Revision 4bdd8a5eb1a7ba462bdef14db5f35b6e37598e94 authored by Louis Tiao on 04 March 2022, 10:28:47 UTC, committed by GitHub on 04 March 2022, 10:28:47 UTC
The `_init_variational_parameters` method was using the instance attribute (`self.q_diag`) as well as input argument `q_diag` which refer to the same thing and should never diverge. In fact, `self.diag` is never used anywhere else so does not have to be an attribute.
1 parent bc68675
tests_requirements.txt
mypy
black==20.8b1
isort>=5.1
pytest>=3.5.0
pytest-cov
pytest-random-order
pytest-xdist # for local tests only
codecov
types-dataclasses
types-pkg_resources # for mypy check of gpflow/versions.py
types-tabulate
types-Deprecated
# Notebook tests:
tensorflow-datasets
nbformat
nbconvert
ipykernel
jupyter_client
jupytext
matplotlib
sklearn # for mixture-density-network notebook
ipywidgets # Required by tensorflow-datasets
Computing file changes ...
