Riva Quickstart 2.2.1 offline en-US models missing

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

Please provide the following information when requesting support.

Hardware - AGX Orin
Hardware - CPU
Operating System Ubuntu 20.04
Riva Version 2.2.1

After a successful install of the Riva Quickstart 2.2.1 following the quick start guide the streaming example for audio file transcription runs well. However, the offline example provides the following error:

root@fb724947c645:/opt# riva_asr_client --audio_file=/work/wav/en-US_sample.wav
I0627 18:45:55.505287   235 riva_asr_client.cc:445] Using Insecure Server Credentials
Loading eval dataset...
filename: /work/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

By looking at the config.sh it seems that the offline models are not loaded:

    if [[ $riva_target_arch == "arm64" ]]; then
      models_asr+=(
      ### Citrinet-1024 Streaming w/ CPU decoder, best latency configuration
          "${riva_ngc_org}/${riva_ngc_team}/models_asr_citrinet_1024_${modified_lang_code}_str:${riva_ngc_model_version}-${riva_target_arch}${riva_arm64_legacy_platform}"
      )

Could somebody provide help to include the offline models. Otherwise the basic https://github.com/nvidia-riva/tutorials/blob/stable/asr-python-basics.ipynb does not work.

1 Like
2

Hi @user108910

Thanks for your interest in Riva

I will check with the team regarding the offline ASR model for embedded and get back to you soon

Thanks for your patience

3

Hi, My name is Ellie White, I have completed my data science training from SynergisticIT recently. And I am facing the same issue.

4

HI @user108910 and @elliewhite7090

Thanks for your interest in Riva,

Please add the below line in your highlighted code (as shown below)

if [[ $riva_target_arch == "arm64" ]]; then
      models_asr+=(
      ### Citrinet-1024 Streaming w/ CPU decoder, best latency configuration
          "${riva_ngc_org}/${riva_ngc_team}/models_asr_citrinet_1024_${modified_lang_code}_str:${riva_ngc_model_version}-${riva_target_arch}${riva_arm64_legacy_platform}"
      ### Citrinet-1024 Offline w/ CPU decoder,
          "${riva_ngc_org}/${riva_ngc_team}/rmir_asr_citrinet_1024_${modified_lang_code}_ofl:${riva_ngc_model_version}"
      )

Let us know if you face issues

Thanks