Matthew

Actually I realised the problem is bigger, since all the other functions take a credentials map rather than accept an AmazonS3Client object itself. It'd be nice if it supported passing...

Thanks, that looks interesting, although sounds like it's addressing a slightly different problem; I'm not so much interested in validation of external inputs, more just in having some basic datatypes...

Re `xr.broadcast(ds)[0].sum(dims)` -- Thanks, that's neat and may be useful as a workaround, but it looks like it could incur significant extra CPU and RAM costs (tiling all variables to...

Just seen this and my 2 cents: I'd value being able to parallelize the data processing more highly than the resumability stuff. Resumability seems to constrain the implementation a lot...

OK, yeah that makes sense. I need some level of reproducibility too, but unfortunately also need to do some relatively expensive transformations on-the-fly, which is why I'm particularly interested in...

I would definitely like to see this -- quite a noticeable gap at the moment. On a related point, how do I make predictions from a pre-trained Model? Right now...

OK, fair enough -- that's useful to know. Maybe worth showing an example of what the recommended approach is to creating some kind of model object with a prediction interface...

Ah that's great, thanks! This solves my problem, but just to give a bit more detail why I was confused about this: With `Rechunk`, it explicitly asks for a dict...

Hello and thanks for the interest! I'm afraid it's unlikely we'll open source this code as it's rather tied up in our proprietary data processing pipelines. That said we do...

While you're discussing this, is there a reason why MergeLayer needs to be a special subclass, could support for multiple incomings be rolled into the Layer superclass, and a single...