolan

5

Have the same issue.

I have Ubuntu 20.04.5 LTS and am trying SuperRes.

OpenCV: FFMPEG: tag 0x34363248/'H264' is not supported with codec id 27 and format 'mp4 / MP4 (MPEG-4 Part 14)'
OpenCV: FFMPEG: fallback to use tag 0x31637661/'avc1'
ERROR: 1: [runtime.cpp::parsePlan::314] Error Code 1: Serialization (Serialization assertion plan->header.magicTag == rt::kPLAN_MAGIC_TAG failed.)
ERROR: Loading TRTModel failed in Super Res
Error: There was a problem deserializing the inference runtime engine

@l.miklosko @user64586 did you have any progress on this?

6

Unfortunately not. I followed all the memos, although had to alter the installation process as I was using Mint - but guessed it would be the same as it uses Ubuntu repositories anyway. Strangely enough, I was able to run it in Windows 11 without much trouble. What a strange times we live in.

7

Yeah. I ended up asking my friend to run it on Windows.

8

i have the same trouble, my envs is CUDA 12.0 TensorRT 8.6.0.12 cudnn is 8.9.0 , and my exec can run in windows 11 sucessfully ,my program is run by 3rd party and i create the same env in linux ,but i just get error Serialization (Serialization assertion plan->header.magicTag == rt::kPLAN_MAGIC_TAG failed.) ,Can somebody help plz!!!