Skip to content

changes.md

Latest commit

 

History

History
154 lines (127 loc) · 9.03 KB

changes.md

File metadata and controls

154 lines (127 loc) · 9.03 KB
layout default
title Change Log
nav_order 2

Change Log

v0.5.1

  1. CI hygiene: drop dgl install pin from all GitHub Actions workflows (matgl now resolves its own backend); fix release.yml permissions/secrets: inherit so the reusable test.yml call satisfies the subset rule and PyPI trusted publishing works.

v0.5.0

  1. RelaxCalc convergence checks.
    • RelaxCalc.calc surfaces is_converged and max_force in its result dict.
    • PropCalc._prerelax raises RuntimeError when the pre-relaxation fails to converge, instead of silently feeding an unrelaxed structure into downstream property calcs.
    • RelaxCalc now copies the structure before perturbing, eliminating a caller-mutation hazard.
  2. PR #218 Extract PropCalc._prerelax helper; fold relax boilerplate by @shyuep
    • Centralises the if self.relax_structure: relaxer = RelaxCalc(...); result |= relaxer.calc(structure); structure = result['final_structure'] pattern duplicated across _phonon3, _eos, _qha, _md, _stability, and _elasticity.
  3. PR #210 / #85 Fix youngs_modulus and add _units metadata across all PropCalcs by @shyuep
    • Self-consistent youngs_modulus computation; every PropCalc result now carries a _units dict, with assertions in the test suite.
  4. Phonon3Calc adapted to the new phono3py conductivity API.
  5. Foundation potential model naming unified. UMA model names updated to match FAIRChem upstream; CHGNet alias typo fixed; small registry cleanups.

v0.4.8

  1. Updated the Hugging Face organization to materialyze (lower case).

v0.4.7

  1. ElasticityCalc can now return moduli in GPa.
    • Added a new units_GPa keyword argument (default False). When set to True, the returned elastic_tensor, bulk_modulus_vrh, shear_modulus_vrh, and residuals_sum are converted from pymatgen's native eV/A^3 units to GPa.
  2. Bug fix in EOSCalc.
    • Fixed a variable shadowing issue where the loop-local relaxation result was overwriting the outer merged result dict, causing the merged relaxation/structure fields returned by EOSCalc.calc to come from a strained calculation rather than the originally relaxed input structure.

v0.4.6

  1. PR #199 Add GBCalc Class for Grain Boundary Calculation (originally PR #111)
    • Author: @rul048
    • Description: Adds GBCalc for generating grain boundary structures and computing grain boundary energies, complementing the existing SurfaceCalc and InterfaceCalc classes.
  2. PR #203 Multiple pressure support in QHACalc
    • Author: @Andrew-S-Rosen
    • Description: Adds support for evaluating multiple pressures in the quasi-harmonic approximation workflow.
  3. PR #174, #176, #163 Phonon and QHA refactor
    • Author: @Andrew-S-Rosen
    • Description: Significant refactor of PhononCalc and QHACalc for clearer logic, more consistent relaxation settings, better error handling, and improved typing.
    • Description: Adds detection and improved handling of imaginary phonon modes.
    • Description: Fixes PhononCalc handling of Atoms inputs and supercell generation, and exposes a min_length kwarg for supercell sizing.
    • Description: Allows users to toggle symprec in Phonopy calls and sets primitive_matrix="auto" by default.
  4. PR #158 Consistent relaxation settings in QHACalc
    • Author: @Andrew-S-Rosen
    • Description: Relaxes atomic positions in the QHA workflow and ensures consistent settings between relaxations.
  5. PR #179 Allow fixing symmetry or specific atoms during relaxation
    • Author: @rul048
    • Description: Adds support in RelaxCalc for fixing symmetry or specifying particular atoms to be held fixed during relaxation.
  6. PR #161 / #206 Better handling when relaxations hit max steps
    • Author: @Andrew-S-Rosen
    • Description: Logs (instead of warning) when a relaxation reaches its maximum step count, making the behavior easier to track in long-running workflows.
  7. PR #189 Remove upper triangular cell call in NVT ensemble
    • Author: @Andrew-S-Rosen
    • Description: Avoids unnecessary cell transformation in MDCalc NVT runs.
  8. PR #136 Default NEB method to improvedtangent
    • Author: @rul048
    • Description: Changes the default NEB method to improvedtangent for more reliable minimum energy paths.
  9. Replaced the custom TrajectoryObserver with ASE's native trajectory observer.

v0.4.5

  1. PR #117 Add InterfaceCalc Class for Interface Structure and Energy Calculation
    • Author: @so2koo
    • Description: Introduces InterfaceCalc for calculating interface structures and energies, similar to SurfaceCalc. Utilizes pymatgen CoherentInterfaceBuilder to generate interface structures from two different bulk structures. Employs RelaxCalc for structure relaxation and determines the lowest interface energy.
    • Todos: Further work may be needed if this is a work in progress.
  2. PR #118 Add AdsorptionCalc Class for Adsorption Energy Calculation
    • Author: @mkphuthi
    • Description: Adds AdsorptionCalc for calculating adsorption energy of single molecules on surfaces. Capabilities include generating slabs with SlabGenerator, placing adsorbates with AdsorbateSitefinder, and directly supplying an adslab. Features include calculation of adsorption energies, support for customizable relaxation via RelaxCalc and ASE optimizers, extensive documentation, and type hints.
    • Todos: Some unrelated tests need to be fixed.

v0.4.4

  1. More data-efficient MEP class for storing NEB results.

v0.4.3

  1. PR #102 fix: add required-environments to enable Linux install by @MorrowChem
  2. PR #97: Make it possible to set COM momentum and/or rotation to zero in MDCalc by @Andrew-S-Rosen
    • Feature: Allows setting the center-of-mass momentum and angular momenta to zero in MDCalc to prevent drift.
  3. PR #96: Allow for MDCalc to be run with relax_cell=True during the initial relaxation by @Andrew-S-Rosen
    • Feature: Enables unit cell relaxation before an MD run.
  4. PR #89: Add support for MTKNPT by @Andrew-S-Rosen
    • Feature: Added support for MTKNPT, the recommended method for NPT simulations in ASE.
  5. PR #90: Add a warning if using NPT by @Andrew-S-Rosen
    • Warning: Raises UserWarning when NPT class is used due to stability concerns.
  6. PR #88: Add IsotropicMTKNPT to _md.py by @Andrew-S-Rosen
    • Feature: Introduced IsotropicMTKNPT support, expanding supported ensemble options.
  7. Progress bars when running benchmarks.

v0.4.2

  • Bug fix for surface calculations (@computron).
  • Update OCPCalculator with the newer FairChemCalculator (@atulcthakur)

v0.4.1

  • Bug fix for bad trajectory snapshotting in MDCalc.
  • Beta LAMMPSMDCalc.

v0.4.0

  • All PropCalcs now support ASE Atoms as inputs, as well as pymatgen Structures.
  • Added MDCalc class for molecular dynamics simulations. (@rul048)
  • Minor updates to EquilibriumBenchmark to make it easier to reproduce matpes.ai/benchmarks results. (@rul048)

v0.3.3

  • SurfaceCalc class for computing surface energies. (@atulcthakur)
  • NEBCalc now returns MEP information. (@drakeyu)
  • All PropCalcs and Benchmarks now support using a string as a calculator input, which is automatically interpreted as a universal calculator where possible. This greatly simplifies the use of PropCalc.
  • PESCalculator.load_universal now is lru_cached so that the same model does not get loaded multiple times.

v0.3.2

  • Added Phonon3Calc for calculation of phonon-phonon interactions and thermal conductivity using Phono3py. (@rul48)

v0.3.1

  • All PropCalc implementations are now private and should be imported from the base package.
  • Added support for Mattersim, Fairchem-Core, PET-MAD, and DeepMD (@atulcthakur)

v0.2.2

  • Added ChainedCalc helper class to performed chaining of PropCalcs more easily.
  • Added matcalc.load_fp alias for easier loading of universal MLIPs.

v0.2.0

  • Major new feature: Most PropCalc now supports chaining. This allows someone to obtain multiple properties at once.
  • SofteningBenchmark added. (@bowen-bd)
  • Major expansion in the number of supported calculators, including Orb and GRACE models. (@atulcthakur)

v0.1.2

  • Emergency bug fix for bad default perturb_distance parameter in Relaxation.

v0.1.1

  • Provide model aliases "PBE" and "R2SCAN" which defaults to the TensorNet MatPES models.
  • New CLI tools to quickly use prop calculators to compute properties from structure files.

v0.1.0

  • Added support for ORB and GRACE universal calculators (@atulcthakur)
  • Option to perturb structure before relaxation (@rul048)
  • Improved handling of stress units (@rul048)
  • Option to relax strained structures in ElasticityCalc (@lbluque)

v0.0.6

  • Checkpointing and better handling of benchmarking.
  • Most PropCalc can now be imported from the root level, e.g., from matcalc import ElasticityCalc instead of the more verbose from matcalc.elasticity import ElasticityCalc.

v0.0.5

  • Initial release of benchmarking tools with Elasticity and Phonon benchmark data.

v0.0.2

  • Minor updates to returned dicts.

v0.0.1

  • First release with all major components.