Hi folks,
Just wanted to let you all know that we are getting ready to wrap up the current changes in greta and submit them to CRAN. I also wanted to make a habit of posting here more regularly so you know that things are happening!
It’s been a long time coming - unfortunately there will still be some rough edges and not as feature complete as we would like.
Rough edges include:
- Python warnings that have been elusive to track down
Not being as feature complete as we would like includes:
- cholesky representation is a newly discovered bug, and might require a substantial re-architecture of the code base. It will work, but it is not fast - this means that we have an opportunity in the future to speed up code that uses cholesky representation.
- There are more opportunities for greta to stay using TF rather than polling it from an R session frequently
- some new TFP optimisers aren’t yet implemented.
Overall this means that there more opportunities to make greta faster and better for users.
Currently I’m experiencing a bit of developer fatigue, so I’m focussing on some easier to tackle issues - which you can track here: https://github.com/greta-dev/greta/issues?q=is%3Aissue+is%3Aopen+label%3A"low+hanging+fruit"
I will keep you posted this week on some changes that we will be making, and will ask for feedback on things like print methods and other things.
Cheers!
Nick
(Maintainer of greta)