All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- Fix error when using automatic mixed precision with spline transforms (#66)
- Add various autoregressive flows using the existing transforms in
nflows
(#62) - Add
scale_activation
keyword argument tonflows.transforms.autoregressive.MaskedAffineAutoregressiveTransform
(uofgravity/nflows#11)
- Drop support for Python 3.7 (#61)
- Addressed a deprecation warning in the
nflows
submodule when using LU decomposition (uofgravity/nflows#10, #57)
- Keyword arguments passed to
glasflow.transform.coupling.AffineCouplingTransform
are now propogated to the parent class. (#51) - Add support
scale_activation
toglasflow.transform.coupling.AffineCouplingTransform
and set the default tonflows_general
. (#52, #54)
- Default scale activation for
glasflow.transform.coupling.AffineCouplingTransform
is changed fromDEFAULT_SCALE_ACTIVATION
in nflows tonflows_general
from glasflow. This changes the default behaviour, the previous behaviour can be recovered by settingscale_activation='nflows'
. (#52, #54)
- fix a bug in
glasflow.nflows/utils/torchutils.searchsorted
, see uofgravity/nflows#9 for details. (#53)
- The
scaling_method
argument inglasflow.transform.coupling.AffineCouplingTransform
is now deprecated in favour ofscale_activation
and will be removed in a future release. (#52)
- Add a multi-layer perceptron (
glasflow.nets.mlp.MLP
). (#40) - Add a resampled Gaussian distribution that uses Learnt Accept/Reject Sampling (
glasflow.distributions.resampled.ResampledGaussian
). (#40) - Add
nessai.utils.get_torch_size
. (#40) - Add a multivariate uniform distribution for Neural Spline Flows (
glasflow.distributions.uniform.MultivariateUniform
). (#47)
- Change logging statements on import to, by default, only appear when an external version of nflows is being used. (#44)
Another patch to fix CI not uploading release to PyPI
Patch to fix CI not uploading release to PyPI
- Add LFS to
publish-to-pypi
workflow (#35)
- Add
RealNVP
- Add
CouplingNSF
(Coupling Neural Spline Flow) - Add
nflows
submodule that replacesnflows
dependency - Add option for user-defined masks in coupling-based flows