Dustin Tran

Thanks for pinging me. That's a great question. My understanding is because you'd like to broadcast `K` along a batch dimension, that should work without having to tile. (At least,...

Note: This is important for using the Observations library for restricted data sets, where you might first need to ask the author or go through some process to download the...

I can look into my research group's funding to see if we can dedicate a server somewhere to host the data. Note this may take at least two weeks.

update: we have the funding. amazon s3 as the solution is most likely. a todo is to decide on the specific pricing option and actually set up the server.

Observations is agnostic to the user's choice of workflow, which is a deliberate design choice. That said, it could be useful to see just how much we can push generic...

Agnostic to choice of workflow as in including the framework too. I think of Observations as having a longer life span than Edward or TensorFlow in that it's more likely...

I thought about it and agree with you. I think it makes sense to have celeba/lsun/etc. functions load and return objects for contrib.data.dataset.

Either as part of `util.py` or a new `util_tf.py` if you think it's substantial and dependency-ridden enough to have separately.

A currently private but soon-to-be-open probabilistic programming library built on PyTorch also uses this library. We should make sure to enable other data loaders and not just TF's.

apologies for the (very) delayed response; from a cursory glance it would be as simple as changing those classes.