Spaces:

---

ivrit-ai

/

README

like 0

Running

Hello,
Thanks for all the great work.
I am trying to explore the world of hebrew ASR, and I have some questions.

I saw in the readme that you wrote that your recommended model is "faster-whisper-v2-d4". (It is a faster_whisper format based on ctranslate2 and not based on the Hugging Face's transformer)

1. What is the most similar model that available for Hugging Face's transformer? Are they same (except the file format)?
2. How did you train "faster-whisper-v2-d4"? I saw this manual for finetuning Hugging Face's transformers models, and I saw the script ct2-transformers-converter that can convert the transformers model to faster-whisper's format, but I don't know what was the base model for the finetuning, and which dataset has been used for the finetuning. (or maybe the finetuning has been done in entirely different way.)