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
When sampling large scale problems, it requires a lot of memory to append and store all samples to history.
CUQIpy offers the functionality to save samples in "batches" of specified sizes. This let's the user save smaller files, but does not change to total storage requirements, as I understand it.
I am interested in functionality where I can pre-specify a thinning parameter, e.g. 10, such that the sampler updates the state 10 times before appending to the history. A variaty of this seems to be implemented for experimental.mcmc.HybridGibbs (num_sampling_steps), but for the regular experimental.mcmc.sampler, this does not seem to be an option.
The text was updated successfully, but these errors were encountered:
When sampling large scale problems, it requires a lot of memory to append and store all samples to history.
CUQIpy offers the functionality to save samples in "batches" of specified sizes. This let's the user save smaller files, but does not change to total storage requirements, as I understand it.
I am interested in functionality where I can pre-specify a thinning parameter, e.g. 10, such that the sampler updates the state 10 times before appending to the history. A variaty of this seems to be implemented for experimental.mcmc.HybridGibbs (num_sampling_steps), but for the regular experimental.mcmc.sampler, this does not seem to be an option.
The text was updated successfully, but these errors were encountered: