Summary:
AGX Xavier updated to latest Jetpack was previously able to run the latest available Riva Quickstart that supported Xavier, but now all new setups fail due to English-US models not getting downloaded except for first 110 bytes (apparently)
Details:
Hardware AGX Xavier Dev Kit (32 GB RAM - 2 TB NVME)
Operating System; Jetpack 5.15 - Jetson Linux 35.6.2 - Ubuntu 20.04.6 LTS - Kernel 5.10.216-tegra
Riva Version - 2.14
TLT Version (if relevant)
How to reproduce the isssue ?
- bash riva_clean.sh
- download riva_quickstart_arm64_2.14.0
- edit config.sh to change riva_tegra_platform=“xavier”, no other changes
- bash riva_init.sh
(downloads just 110 bytes for each model and then says it’s ready)
- bash riva_start.sh
- now any attempt to use TTS or ASR results in model not available
This sequence worked beautifully in January 2025. Now the initialization thinks it succeeds but the models don’t actually download. Using Quickstart 2.19 results in models downloaded on the same system, but then it doesn’t work on Xavier even if I edit the config.sh file to specify xavier instead of orin.
Here is the console output from riva_init (version 2.14):
carter@carter-010:~/nvidia/riva_quickstart_arm64_2.14.0$ bash riva_init.sh
Logging into NGC docker registry if necessary...
Pulling required docker images if necessary...
Note: This may take some time, depending on the speed of your Internet connection.
> Pulling Riva Speech Server images.
> Image nvcr.io/nvidia/riva/riva-speech:2.14.0-l4t-aarch64 exists. Skipping.
Downloading models (RMIRs) from NGC...
Note: this may take some time, depending on the speed of your Internet connection.
To skip this process and use existing RMIRs set the location and corresponding flag in config.sh.
2025-09-29 23:14:55 URL:https://xfiles.ngc.nvidia.com/org/nvidia/team/ngc-apps/recipes/ngc_cli/versions/3.26.0/files/ngccli_arm64.zip?versionId=7Hjg08XzaQxBan_0sPP4JvVk29zu_uTK&Expires=1759274094&Signature=CvnPbDcvAmozEpXBIT91H3KFpSnM11Q4eD9RZf9XyU3jwbFmZ2c-O8aVob3tEqPaqdtCELjeuRglgmZ81xjyD24NaBnPLsfmNLf7MV3N-4xymYOy3~OmMDz4PVhSOrnaRDnhsEo8-njerZmT1skUYYcQTku6TYhscrnBNvLvs-4hGnA6VsOOQX32JfwMBhFDAMbJxGBxs6yEYVBJxXmygjcUwxvusIU60jiKy9rsI8EF6dRfJVM0u75H3cXSA7m47U78x7g3eEywBEaCghJZxnwwQiwEIkb7psW-A0cI1JskWANo2io7lWLDW3JoDFPppuriX0zXy1AXBK7GsbyWuw__&Key-Pair-Id=KCX06E8E9L60W [47840604/47840604] -> "ngccli_arm64.zip" [1]
/opt/riva
CLI_VERSION: Latest - 4.3.0 available (current: 3.26.0). Please update by using the command 'ngc version upgrade'
Getting files to download...
⠏ ━━ • 0… • Remaining: … • ? • Elapsed: 0… • Total: 1 - Completed: 1 - Failed: 0
M…
--------------------------------------------------------------------------------
Download status: COMPLETED
Downloaded local path model: /data/artifacts/models_asr_conformer_en_us_str_v2.14.0-tegra-xavier
Total files downloaded: 1
Total transferred: 110 B
Started at: 2025-09-29 23:15:02
Completed at: 2025-09-29 23:15:03
Duration taken: 0s
--------------------------------------------------------------------------------
Getting files to download...
⠧ ━━ • 0… • Remaining: … • ? • Elapsed: 0… • Total: 1 - Completed: 1 - Failed: 0
M…
--------------------------------------------------------------------------------
Download status: COMPLETED
Downloaded local path model: /data/artifacts/models_nlp_punctuation_bert_base_en_us_v2.14.0-tegra-xavier
Total files downloaded: 1
Total transferred: 110 B
Started at: 2025-09-29 23:15:07
Completed at: 2025-09-29 23:15:08
Duration taken: 0s
--------------------------------------------------------------------------------
Getting files to download...
⠧ ━━ • 0… • Remaining: … • ? • Elapsed: 0… • Total: 1 - Completed: 1 - Failed: 0
M…
--------------------------------------------------------------------------------
Download status: COMPLETED
Downloaded local path model: /data/artifacts/models_nlp_punctuation_bert_base_en_us_v2.14.0-tegra-xavier
Total files downloaded: 1
Total transferred: 110 B
Started at: 2025-09-29 23:15:12
Completed at: 2025-09-29 23:15:12
Duration taken: 0s
--------------------------------------------------------------------------------
Getting files to download...
⠇ ━━ • 0… • Remaining: … • ? • Elapsed: 0… • Total: 1 - Completed: 1 - Failed: 0
M…
--------------------------------------------------------------------------------
Download status: COMPLETED
Downloaded local path model: /data/artifacts/models_tts_fastpitch_hifigan_en_us_ipa_v2.14.0-tegra-xavier
Total files downloaded: 1
Total transferred: 110 B
Started at: 2025-09-29 23:15:16
Completed at: 2025-09-29 23:15:17
Duration taken: 0s
--------------------------------------------------------------------------------
+ [[ tegra != \t\e\g\r\a ]]
+ [[ tegra == \t\e\g\r\a ]]
+ '[' -d /home/carter/nvidia/riva_quickstart_arm64_2.14.0/model_repository/rmir ']'
+ [[ tegra == \t\e\g\r\a ]]
+ '[' -d /home/carter/nvidia/riva_quickstart_arm64_2.14.0/model_repository/prebuilt ']'
+ echo 'Converting prebuilts at /home/carter/nvidia/riva_quickstart_arm64_2.14.0/model_repository/prebuilt to Riva Model repository.'
Converting prebuilts at /home/carter/nvidia/riva_quickstart_arm64_2.14.0/model_repository/prebuilt to Riva Model repository.
+ docker run -it -d --rm -v /home/carter/nvidia/riva_quickstart_arm64_2.14.0/model_repository:/data --name riva-models-extract nvcr.io/nvidia/riva/riva-speech:2.14.0-l4t-aarch64
+ docker exec riva-models-extract bash -c 'mkdir -p /data/models; \
for file in /data/prebuilt/*.tar.gz; do tar xf $file -C /data/models/ &> /dev/null; done'
+ docker container stop riva-models-extract
+ '[' 0 -ne 0 ']'
+ echo
+ echo 'Riva initialization complete. Run ./riva_start.sh to launch services.'
Riva initialization complete. Run ./riva_start.sh to launch services.
Note the above says just 110 bytes for each model ! but no error message.
Then riva_start.sh loads suspiciously fast with no indication of any errors.
carter@carter-010:~/nvidia/riva_quickstart_arm64_2.14.0$ bash riva_start.sh
Starting Riva Speech Services. This may take several minutes depending on the number of models deployed.
Waiting for Riva server to load all models...retrying in 10 seconds
Riva server is ready...
Use this container terminal to run applications:
root@70484a868c3d:/opt/riva#
So still no error but nothing works. For example:
python3 examples/talk.py --text "hello I am your favorite warehouse robot" --play-audio --voice English-US.Female-1 --output-device=23 --sample-rate-hz=32000
results in a bunch of expected messages from ALSA and then:
Generating audio for request...
Traceback (most recent call last):
File "examples/talk.py", line 114, in <module>
main()
File "examples/talk.py", line 97, in main
resp = service.synthesize(
File "/usr/local/lib/python3.8/dist-packages/riva/client/tts.py", line 68, in synthesize
return func(req, metadata=self.auth.get_auth_metadata())
File "/usr/local/lib/python3.8/dist-packages/grpc/_channel.py", line 1160, in __call__
return _end_unary_response_blocking(state, call, False, None)
File "/usr/local/lib/python3.8/dist-packages/grpc/_channel.py", line 1003, in _end_unary_response_blocking
raise _InactiveRpcError(state) # pytype: disable=not-instantiable
grpc._channel._InactiveRpcError: <_InactiveRpcError of RPC that terminated with:
status = StatusCode.INVALID_ARGUMENT
details = "Model is not available on server: Voice English-US for language en-US not found. Please specify the voice name in your SynthesizeSpeechRequest."
debug_error_string = "UNKNOWN:Error received from peer {grpc_message:"Model is not available on server: Voice English-US for language en-US not found. Please specify the voice name in your SynthesizeSpeechRequest.", grpc_status:3, created_time:"2025-09-29T23:27:22.379586613+00:00"}"
Again this all worked very nicely earlier this year. Even when using different voices. Now all the voices result in the same error: Voice English-US for language en-US not found.
Quickstart 2.19 doesn’t support Xavier but has no problem downloading models on the same system.
This AGX Xavier dev kit has plenty of storage, only 10% of 2TB used, and runs the latest available Jetpack.
i would appreciate any help in getting this up and running again like it worked earlier this year.