Finetuning:

https://github.com/booydar/t5-experiments/blob/scaling-repor...

I found this interesting link in paper, so this can be used as ready code for RMT Transformer

https://github.com/booydar/t5-experiments/tree/scaling-repor...

Whisper JAX is an optimised implementation of the Whisper model by OpenAI. It runs on JAX with a TPU v4-8 in the backend. Compared to PyTorch on an A100 GPU, it is over 70x faster, making it the fastest Whisper API available.

how does it compare to whisper.cpp?

Not a good comparison. CPU vs TPU. Also, whisper.cpp perf varies a ton depending on hardware. Avx512 or neural engine?

thanks