Hello 
I would like to share some pretrained transformer models on HuggingFace hub. The code (both training and inference code) has been written from scratch in Jax using the Haiku library (GitHub - deepmind/dm-haiku: JAX-based neural network library). I am wondering what would be the best way to share these so that people who will use them can benefit from all huggingface functionalities? I would be fine to rewrite/adapt some parts of the code to make in fit in a framework such as transformers but I would like to avoid making big changes such as translating everything from Haiku to Flax.
Thanks,