Docker - Riva fails to launch Model not specified

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

I get the Riva Speech Skill docker image path: nvcr.io/nvidia/riva/riva-speech:2.8.1
It doenst load the model and the bin file only has 2 executables: start-riva and riva_server

I am confused since I cant find any reference on the documentation.

riva issue
riva issue837×469 15.7 KB

2

Hi @Pappachuck_renan where do you have your models?

The riva_start.sh under the hood launches this:

docker run -d \
    --init \
    --gpus '"'$gpus_to_use'"' \
    -p $riva_speech_api_port:$riva_speech_api_port \
    -e "LD_PRELOAD=$ld_preload" \
    -e "RIVA_API_KEY=$RIVA_API_KEY" \
    -e "RIVA_API_NGC_ORG=$RIVA_API_NGC_ORG" \
    -e "RIVA_EULA=$RIVA_EULA" \
    -v $riva_model_loc:/data \
    $ssl_vol_args \
    --ulimit memlock=-1 --ulimit stack=67108864 \
    --name $riva_daemon_speech $docker_run_args

Models are in $riva_model_loc variable path and then mounted inside the docker in /data. Can you check that inside /data you have the models and it is not empty?

Then try to launch

    start-riva \
      --riva-uri=0.0.0.0:50051 \
      --asr_service=true \
      --tts_service=true \
      --nlp_service=true

To start riva server.

If you are still having problems you can paste all the container log using the docker logs command.

Regards!

3

Hi @Pappachuck_renan

Thanks for your interest in Riva

Thank you so much @jlamperez10 for your kind help and inputs, really appreciate

@Pappachuck_renan are you trying to setup and start using the quickstart guide
https://docs.nvidia.com/deeplearning/riva/user-guide/docs/quick-start-guide.html
Were bash riva_init.sh and bash riva_start.sh successful

Thanks

4

We found the error:
riva-trt-conformer-en-US-asr-streaming-am-streaming

You must remove from production until fixed.