Question: Build my own or use elm

[shaunc]

I have my own predictors, and am using dask and Xarray. I am wondering if I should adopt elm's pipelines to manage the training process.

Elm says it is still experimental; it has not seen activity in a while -- is elm still maintained/progressing?

[jbednar]

Elm is not currently being maintained. We took a different approach in the end, focusing on improving various existing tools, mainly Xarray and Dask-ML, to achieve similar goals. See https://earthml.holoviz.org for some examples, though many of those focus on viz rather than ML per se. This repo is still available for anyone who wants to crib things from it, but it's not currently planned to go forward from where it is.

[shaunc]

Thanks for the reply! I'm using an overlapping set of libraries so it looked interesting -- but great: I'll probably just pick out a few bits that look useful! :)