GitLab

The `copy` subcommand of `pop_models_h5_cleaned_samples` now has an
optional `--unique` flag which discards any duplicate posterior
samples.  This is useful when dealing with runs with stuck MCMC
walkers, particularly when using such a run to seed another run in
hopes of becoming un-stuck.

Previous dependency list did not allow h5py 3.0 or above as it was not
released and could break compatibility.  Now that it is out and does
not appear to break anything, released prior to the future 4.0 release
are allowed.

Specifically for the PL+G model, there is now a tool for computing the population-weighted average of a single event's likelihood.

Improved sampling

See merge request !21

Introduces generic wrappers for rescaling chosen parameters before
passing to `emcee.moves.Move` objects, with one implementation for
log-scaling.  Can currently be used via API or the PL+G model.

Chain plotting now supports overlaying a rolling average of the
chains.

MCMC sampling now supports rescaling select hyperparameters, but this
is currently introducing numerical problems, and should not be used.

Fixed a simple yet critical bug in the H5RawPosteriorSamples  corruption recovery code, where we miscounted the number of samples in each subfile after the first, resulting in all uninitialized values from the last subfile being included.

Also made some tweaks to the chain plotting code, in preparation for an improved, color-coded option.

Injection vt

See merge request !20

Fixed bugs in `ReweightedInjectionPoissonMean` through a rewrite of all
mathematical expressions.  Removed redshift dependance, and overrode with
zero-redshift model.

Added a new generic `SummedPoissonMean` class, which takes one or more
`PoissonMean` objects of any type, and takes the sum of their output when
producing its output.

Removed erroneous `close()` call, as the returned object is in fact `None`.

Added missing storing of metadata.