Dpgen

New features:
- Support gromacs in model deviation (427)
- Support DeePMD-kit v2 (420)
- Added ABACUS interface for first principle calculations (433)
- Allow assigning different activation functions to embedding and fitting nets (461)

Enhancement
- Remove the support for deepmd-kit v0.x (397 )
- Support for new dpdispatcher (449)
- Example machine file for 4 GPUs on a local machine (400 412 )
- Better error message in too many failed FP jobs (428 )
- Slurm customized flags (368 )
- ssh session is not active (382 )
- Retry slurm commands when failing (381 )
- Store sftp and reduce the frequency of opening sftp (387 )
- Taskwise restart for manual cuda device (383 )
- Add Cu's init_surf examples. (406)
- Add input files of enhanced sampling method (446)
- Update CP2K new example and slurm-lsf-mix machine (456)

Bug fixings:
- PBS error message (396 )
- Bug `inter_type` in auto_test Lammps support (423 )
- Bug in auto_test multicomponent surface systems (378 )
- Too long argument list for `tar` command (380 )
- force overwrite when using `gzip` (384 )
- MPI execute error (405 )
- Support hybrid descriptor (fix 458) (459)
- Fix parameter error for LSF GPU tasks (463)
- Fix NSW value in INCAR; fix local_root check (464)

Bug fixing:
- stop printing `pytmatgen.__version__` that is not supported by version 2022

Bug fixing:
- Solve the comparability issue of the pymatgen v2022.0.* 358

New feature:
- New autotest

Improvement:
- Improve support for batch systems 303 313 337 341 352
- Support different template settings for different systems 328 346
- Support cp2k template 348
- Support restarting in lammps 335 336

Bug fixings
- 332 334 340 342 343 344 345

New feature:
- support stress in cp2k

Bug fixings:
several know issues

New features:
- support pwmat
- support nopbc
- support init model parameter from old models
- support aniso kspacing for vasp
- support plm_path that defines CV in a separate file
- support checking bad fp confs
- support collect data
- support initializing from an existing model
- support simplify for LabeledSystem and System
- with accurate thresthold, decay number of fp tasks to 0 if it is already accurate enough
- support different activation func for different models

Updates:
- set default multiplicity to auto; add parameter fragment_guesses
- better suppport on deepmd-kit >= v1.1
- tag failures on allow_failure
- better cvasp support
- better ali support
- restart shell by uuid
- restart training if interrupted
- retry slurm commands when failing

Bug fixings:
- error in checking slurm status
- error in ensuring alive
- shell submit script should exit with code 1
- transfer type_map.raw. enforce the the same type_map in the training input.json
- bad surf configurations
- check if the length of batch size list is enough
- auto test user provided INCAR
- getting nframes in make_train
- fix issue 247 add = to flags starting with --
- nve doesn't set the initial temperature