- refactor slice implementation
- improvement: options to slice traj in model_devi
- fix: inplace should support ignore_error
- fix: model_devi read file
- improvement: options to ignore model_devi error
- fix: add model_devi tool to entry point
- breaking: modify model deviation tool
- fix: ignore error data in dpdata read
- fix: ignore error data in dpdata read
- fix: filter in model deviation tool
- fix: dplr in ase tool
- fix: make viber adapted for dp v3
- feat: support wannier charge != -8 (via
model_charge_map
) - fix: broken dp_train_opts options
- fix: wrong filter in model devi tool
- feat: model deviation toolkit
- allow to override potential in lasp
- implement dplr and refactor dpff
- feat: handle cp2k/dplr in dpdata tool
- fix: dpdata tool wrong params
- fix: vasp wrong params
- fix: vasp wrong params
- improvement: ignore cutoff in viber
- improvement: vasp support attribute variable
- improvement: new ensable for lammps
- fix: pydantic 2 broken issue
- feat: extra options for dp train command
- add netCDF
- upgrade pydantic to 2.8.2
- fix: hpc tool typo
- feat: support use multiple gpus in the same batch job
- improve: dpdata: allow user to ignore broken data
- fix: dpdata eval wrong type mapping
- improve:add label to fes graph and dump result to json file
- fix cp2k-inc for ase write_frames
- fix out of date documents
- tesla: generate single dump file in lammps
- tool: refactor frame tool
- frame file process command line tool
- command line interface for reweighting algorithm
- deepmd: support
input_modifier_fn
- lammps: use relative path to reference plumed input
- fix lammps data util for dplr
- fix cp2k cell digit limit
- refactor: ase write_each_frame to write_frames
- feat: reweighting algorithm
- feat: dpdata tool support eval
- improve: support covert between ase and dpdata format
- chore: replace
pylint
withruff
- fix: redefine variable in for-loop
- fix: anyware execution
- fix: anyware output artifact
- fix: post_process_fn in anyware
- feat: cmd to write dplr lammps data
- feat: anyware for explore
- breaking change: new checkpoint system
- cp2k: use fp32 for input data
- lammps: support custom force field via
custom_ff
option
- ai2-cat: refine the basis set and potential selection
- ai2-cat: support multiple basic set and potential files
- ai2-spectr: fix wrong atom order
- fix hpc tool
- fix deepmd restart
- improve gen lammps script tool
- fix loading cp2k viber data
- fix ai2cat cp2k data path
- fix ai2spectr entry
- fix dpdata tool
- fix dpdata read
- fix fep pka bugs
- fix dpdata wrong arguments
- improve fep pka mode
- fix edge case of efield
- for dflow-galaxy: reference lammps data file use relative path
- feat: tool to format cp2k input
- hotfix: remove cp2k wfn if warmup step is failed
- hotfix: executor upload/download issue
- feat: temporary support for init from previous model
- refactor: make
asaplib
optional
- refactor: python remote execution solution
- feat: upload local python package to remote
- improvement: move some dependencies to optional
- improvement: refactor slurm done file
- improvement: no need to return in wait method for slurm tool
- feat:
ai2-kit tool hpc
for managing HPC jobs
- improvement: add
ignore_error
for deepmd.
- improvement: add
ignore_error
for cp2k.
- fix: wrong lammps template for fep training
- improvement: select structure for next round from good and decent structures
- fix: wrong lammps template for fep training
- improvement: optimize job running indicator
- improvement: allow to override fix_statement and ensemble in artifact
- fix: rsplit max should be 1
- improve: don't keep good and poor structures to save storage
- feat: support max_decent_per_traj
- fix: wrong softlink path
- improvement: create softlink for dw_model
- fix: wrong lammps template_vars
- fix: fail to load config when it contains key starts with
- fix: wrong lammps template_vars
- feat: support
dpff
training mode - improvement: strict check of workflow configuration
- fix: wrong specorder of
fep-pka
mode - breaking:
mode
move to general config section - breaking: atom id of
DELETE_ATOMS
offep
mode now start from 1 instead of 0 (the same as lammps) - breaking: require
deepmd-kit
when usedpff
mode
- fix: incomplete lammps template_vars
- fix: pickle issue
- fix: typo
- feat: cll-training-workflow support
fep-redox
mode, details here - feat: dpdata support
set_fparam
- fix: ensure no duplicate dataset in deepmd input
- fix: lammps vars should also be template vars
- improvement: plumed config support lammps template vars, e.g
$$TEMP
- feat: batch toolkit
- fix: openpbs workdir
- feat: support pbs job scheduler
- fix: dpdata filter
- refactor: aos-analysis options
- fix: dp train should support restart
- chore: update ai2cat default value
- chore: update ai2cat workflow template
- feat: Amorphous Oxides Structure Analysis toolkit, for more detail please check here
- fix: scale factor of fep lammps template
- feat: support customize fix statement
- fix: formily schema
- fix: ai2cat lammps template
- feat: release of ai2cat notebook
- fix: missing files
- feat: interactive ui for generate lammps input
- feat: yaml tool
- feat: ai2cat notebook
- chore: upgrade jupyter-formily
- wip: ai2cat tool and notebook
- fix: asap log error
- feat: ase tool support
set_by_ref
andlimit
method - feat: integrate
jupyter-formily
- wip: ai2cat tool and notebook
- fix: deepmd data path
- feat: deepmd option to group data by formula
- fix: fep lammps template
- fix: copy dataset in deepmd
- fix: fep lammps template
- fix: cp2k potential/basic-set parser
- fix: lammps config validator
- fix: fep lammps template
- fix: squeue options
- feat: support
broadcast_vars
in explore step, for more detail please check here
- refactor: deprecate
type_order
for deepmd >= 2.2.4, more detail please check here - fix: HPC executor resubmit error
- fix: asap selector
- fix: doc typo
- fix: inject locals to eval
- fix: selector screening function
- improvement: avoid unnecessary job submission
- fix: plumed template
- fix: deepmd template
- fix: cache for py38
- feat: new ai2-cat commands
- fix: drop deepmd v1 config format support as it is not compatible with v2.2.5
- feat: cll-workflow support sort structure by energy, for more detail please check here
- feat: cll-workflow support screening by energy, for more detail please check here
- feat: cll-workflow deepmd support fixture_models, for more detail please check here
- feat:
ai2cat
support gen plumed input.
- feat: handle outlier data in deepmd training stage, for more detail please check here
- feat: dpdata tool support filter systems by lambda expression, for more detail please check here
- improvement: cll-workflow now allow user to change init_dataset in each iteration.
- feat:
ase
toolkit support save atoms ascp2k-inc
file, which can be used as@include coord_n_cell.inc
macro in CP2K input file.
- feat:
ai4cat
toolkit, for more detail please check here
- improvement: sorted input files for dpdata and ase tools
- You may need to quote the input file name or else the file expansion will be handled by shell. e.g.
ai2-kit ase tool read './path/to/*.xyz'
- You may need to quote the input file name or else the file expansion will be handled by shell. e.g.
- fix: ase tool read data
- improvement: ignore
squeue
error - fix: upgrade
dpdata
to 0.2.16 to support CP2K 2023.1
- fix: condition of selecting new explore systems
- feat: update explore systems for each iteration,for more detail please check 1, 2
- This version breaks the compatibility of checkpoint file, you may remove the checkpoint file before running this version.
- improvement: allow user to ignore lammps error
- fix: itemgetter bug: https://stackoverflow.com/a/48988896/3099733
- BREAKING CHANGE: cp2k input template no longer support
dict
, system data should be loaded via@include coord_n_cell.inc
macro. More detail here
- fix: cp2k input file generation bug
- feat: [cll-workflow] support wave function warmup for cp2k, more detail here
- improvement: [cll-workflow] raise invalid artifact error earlier
- fix: ase tool arguments
- improvement: [cll-workflow] save good/decent/poor structures in xyz file.
- feat: implement dpdata tool, more details here
- fix: data handling bug
- feat: Support training DP model for FEP simulation
- feat: ase tool support delete_atoms
- BREAKING CHANGE: [cll-workflow] LAMMPS configuration has been change, see example
- improvement: change the squeue polling command
- feat: support
load_text
andload_yaml
tag in yaml parser - doc: config samples for cll-mlp-training workflow
- fix: model_devi_decent.xyz not created when no dumps are selected
- fix: atoms from lammps-dump-text not ordered by id
- fix: sort atoms before exporting as POSCAR
- improvement: suppress numba warning
- improvement: checkpoint rm support exclude pattern
- BREAKING CHANGE: [cll-workflow] checkpoint file may not compatible with < v0.5.0
- feat: select distinct structures by global descriptor grouping
- BREAKING CHANGE: [cll-workflow] checkpoint file may not compatible with < v0.4.0
- BREAKING CHANGE: [cll-workflow] config file change:
select.by_threshold
=>select.model_devi
- feat: [cll-workflow] support explore with LASP
- refactor: [cll-workflow] data conversion system
- chore: asap tool integration
- refactor: list_sample method
- fix: use len(dp_system) > 0 to filter broken data
- fix: ignore data merge when SCF is not converged
- refactor: simplify data handling
- fix: slurm job checkpoint key
- fix: cp2k command line
- fix: use override instead of append for cp2k output
- improvement: checkpoint tool command line
- fix: slurm job checkpoint key
- improvement: pretty table output
- fix: create missing dir
- fix: cpk2 command
- improvement: cp2k support post_cp2k_cmd option
- improvement: selector will save result in csv format
- fix: dp empty validation data error
- doc: generate online doc
- chore: github workflow
- improvement: support random sampling if selected structures hit limitation
- fix: add type_map when converting data to deepmd/npy
- feat: distance analysis for proton transfer
- fix: convert vasp data for deepmd
- improvement: sample list evenly instead of truncate when selecting new structures
- fix: use max_devi_f instead of avg_devi_f to select structures
- fix: label all init data
- feat: label all init data
- feat: support specify deepmd validation data
- fix: model deviation selection rules
- improvement: cll: stop workflow if no more data is generated
- refactor module names
- fix: pass down ancestor attributes to child artifacts
- chore: remove cp2k-input-tools
- feat: implement cpk2 parse module
- fix: raise error on job failure
- feat: ase tool support set_cell and set_pbc commands
- feat: support VASP in cll training workflow
- feat: support
compress_model
option in deepmd step
- feat: implement read tag for yaml parser
- feat: support plumed config in cll workflow
- feat: support artifacts attributes inheritance in workflow
- feat: implement join tag for yaml parser
- feat: apply checkpoint to all workflows and job submission steps
- feat: add checkpoint command line interface
- refactor: using async.io to implement tasks and workflows
- improvement: support relative work_dir
- feat: async job polling
- feat: ase toolkit
- fix: fep training workflow
- improvement: speed up command line interface with lazy import
- feat: proton transfer analysis toolkit
- feat: cll-mlp-training workflow
- feat: fep-mlp-training workflow
- feat: support providing initial structures for label task at first iteration.
- feat: support dynamic configuration update for each iterations.
- improvement: reduce size of remote python script by compressing it with bz2.
- fix: ssh connection closed when running large remote python script.
- refactor: design common interface for CLL workflow.
- refactor: use ase.Atoms as the universal in-memory data structure.
- refactor: use ResourceManager to manage executors and artifacts.
- feat: utilities to execute and manage jobs in local or remote HPC job scheduler.
- feat: utilities to run python functions in remote machines directly.
- feat: utilities to simplified automated workflows development with reusable components.
- feat: fundamental CLL & FEP workflows.