Alex Lyttle

**Update**: This seems to work for saving the posterior, using `xarray`'s dataset to netcdf function with the `'scipy'` engine. ```python trace.posterior.to_netcdf('posterior.h5', engine='scipy') ``` reload with ```python from xarray import open_dataset...

Thanks @ericmjl for tagging the relevant people! Hi @AlexAndorra thanks for the response. I understand this isn't priority atm but thought it worth mentioning. With regards to PyMC4 not being...

@lucianopaz thanks for your interest! It's possible the main contribution to the speed-up is that we are using a regression neural network (fully-connected 6 x 128 hidden neurons) trained using...

Thanks for the advice @lucianopaz! Actually, I have been noticing divergences and some other issues with the chains. With regards to changing the `step_size` I tried implementing something similar to...

Hi @OlaRonning, I would love to help and I have been giving this some thought. At first I thought the inclination angle of stars may be a good example, but...

@OlaRonning > Regarding orbits, I think the angle tends not be important as we care more about the period and distance from the star. I was wrong here... I think...

It appears the above workaround no longer works for Calendar.app for me. It does seem to be working for Mail.app still, so maybe Apple have changed the way suggestions in...

> Ok, so I figured out a quick fix. Just add `jax.distributed.initialize(local_device_ids=range(8))`. Works like a charm. > > ``` > import jax > import os > jax.distributed.initialize(local_device_ids=range(8)) > print("jax.__version__", jax.__version__)...