Request for Optimal Brain Surgeon -- SparseGPT

Hi and thanks for the amazing repo.

I have a bit of tall request. SparseGPT uses a per-layer optimal brain surgeon approach to pruning. Here is the [pytorch code](https://github.com/IST-DASLab/sparsegpt).

Having this in jax would really help push the boundaries of what we can do.

Thank you,
Omead

The request points to the SparseGPT PyTorch implementation as the reference. Start by studying that linked implementation and then inspect jaxpruner's existing pruning entry points to determine the integration scope. Done means SparseGPT-style pruning is available in the JAX project and its behavior is validated against the intended reference.

Written by the indexing model from the issue text.