There were many contributions to the PyMC4 repository in 2019, but few since then. Have the contributors lost interest? Decided that the project was a mistake? Decided to model directly in TensorFlow Probability? Or perhaps everyone is busy with other projects, and plan to return to PyMC4 when they free up?
We have quite a few new GSOC projects to add more features, so stay tuned There were some road blocks on the sampling front on the TFP side (having a fast and reliable way to tune mass matrix), but we are making some good progress (e.g., https://github.com/tensorflow/probability/tree/master/discussion/turnkey_inference_candidate)
In my particular case, with the covid lockdown I have to focus the little time I have in between family and house keeping activities to my paying job. I have been trying to keep contributing small infrastructure things like docstrings, but I won’t be able to put time in the near future into the core work that still needs to be done.
I’m also waiting on a progress bar for the samplers…
Update: PyMC4 is dead. The future of PyMC3 is PyMC3, with a JAX backend.