The adoption of #92 caused warnings to be logged by default during model initialization, because the special nodes for the log likelihood and log probability cannot always be updated directly on their initialization. This PR just sets update_on_init=False for these nodes, which solves the issue.
The adoption of #92 caused warnings to be logged by default during model initialization, because the special nodes for the log likelihood and log probability cannot always be updated directly on their initialization. This PR just sets
update_on_init=False
for these nodes, which solves the issue.