Missing GLIBCXX_3.4.32 when starting riva speech server on jetson thor using 2.19.0

Hello,

Could you who help give me a advice about how to start RIVA speech server on local or docker? Thks!

Hardware: Jetson-thor

Software: jetson sdk 7.0

Software part of jetson-stats 4.3.2 - (c) 2024, Raffaello Bonghi
Jetpack missing!

  • Model: NVIDIA Jetson AGX Thor Developer Kit
  • L4T: 38.2.1
    NV Power Mode[0]: MAXN
    Serial Number: [XXX Show with: jetson_release -s XXX]
    Hardware:
  • P-Number: p3834-0008
  • Module: Not available
    Platform:
  • Distribution: Ubuntu 24.04 Noble Numbat
  • Release: 6.8.12-tegra
    jtop:
  • Version: 4.3.2
  • Service: Active
    Libraries:
  • CUDA: 13.0.48
  • cuDNN: 9.12.0.46
  • TensorRT: 10.13.3.9
  • VPI: 4.0.0~er5
  • Vulkan: 1.4.304
  • OpenCV: 4.8.0 - with CUDA: NO

The issue on my running riva speech server:

I followed the guide of riva_quickstart_arm64_2.19.0, but, the riva server can't be stated, and the log showed: 


  > Riva waiting for Triton server to load all models...retrying in 1 second
I1023 01:24:34.833186 53 pinned_memory_manager.cc:277] "Pinned memory pool is created at '0x207e00000' with size 268435456"
I1023 01:24:34.840746 53 cuda_memory_manager.cc:107] "CUDA memory pool is created on device 0 with size 1000000000"
I1023 01:24:34.868128 53 model_lifecycle.cc:472] "loading: conformer-en-US-asr-streaming-asr-bls-ensemble:1"
I1023 01:24:34.868169 53 model_lifecycle.cc:472] "loading: riva-onnx-fastpitch_encoder-English-US:1"
I1023 01:24:34.868206 53 model_lifecycle.cc:472] "loading: riva-punctuation-en-US:1"
I1023 01:24:34.868226 53 model_lifecycle.cc:472] "loading: riva-trt-conformer-en-US-asr-streaming-am-streaming:1"
I1023 01:24:34.868246 53 model_lifecycle.cc:472] "loading: riva-trt-hifigan-English-US:1"
I1023 01:24:34.868262 53 model_lifecycle.cc:472] "loading: riva-trt-riva-punctuation-en-US-nn-bert-base-uncased:1"
I1023 01:24:34.868277 53 model_lifecycle.cc:472] "loading: spectrogram_chunker-English-US:1"
I1023 01:24:34.868292 53 model_lifecycle.cc:472] "loading: tts_postprocessor-English-US:1"
I1023 01:24:34.868315 53 model_lifecycle.cc:472] "loading: tts_preprocessor-English-US:1"
I1023 01:24:34.869711 53 onnxruntime.cc:2718] "TRITONBACKEND_Initialize: onnxruntime"
I1023 01:24:34.869739 53 onnxruntime.cc:2728] "Triton TRITONBACKEND API version: 1.19"
I1023 01:24:34.869745 53 onnxruntime.cc:2734] "'onnxruntime' TRITONBACKEND API version: 1.16"
I1023 01:24:34.869751 53 onnxruntime.cc:2764] "backend configuration:\n{\"cmdline\":{\"auto-complete-config\":\"false\",\"backend-directory\":\"/opt/tritonserver/backends\",\"min-compute-capability\":\"5.300000\",\"default-max-batch-size\":\"4\"}}"
E1023 01:24:34.886243 53 model_lifecycle.cc:642] "failed to load 'riva-trt-conformer-en-US-asr-streaming-am-streaming' version 1: Not found: unable to load shared library: /usr/lib/aarch64-linux-gnu/libstdc++.so.6: version `GLIBCXX_3.4.32' not found (required by /usr/lib/aarch64-linux-gnu/nvidia/libnvdla_runtime.so)"

Thks!

riva-start.log (51.0 KB)

Hi,

RIVA is not available on the Thor yet.
But our internal team is actively working on adding the support.

Thanks.

Hi,

Thanks for your quick response.

Though, which means no any hints to get up.

I check the container has gcc11.4 version, don’t know if gcc update can solve the issue.

Regards

Hi,

It is not working right now.

The package (nvidia/riva/riva_quickstart_arm64:2.19.0) is built for Orin, which uses the nvgpu Driver.
But Thor’s GPU driver has been changed to the SBSA driver, so the package cannot work.

Please wait for our upcoming release for the RIVA spport on Thor.
Thanks.