ASR offline model not found on Jetson Orin

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

Using Jetson Orin NGX 64GB

Operating System - Ubuntu 20.04.5 LTS
Riva Version - arm64 v2.7.0

First I run bash riva_init.sh and then I try running script from tutorial:
riva_asr_client --audio_file=/opt/riva/wav/en-US_sample.wav

I got follwing error:

Client output:

I1128 10:22:37.429924   420 riva_asr_client.cc:445] Using Insecure Server Credentials
Loading eval dataset...
filename: /opt/riva/wav/en-US_sample.wav
Done loading 1 files
RPC failed: Error: Unavailable model requested. Lang: en-US, Type: offline
Done processing 1 responses
Some requests failed to complete properly, not printing performance stats

Server output:

I1128 10:22:37.434805   197 grpc_riva_asr.cc:492] ASRService.Recognize called.
I1128 10:22:37.434899   197 riva_asr_stream.cc:214] Detected format: encoding = 1 numchannels = 1 samplerate = 16000 bitspersample = 16
I1128 10:22:37.434916   197 grpc_riva_asr.cc:558] ASRService.Recognize performing streaming recognition with sequence id: 1905442010
E1128 10:22:37.434995   197 grpc_riva_asr.cc:407] Error: Unavailable model requested. Lang: en-US, Type: offline

server-output.txt (10.4 KB)
config.sh (9.7 KB)

I am not seeing any citrinet_256 RMIR file in model_repository. I’ve also tried default ASR conformer model but the output was the same. Somehow script which should initialize riva server do not download all required files.

Any ideas?

2

HI @tymonzar

Thanks for your interest in Riva,

Thanks for sharing the config.sh file,

I can see the offline model is present (uncommented)

I kindly request to try

  1. bash riva_clean.sh
  2. rerun bash riva_init.sh
    And let me know whether it works,
    Also if still not working, please send docker logs riva-speech complete output as file in this thread

Thanks

3

Hi @rvinobha,

thanks for contact - unfortunately it did not help. Here is the log:
I am also attaching full output after issuing bash riva_start.sh (as riva-full-log.txt file)

riva-speech-logs.txt (1.6 KB)
riva-full-log.txt (88.0 KB)

4

Any updates on that issue?

5

Hi @tymonzar

Apologies, I will check with the team and provide updates

Thanks

6

Hi @tymonzar

Offline Model is not packaged as prebuilt for embedded by default, Please fins the instruction below to add

  1. Open config.sh
  2. find models_asr and add the below highlighted change
"${riva_ngc_org}/${riva_ngc_team}/rmir_asr_${asr_acoustic_model}_${modified_lang_code}_ofl:${riva_ngc_model_version}"

After the change the code will look like below

image
image1567×227 8.35 KB

Now your offline model will work after this addition

Thanks

7

I believe you should add this info somewhere. The quick start instruction is not correct for that case.

8

Thanks @tymonzar for your feedback

This will be added as code in config.sh in upcoming version

Thanks

9

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