Libnvos.so not found

AI & Data Science Deep Learning (Training & Inference) TensorRT
, ,
1

Description

I’m upgrading devices to jetpack 6.2

libnvos.so not found when loading models in triton server
model | 1 | UNAVAILABLE: Not found: unable to load shared library: libnvos.so: cannot open shared object file: No such file or directory
already I had issues with libnvdla_compiler.so but solved with this

root@67d8676:~# /opt/tritonserver/bin/tritonserver --model-repository=/models-management/models/
W0320 19:40:13.421139 232115 pinned_memory_manager.cc:273] “Unable to allocate pinned system memory, pinned memory pool will not be available: CUDA driver version is insufficient for CUDA runtime version”
I0320 19:40:13.421251 232115 cuda_memory_manager.cc:117] “CUDA memory pool disabled”
E0320 19:40:13.421364 232115 server.cc:241] “CudaDriverHelper has not been initialized.”
I0320 19:40:13.422100 232115 model_lifecycle.cc:472] “loading: model:1”
E0320 19:40:13.423092 232115 model_lifecycle.cc:642] “failed to load ‘detector’ version 1: Not found: unable to load shared library: libnvos.so: cannot open shared object file: No such file or directory”
I0320 19:40:13.423139 232115 model_lifecycle.cc:777] “failed to load ‘detector’”
I0320 19:40:13.423243 232115 server.cc:604]

Environment

I’m using base image to build a docker container on jetson using BalenaOS
nvcr.io/nvidia/tritonserver:24.08-py3-igpu

Now I’m having problems due to
Device is a Jetson Orin Nano running r36.4.3

Linux 67d8676 5.15.148-l4t-r36.4-1012.12 #1 SMP PREEMPT Thu Oct 10 15:09:40 UTC 2024 aarch64 aarch64 aarch64 GNU/Linux

Jetson Linux 36.4.3 is part of JetPack 6.2

CUDA 12.6.10

cuDNN 9.3.0

TensorRT 9.3.0

Any suggestions?

Also, when will Jetpack full image for 6.4.3 be available?

Thanks!

2

Hi @andye ,
Would request you to pls raise the concern to Jetson Forum.

Thanks