Skip to content

Latest commit

 

History

History
716 lines (549 loc) · 25.5 KB

CHANGELOG.rst

File metadata and controls

716 lines (549 loc) · 25.5 KB
Raw

Change Log

All notable changes to this project will be documented in this file. This project adheres to Semantic Versioning.

1.0.0

  • Bump the minimum nano-utils version
  • Issue a warning in the psf recipe if the qd or ligands don't have any bonds
  • Fix an issue with parsing multiple ligands in generate_psf2`
  • Add formal Python 3.11 support
  • Pin pandas to <2.0.0
  • Test using CP2K 2023.1
  • Add a citation.cff file

0.10.2

  • Fix the atom-pairs not being sorted if specific parameters are guessed.
  • Add a workflow for computing Debye scattering.
  • Fix various PES-averaged ARMC issues.
  • Fix improper dihedral-indices being incorrectly sorted.
  • Make the .xyz parser more robust.
  • Export warnings to the logger when running ARMC.
  • Fix a Boost issues with the PLAMS master branch.

0.10.1

  • Avoid schema 0.7.5.

0.10.0

  • Add a new MultiMolecule method for constructing supercells.
  • Add a timestep parameter to MultiMolecule.init_power_spectrum.
  • Fixed an issue wherein init_power_spectrum could raise when the atom subset is specified.
  • Raise a ValueError if an atom type with multiple charges is found.
  • Fixed an issue wherein net charges weren't properly conserved in PES-averaged ARMC.
  • Allow users to pass custom error functions.
  • Make rdkit an optional dependency.
  • Fixed an issue wherein auto-FOX would not check the status of jobs.
  • Add the new atom_pairs keyword to init_rdf and init_adf.
  • Round the net charge to the nearest integer.

0.9.1

  • Added the new segment_dict parameter to the PSFContainer.generate_x() functions.
  • Added the new PSFContainer.sort_values method.
  • Store metadata and net charges for each individual system in PES-averaged ARMC and ARMCPT.
  • Fixed an issue wherein frozen atomic charges were ignored when not explicitly specified.
  • Fixed in issue wherein PSFContainer.sort_values did not not update the residue ID.
  • Fixed an issue wherein guessed parameters could overwrite ones that were explicitly specified.
  • Fixed an issue wherein frozen parameters weren't properly sorted.
  • Fixed an issue wherein the mean pair-density was computed using lattice vectors in non-periodic calculations.

0.9.0

  • Fixed an issue wherein PSFContainer.to_atom_alias_dict would assign incorrect indices to atom aliases.
  • Fixed an issue wherein treating an entire ARMC parameter block as frozen could raise.
  • Added the FOX.MultiMolecule.lattice property for storing the lattice vectors of periodic systems.
  • Added FOX.io.lattice_from_cell, a function for reading lattice vectors from CP2K .cell files.
  • Added support for periodic ADF and RDF calculations.
  • Added the new job.lattice keyword to the ARMC workflow for specifying lattice vectors of the reference system.
  • Lattice vectors are now preserved when interconverting between PLAMS, ASE and Auto-FOX.
  • Read the pressure from the CP2K .out file when calculating the bulk modulus, rather than calculating it from scratch.

0.8.12

  • Added a recipe for calculating the similarity between 2 MD trajectories.
  • Added improved support for custom atom types.
  • Made the ligand optional in the FOX.recipes.generate_psf() recipe.
  • Fixed an issue where non-absolute distances were used when calculating the pressure.
  • Fixed an issue where non-charge parameters were updated incorrectly.
  • Fixed an issue where parameter guessing could fail if no .prm file was provided.

0.8.11

  • Fixed an issue where the .hdf5 status would not be properly cleared (if necessary).
  • Grab the cell parameters from previous jobs if applicable.
  • More thouroughly check the qmflows.Result status before accessing it.
  • Automatically determine appropiate chunk sizes when calculating the RDF.
  • Added a new sub-module dedicated to the calculation of properties: FOX.properties.
  • Fixed an issue where parsing the unit would fail when parameter guessing.
  • Fixed an issue where writing nan to armc.xyz.hdf5 would fail.

0.8.10

  • Reorganized the dataframes in FOX.armc.ParamMapping.
  • Added the pes_validation keyword. Used for constructing a set of PES descriptors for the purpose of validation.
  • Dissallow the specification of duplicate .yaml keys.
  • Allow users to provide unique settings for each molecule in PES-averaged ARMC.
  • Added tests using CP2K, python 3.9, pre-releases and the minimum supported version of dependencies.
  • Removed the plotting and .csv-related entry points.
  • Added the param.validation.enforce_constraints keyword; used for checking whether or not initially supplied parameters satisfy all user-provided constraints.
  • Added the unit field to the param_metadata dataset.
  • Added the new pes.block.ref keyword; used for constructing PES descriptors from qmflows.Result objects rather than the FOX.MultiMolecule instances constructed therefrom.

0.8.9

  • Fixed an issue where frozen parameters were not respected when performing contrained parameter updates.
  • Fixed an issue where the ARMCPT parameters weren't properly swapped.
  • Added two new ARMC options: param.validation.allow_non_existent and param.validation.charge_tolerance.
  • Update the type hints within Auto-FOX.
  • Allow the ARMC(PT) input to, once again, be dumped to a .yaml file via the armc2yaml command.
  • Dump more information into the ARMC logger.
  • Re-enable the ARMC restart option.

0.8.8

  • Added recipes for calculating time-resolved angular/radial distribution functions.
  • Various documentation-related updates.
  • The order in wich atom-pair/-triplet based parameters are provided is now irrelevant. For example Cd Se and Se Cd are now treated as equivalent, as well as O C H and H C O (but not O H C).
  • Fixed an issue where guessed parameters were not properly parsed.
  • Relaxed the PLAMS version requirement.
  • Log all local variables whenever an exception is encountered.
  • Move the ARMC test files to their own repo.
  • Export parameter metadata to the .hdf5 file.

0.8.7

  • Moved from PRMContainer.__dict__ to a PRMContainer.__slots__ based class structure.
  • Cleaned up the PRMContainer code; updated annotations, etc..
  • Removed assertionlib.AbstractDataClass as base class from PRMContainer.
  • Do not read or write comments to and from a .prm file.
  • Upped the minimum Sphinx version to 2.1.
  • Removed sphinx-autodoc-typehints.

0.8.6

  • Import AbstractFileContainer from Nano-Utils.
  • Removed TypeMapping in favor of TypedDict.
  • Remove travis in favor of GitHub Actions.

0.8.5

  • Moved a number of functions to the Nano-Utils Package.

0.8.4

  • Updated the ARMC documentation.

0.8.3

  • Updated the ARMC tests.
  • Renamed FOX.test_utils to FOX.testing_utils.
  • Added flake8 and pydocstyle to the tests.

0.8.2

  • Fixed and generalized the frocefield parameter guessing procedures (#100 and #112).
  • Log the optimum ARMC cycle in get_best() and overlay_descriptor() (#111).
  • Fixed an issue where certain ADF atom-subset-permutations were ignored (#110).
  • Aux error: Ensure that the summation over qm occurs row-wise (#108).

0.8.1

  • WiP: Introduction of the ARMCPT class.

0.8.0

  • Move all ARMC related modules to the new FOX.armc module.
  • Switched from plams.Job to qmflows.Package runners.
  • Introduced the PhiUpdater class for handling and updating the phi parameter.
  • Introduced the ParamMapping class for handling and updating the forcefield parameters.
  • Introduced the PackageManager class for handling the and managing the qmflows.Package instances, including the running of jobs.
  • Store the Auto-FOX __version__ in the .hdf5 file.
  • Changed the .yaml input to closer resemble the actual class structure.
  • Overhauled the .yaml input parsing.
  • Bumped the minimum Python version to 3.7.
  • Marked Auto-FOX as a typed package.
  • Added qmflows and noodles as new dependencies.
  • Added typing_extensions as a new dependency for Python < 3.8.

0.7.4

  • Increased the assertionlib version requirement to >= v2.1.
  • Generalized PRMContainer.overlay_cp2k_settings() to work for all forcefield parameters (potentially) specified in CP2K settings.
  • Added the PRMContainer.overlay_mapping() function for overlaying an arbitrary (nested) Mapping with the PSFContainer.
  • Added the TypedMapping class, a baseclass for creating typed Mappings, i.e. Mappings with a set number of specific keys.
  • Exchange a number of Settings.__contains__() operations for Settings.get().

0.7.3

  • Updated the default CP2K Settings template.
  • Employ more rigorous index-sorting when creating the .psf bonds, angles, dihedrals and impropers sections.
  • Fixed a bug where values reported by degree_of_separation() were incorrectly ordered when dtype != bool.
  • Added the shift_cutoff keyword for the calculation of all forcefield non-bonded potential energies. Sets the value of the potentials to zero at the specified distance_upper_bound.
  • Fixed an issue where sigma-values produced by .prm files were not properly parsed.
  • Fixed an issue where multiple potentials for a single set of dihedrals were not properly parsed.
  • Further miscellaneous improvements and fixes to the FOX.ff modules.

0.7.2

  • All forcefield related energies are now returned in their entirety, rather than averaging them with respect to the number of MD iterations.
  • Add recipes for the analyses of forcefield energies.
  • Increased the flexibility of the recipes.plot_descriptor() function.
  • #77 & #78: Combine NumPy vectorization with for-loops during the calculation of inter-/intra-ligand non-bonded interactions if array sizes start to exceed 100 million elements.
  • #78: Truncated distance matrices can now be used for the calculation of inter-ligand non-bonded interactions.

0.7.1

  • Renamed the csv module to csv_utils.
  • Fixed the previously broken MultiMolecule.delete_atoms() method.
  • Ensure that MultiMolecule._get_atom_subset() can handle all Iterables.
  • When assigning new bonds (MultiMolecule.bonds) all bond orders will default to 1 if not explicitly specified.
  • Cleaned up the LJDataFrame() class.
  • Implemented multiple bugfixes related to the calculation of intra-moleculair non-bonded interactions.
  • Lennard-Jones and Electrostatic scaling factors can now be applied for the calculation of 1,4 non-bonded interactions, similiar to the CP2K EI_SCALE14 and VDW_SCALE14 keywords.
  • Ensure that the IMPROPER / IMPROPERS .prm block is always written as IMPROPER. While both of them are in principle valid block-names, CP2K will only accept IMPROPER.
  • Introduced code style changes to the recipes.psf module.

0.7.0

  • Multiple updates to the FOX.ff modules:
  • Fixed a missing +1 addition in the calculation of the dihedral potential.
  • Wildcard atoms ("X") are now properly parsed.
  • 1,4-nonbonded interactions (intra-moleculair) are now calculated.
  • 1,3-nonbonded interactions (intra-moleculair), aka the Urey-Bradley terms, are now calculated.
  • Non-bonded interactions between explicitly specified atom-pairs are now calculated.
  • Fixed a number of issues introduced in #74.

0.6.21

  • Fixed an issue where a MultiMolecule() couldn't be converted into a Molecule().
  • Upped the version requirement from the assertionlib package to >= 2.

0.6.20

  • Cleaned up how PES descriptors are generated & stored in the ARMC() class.
  • Atom names specified in .PSF files are now accessible by MultiMolecule() instances during the ARMC procedure.
  • Generalized dekekulize() to work for all non-integer bond orders; not just 1.5.

0.6.19

0.6.18

  • Added the MultiMolecule.add_atoms() method.
  • Added a new recipe (FOX.recipes.ligands) for generating radial distribution functions using the center of mass of ligands (doc).

0.6.17

0.6.16

  • There is no v0.6.16.

0.6.15

  • Added recipes for generating .psf files in FOX.recipes.
  • #65 : Fixed a bug where ARMC parameter constraints were not properly parsed.
  • #66 : Added new ARMC tests.

0.6.14

  • Fixed an issue where valid .xyz files were not properly read during the ARMC procedure.
  • Added a precaution against reading faulty .xyz files.
  • Fixed an issue where some of datasets in the armc.xyz.hdf5 file were of incorrect shape.
  • Change the datatype from the armc.xyz.hdf5's datasets from np.float64 to np.float16 in order to reduce disk space.
  • Added a precaution against reading faulty .xyz files.
  • #60 : .hdf5 files are now forcibly closed (if necessary) upon restarting an ARMC procedure.
  • #61 : Updated the recipe examples; plot_descriptor() no longer crashes when encountering a DataFrame() with a single column.
  • #62 & #63 : Ensure that the ARMC restarting starts from the last iteration whose error is not np.nan.

0.6.13

  • Introduced a new logger; see #33.
  • Change the fillvalue of all float-based .hdf5 Datasets to np.nan.
  • Atoms and bonds are now, again, properly stored in the .xyz.hdf5 file.

0.6.12

  • The ARMC input parser no longer expects ARMC.param and the .psf file(s) to form identical sets.
  • All atomic charges in the ARMC .psf files are now set to 0.0. Charges are handled, exclusively, by the cp2k input file.
  • Fixed an issue where atom-types were not properly updated in the .psf file.
  • Fixed an issue where the ARMC .xyz.hdf5 file was not properly updated.
  • Ensure that None object encountered during the ARMC procedure are always converted into np.nan. Contrary to NumPy or Pandas, h5py will not automatically convert None to np.nan when assigning items to a Dataset.
  • Raise a RuntimeError if a job hard-crashes in the first ARMC iteration.
  • Always create a shallow copy of (to-be mutated) input parameters when calculating (forcefield-based) interactions.
  • Fixed the atom-pair hashing in get_bonded().
  • Prevent double counting non-bonded interactions when i == j in get_intra_non_bonded().
  • Potentials are now (properly) averaged over all molecules within an MD trajectory in get_intra_non_bonded().
  • Import scipy's fftconvolve() with a try/except approach; importing has a tendancy of raising RecursionErrors.
  • Log the super- & sub-iteration upon ARMC() restarts.

0.6.11

  • .psf files can now be directly supplied in the ARMC .yaml input.

From #52:

  • Added the option to estimate non-bonded parameters using either UFF or the RDF.
  • ARMC() instances can now be converted into ARMC.from_yaml()-compatible .yaml files. See the armc2yaml entry point.

0.6.10

  • Added the option to provide multiple .rtf files for state-averaged ARMC runs.

0.6.9

  • FOX.get_example_xyz() has been deprecated in favor of FOX.example_xyz.
  • Moved the psf_to_atom_dict() function to PSFContainer.to_atom_dict().

From #52:

  • Repos of script to analyze AMRC data.
  • Simultaneous fitting of different trajectories with different atom types; ensure that the PES descriptor generators can have different arguments for each trajectory.
  • Restart procedure for ARMC.

0.6.8

  • Added a workflow for calculating covalent intra-ligand interactions using harmonic- + cosine-based potentials: FOX.get_bonded(). Complementary to the in 0.6.4 introduced FOX.get_non_bonded().
  • Added a workflow for calculating non-covalent intera-ligand interactions using electrostatic + Lennard-Jones potentials: FOX.get_intra_non_bonded(). Complementary to the in 0.6.4 introduced FOX.get_non_bonded().
  • Added a number of useful workflows as stand-alone scripts.
  • Added the FOX.ff directory for all forcefield related modules.
  • Slimmed down the number of exposed functions and classess.
  • Changed PSFContainer._SHAPE_DICT and ._HEADER_DICT to instances of MappingProxyType().
  • Fixed a bug where some PSFContainer() dihedral angles where ordered incorectly.

0.6.7

  • FOX.estimate_lj() can now estimate sigma based on either the base or the inflection point of the first RDF peak.

0.6.6

  • Made Auto-FOX compatible with Python 3.6.
  • Added tests for Python 3.6 and 3.8.
  • Permanently moved a number of modules from (nano-)CAT to Auto-FOX.
  • Added the MutliMolecule.loc property; allowing for the slicing of MultiMolecule (directly) using atomic symbols. Usage examples: mol.loc['Cd'] and mol.loc['Cd', 'Se', 'O']. The Equivalent to mol[mol.atoms['Cd']].

0.6.5

  • Cleaned up the angular distribution code & atom subset code.
  • Added a module for constructing UFF Lennard-Jones parameters.
  • Added the option to specify constant parameter values in the ARMC .yaml input.

0.6.4

  • Updated the read_prm module.
  • Added a workflow for calculating inter-ligand and core/ligand interactions using electrostatic + Lennard-Jones potentials: FOX.get_non_bonded().

0.6.3

  • Added a function, FOX.estimate_lj(), for estimating Lennard-Jones parameters using radial dsitribution functions.

0.6.2

0.6.1

  • Added an additionl memory consumption level to MultiMolecule.init_rdf().
  • Ensure that the 'constraints' column is always present in the ARMC parameter DataFrame.
  • _xyz_to_hdf5() no longer crashes when mol_list=None.
  • Switched the AssertionLib package from GitHub to PyPi.

0.6.0

0.5.0

Added

  • Added the option for state-averaged ARMC parameter optimizations; i.e. simultaneously optimizing a single parameter set based on the auxiliary error of multiple MD trajectories.
  • [reprlib](https://docs.python.org/3/library/reprlib.html) is now used in MultiMolecule.__str__().

Changed

  • Updated all module-level docstrings. Now includes an autosummarry and autodoc description of the module.

0.4.4

Added

  • Added new methods for constructing the velocity autocorrelation function (VACF), .MultiMolecule.get_vacf, and VACF-derived power spectra, .MultiMolecule.init_power_spectrum.

0.4.3

Added

  • Generation of angular distribution functions, .MultiMolecule.init_adf, is now conducted in parallel if DASK is installed.
  • A distance cutoff can now be specified in .MultiMolecule.init_adf.

Changed

0.4.2

Added

  • Minimum and maximum allowed values can now be specified for all ARMC paramaters.
  • Added a commandline interface for generating and exporting plots & .csv files.
  • Added a function for translating strings to callable objects.

Changed

  • Split the armc.job.settings block into .job.md_settings & .job.preopt_setting.
  • Removed the unused FrozenSettings class.
  • Further generalized the param section; a path of keys now has to be specified for each block.
  • Removed a couple of unused functions.
  • Cleaned up the ARMC input parsing; now utilizes schema.
  • Updated many docstrings with examples.

0.4.1

Added

  • Potential energy surfaces, over the course of last ARMC super-iteration, are now stored in .hdf5 format.
  • Added increased control over the non-bonded inter-atomic potential.

Changed

  • Molecular dynamics (MD) jobs are now preceded by a geometry optimization.
  • MD simulations can now be skipped of the geometry optimization RMSD is too large.
  • Docstrings changed to NumPy style.
  • Cleaned up the AMRC code.
  • Comments in .xyz files are now parsed.

0.4.0

Added

  • Added an entry point for accessing .ARMC.init_armc.
  • Expanded io-related capabilities, including the option to read KF PDB, PSF, PDB and PRM files.

Changed

  • Formatting of docstrings in accordance to PEP257.
  • Implementation of type hints. Support for python versions prior to 3.7 has been dropped.
  • .ARMC was moved to its own seperate modules.

0.3.2

Added

  • Simplified the FOX.ARMC input and updated its documentation. (see #33)
  • Added 2 new methods to the FOX.MultiMolecule class for calculating average and time-averaged atomic velocities.
  • Added 2 modules for handling atomic charges and .prm files.

0.3.1

Added

  • Added new tests for the FOX.MultiMolecule class. (see #18)

Changed

  • Minor style changes to the documentation and the .xyz reader.
  • The FOX.MultiMolecule has been changed into a np.ndarray subclass. (see #30)

0.3.0

Added

  • Wrapped up implementation of the Monte Carlo forcefield optimizer. (see #17)

0.2.3

Added

  • Introduced two new methods to the FOX.MultiMolecule class for identifying shell structures in, e.g., nanocrystals or dissolved solutes. (see #29)

0.2.2

Added

  • Introduced an angular distribution generator in the MultiMolecule class.

Changed

  • Fixed a renormalization bug in the 0.2.1 improved get_rdf() function.

0.2.1

Added

  • Introduced new FOX.MutliMolecule methods for slicing MD trajectories.
  • Added the MonteCarlo API to the documentation.
  • WiP: Split the MonteCarlo class into 2 classes: MonteCarlo & ARMC (subclass).

Changed

  • Minor update to copy/deepcopy-related methods.
  • Improved the get_rdf() function.

0.2.0

Added

  • Added a root mean squared displacement generator (RMSD).
  • Added a root mean squared fluctuation generator (RMSF).
  • Introduced the FOX.MultiMolecule class for handling and storing all atoms, bonds and coordinates.

0.1.0

Added

  • Added a reader for multi-xyz files.
  • Added a radial distribution functions generator (RDF).

[Unreleased]

Added

  • Empty Python project directory structure.