st--

In case it's not easy to figure out how to get the suggestions on fork PRs: how about providing `Makefile`s for this - e.g. to allow `make format` (make is...

Hi, suggestions (e.g. for adding a package to some category) still seems to be broken - is there any other way of editing categories for our packages?

What are the PDMats.jl maintainers' thoughts on this?

@markvdw have you had any further thoughts on this? would you be up for looking into it ?

If not fixed outright (#1209), then at least the documentation should be clearer to prevent people from spending their time trying to make it work, e.g. https://stackoverflow.com/questions/61852379/computing-the-log-marginal-likelihood-of-gpr-model-with-a-linearcoregionalizatio

This is actually still an issue - moreover, there seems to be a bug in the expectations code, even the `fHmmm = False` branch doesn't work! I've updated the MWE...

For the `fHmmm = False` case (analytic expectation), the traceback (without the tf.function, cut to the relevant part) is: ``` File "/home/st/Code/gpflow2/gpflow/models/gplvm.py", line 148, in log_likelihood (self.kernel, self.inducing_variable)), File "/home/st/Code/gpflow2/gpflow/expectations/expectations.py",...

Good catch, I think that is definitely a bug. Fixed in #1178. I guess the active_dims=slice(...) issue remains open though?

@ialong hi did you have a chance to look at this in GPflow 2 ? would be great if you could confirm whether or not this is still an issue...

Makes sense! if you do want to look at it after NeurIPS that'd be great, and join us on the gpflow.slack.com chat (invite link in the README.md) if you want...