You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
do we lose NUTS step size here (epsilon, etc)? .. if so maybe we can print a warning
I also wonder if setting sampler.step_size = sampler.epsilon would be better than doing nothing, for NUTS. But it could be for another issue. What I mean is something like:
ifisinstance(sampler, NUTS):
sampler.initial_point=sampler.current_pointstep_size=sampler.step_size
.
.
ifnotisinstance(sampler, NUTS): # Again, special case for NUTS.sampler.set_state(sampler_state)
sampler.set_history(sampler_history)
else:
sampler.step_size=step_size_Originallypostedby @amal-ghamdiinhttps://github.com/CUQI-DTU/CUQIpy/pull/503#discussion_r1750386017_
The text was updated successfully, but these errors were encountered:
Let us work on getting the NUTS sampler fully stateful in HybridGibbs. This would mean no special case handling for NUTS in HybridGibbs.
Also relates to #474 and #505
Comment
do we lose NUTS step size here (epsilon, etc)? .. if so maybe we can print a warning
I also wonder if setting sampler.step_size = sampler.epsilon would be better than doing nothing, for NUTS. But it could be for another issue. What I mean is something like:
The text was updated successfully, but these errors were encountered: