Hi all,
First congrats for this amazing project! Sherpa is really awesome.
But there is so many models, I am a little bit lost ... and maybe there is not so many model for my use case : streaming, 4Go RAM, French.
In order to be efficient in real time, I looked for transducer based models but I only found these ones for the French language;
- sherpa-onnx-nemo-fast-conformer-transducer-en-de-es-fr-14288.tar.bz2
- sherpa-onnx-nemo-fast-conformer-transducer-be-de-en-es-fr-hr-it-pl-ru-uk-20k.tar.bz2
- icefall-asr-multidataset-pruned_transducer_stateless7-2023-05-04.tar.bz2
And those models are probably to big to be deployed and used on A32....
So, does anyone know which model could be a good alternative to Kroko for the french language ? The targeted smartphone is typically Samsung A32 with RAM 4 Go.
I tried the "basic" zipformer from 2023 in the int8 format (sherpa-onnx-streaming-zipformer-fr-2023-04-14.tar.bz2) but I have a lot a mistakes in the transcriptions. And injecting hotwords will be not be sufficient to solve the kind of errors that I saw in the transcriptions.
Any help or suggestion is welcome
Thank you
Hi all,
First congrats for this amazing project! Sherpa is really awesome.
But there is so many models, I am a little bit lost ... and maybe there is not so many model for my use case : streaming, 4Go RAM, French.
In order to be efficient in real time, I looked for transducer based models but I only found these ones for the French language;
And those models are probably to big to be deployed and used on A32....
So, does anyone know which model could be a good alternative to Kroko for the french language ? The targeted smartphone is typically Samsung A32 with RAM 4 Go.
I tried the "basic" zipformer from 2023 in the int8 format (sherpa-onnx-streaming-zipformer-fr-2023-04-14.tar.bz2) but I have a lot a mistakes in the transcriptions. And injecting hotwords will be not be sufficient to solve the kind of errors that I saw in the transcriptions.
Any help or suggestion is welcome
Thank you