Canary 1b producing 'x's as transcription on Arabic audio

[pruthvidhar.nanda]

Please provide the following information when requesting support.

Hardware - GPU (A10)
Hardware - CPU
Operating System: Ubuntu 22.04
Riva Version: 2.18.0
TLT Version (if relevant)
How to reproduce the issue ? (This is for errors. Please share the command and the detailed log here)
I ran the below through python-clients and got this error for various audio files. I also found that on huggingfaces only English, German, French, Spanish is supported but the model card in riva it also supports Arabic. Would be great if someone helped me about this.

python scripts/asr/transcribe_file_offline.py --model-name ‘wz0YOwXoBMKKYtXE.wav’ --input-file ‘Paragraph_test.wav’ --language-code ar-AR
results {
alternatives {
transcript: "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx "
}
channel_tag: 1
audio_processed: 30
}
results {
alternatives {
transcript: "ءْ إِلَيْنَا الْيَوْمَ وَاِسْتَعَدُّوا لِاِنْتِقَالْ إِلَى مُسْتَوَى جَدِيدٍ لِلنَّجَاحْ "
}
channel_tag: 1
audio_processed: 34.1135
}
id {
value: “9be402f7-dac9-4094-bfc7-dc492b9f5da6”
}

Final transcript: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx ءْ إِلَيْنَا الْيَوْمَ وَاِسْتَعَدُّوا لِاِنْتِقَالْ إِلَى مُسْتَوَى جَدِيدٍ لِلنَّجَاحْ

[jkh]

Hey! I think these transcripts include diacritics! Can you please share the model name/link as well as Riva configuration? Did you use one of the pre-built models (e.g. this )? @pruthvidhar.nanda

[pruthvidhar.nanda]

Of course.
This is the config.sh
@jkh sorry for the terrible config quality. Here you go again.

Download config_canary_1b.sh | LimeWire

Please do let me know if you are facing the same issue.

[jkh]

Cannot download the file. You can use
https://wetransfer.com

[pruthvidhar.nanda]

I’m afraid I don’t have your email.

[jkh]

There is no need for an email, once can just generate a link and share it.