Performance on M1 chips compared to PyTorch implementation?

[andrewschreiber]

Looks like JAX does not support accelerated M1 / Apple Neural Engine.

Curious if anyone has done a benchmark comparison. Reference:

jax-ml/jax#8074

openai/whisper#382

[sanchit-gandhi]

This would be super cool 🙌 Think we need JAX working on M1 first (as per the first link you've shared)