Releases: vosjo/nnaps
v0.2.1
v0.2.0
New release: v0.2.0
Removed XGboost predictors in favour of sklearn native gradient boosted predictors. The basic gradient boosted classifiers are not as good as those of XGBoost, but are more than sufficient for this package and reduce the number of dependencies.
Function usage remains the same, only the naming of the class is changed: XGBPredictor -> GBPredictor
Bugfix: Fixed bug in He ignition phase detection. Core ignition is now taken as He ignition for stars that ignite non degenerately. Previous fix didn't work for all stars.
Documentation: Updated documentation to reflex name change in GBPredictor.
Other: Updated dependencies and added minimum required versions for the most important ones.
v0.1.0
New Release: v0.1.0
Upgraded development status from Alpha to Beta
Streamlined command line interface to be more consistent using command groups. Processing of MESA files is not split into two main tools: compress and extract with their own interface with consistent parameter naming.
Code refactor to keep naming of mesa tools consistent.
Documentation update with new CLI, and extended parts that were not clear.
Implemented CI/CD using github actions.
v0.0.10
v0.0.9
Features:
- Added option to flatten the csv output from extract_mesa. So if multiple mass loss phases (or others) are requested, the output csv file can now be flat instead of having nested lists in columns.
Bug fixes:
- Fixed small bug in calculation of M_div_Mdot_div_P
- Fixed bug in fileio.write_hdf5. In versions <3.0 of h5py files are opened in mode 'a' by default, added this in explicitly as in version >=3.0 files are opened in r+ mode and will fail to create a new file.
- The HeIgnition phase now takes into account non degenerate ignition correctly, and for those systems marks the ignition phase based on core ignition criteria instead of a maximum in LHe.
v0.0.8
Features:
- extract_mesa can now extract multiple ML phases and include them in the result csv file as list objects. Added n_ml_phases keyword to setup file to deal with this.
- mass loss phases are now counted by default from RLOF alone, wind mass loss is not included.
- ML phase can now distinguish between rlof, wind or total mass loss.
- count_ml_phases function can now distinguish between rlof, wind or total mass loss.
- NNaPS version is added to hdf5 files created by mesa-2h5
Bug fixes:
- count_ml_phases, prevent crashing if the end of the model is reached at the same time of the end of a mass loss phase.
- Fixed issue in determining the start of the MS phase. log(L) wasn't calculated correctly.
Documentation:
- Added support for jupyter notebooks in docs
- Added a tutorial on predicting orbital properties of hot subdwarf binaries.
- Added API section
- Expanded docs on common envelope ejection
- Expanded docs on evolution phases
- Expanded docs on error checking
Other:
- read_mesa is renamed compress_mesa to better fit with future development plans (internal use only)
- extracted duplicated functionality from extract mesa (internal use only)
v0.0.7
New Features:
- Added start and end phases for MS and RGB
- Extract mesa counts number of distinct mass loss phases
- Extract mesa check for several possible errors that can happen during the evolution and flags them.
Bug fixes:
- bug fix in RGB evolution phase, end is now always after the starting age.
- bug fix in predictors.plot_training_history
Other:
- extended documentation
- added tests and rearranged test suite.
v0.0.6
new release: version 0.0.6,
- nnaps-mesa can now accept different processing parameters for each model that gets processed.
- the score function in the predictors also displays the mean of the regressors
- newer versions that sklearn 0.22 are now also supported for save and load to hdf5 format.