st--

@sdwebb if you know how to fix something, you can also edit the files directly and open a pull request, but yes please do open [GitHub issues related to documentation](https://github.com/GPflow/GPflow/issues/new?template=doc-issue.md)!

There isn't a code problem but we do need to improve how it's documented

You're probably right; as mentioned on the corresponding PR, it'd be great to get this fixed -would be good to have a test that demonstrates this issue (and its fix)...

@j-wilson I just noticed this PR is still not merged despite @vdutor having approved it: it seems you just need to get the tests to pass (and merge latest develop...

@awav Could you add some more thoughts on how you think we could integrate better with TFP?

TODO: remove the comment in the varying-noise notebook (#962) once this is fixed.

TODO: un-skip the MultiClass likelihood tests that were failing due to ndiagquad limitations (#1083, #1091).

@gustavocmv can we address these TODOs with the new Quadrature classes?

Here's the comment in the varying noise notebook: https://github.com/GPflow/GPflow/blob/b74e19e74c1a0e705e69a6bbee7709354e12563e/doc/source/notebooks/advanced/varying_noise.pct.py#L100 And here are the missing tests for the MultiClass likelihood: https://github.com/GPflow/GPflow/blob/b74e19e74c1a0e705e69a6bbee7709354e12563e/tests/gpflow/likelihoods/test_multiclass.py#L172 (original tests: https://github.com/GPflow/GPflow/commit/5c89c1e5ab1b9a91b6df60f7ce8e75cac810f14b#diff-afea7f6ff7ae8e983b82e215a2bbfbf5L169 and code in old commit: https://github.com/GPflow/GPflow/blob/dda8c39889acb6edf237613663ccbcb09888970d/tests/gpflow/likelihoods/test_likelihoods.py#L169)

@awav could you clarify the issue description with what/how you'd like to see GPflow integrated into Keras ?:)