Deepmd-kit2.2.7 version errors during training

[919607]

Hello teachers, I am currently experiencing two problems when training with dpkit2.2.7.
Firstly, I got the following error while training the potential function with dpkit2.2.7, but when I put the same input.json file in another server to train dpkit2.2.7, it is trained normally, both servers use the same method to compile dpkit2.2.7.
Secondly, I found that when the potential function trained by dpkit2.2.3 was used for performance calculation using the lammps carried by the dpkit226 version, the calculated output structure was completely wrong, and the structural atoms flew directly. The model used by the 223 version was se_atten.

Traceback (most recent call last):
  File "/public/home/mgi/local/deepkit-2.2.7-gpu/bin/dp", line 10, in <module>
    sys.exit(main())
  File "/public/home/mgi/local/deepkit-2.2.7-gpu/lib/python3.10/site-packages/deepmd_cli/main.py", line 635, in main
    deepmd_main(args)
  File "/public/home/mgi/local/deepkit-2.2.7-gpu/lib/python3.10/site-packages/deepmd/entrypoints/main.py", line 74, in main
    train_dp(**dict_args)
  File "/public/home/mgi/local/deepkit-2.2.7-gpu/lib/python3.10/site-packages/deepmd/entrypoints/train.py", line 146, in train
    jdata = normalize(jdata)
  File "/public/home/mgi/local/deepkit-2.2.7-gpu/lib/python3.10/site-packages/deepmd/utils/argcheck.py", line 2007, in normalize
    base = Argument("base", dict, gen_args())
  File "/public/home/mgi/local/deepkit-2.2.7-gpu/lib/python3.10/site-packages/deepmd/utils/argcheck.py", line 1778, in gen_args
    model_args(),
  File "/public/home/mgi/local/deepkit-2.2.7-gpu/lib/python3.10/site-packages/deepmd/utils/argcheck.py", line 855, in model_args
    pairwise_dprc(),
  File "/public/home/mgi/local/deepkit-2.2.7-gpu/lib/python3.10/site-packages/deepmd/utils/argcheck.py", line 987, in pairwise_dprc
    qm_model_args = model_args(exclude_hybrid=True)
  File "/public/home/mgi/local/deepkit-2.2.7-gpu/lib/python3.10/site-packages/deepmd/utils/argcheck.py", line 863, in model_args
    Argument("type_map", List[str], optional=True, doc=doc_type_map),
  File "/public/home/mgi/.local/lib/python3.10/site-packages/dargs/dargs.py", line 162, in __init__
    self._reorg_dtype()
  File "/public/home/mgi/.local/lib/python3.10/site-packages/dargs/dargs.py", line 211, in _reorg_dtype
    self.dtype = {dt if type(dt) is type else type(dt) for dt in self.dtype}
  File "/public/home/mgi/.local/lib/python3.10/site-packages/dargs/dargs.py", line 211, in <setcomp>
    self.dtype = {dt if type(dt) is type else type(dt) for dt in self.dtype}
  File "/public/home/mgi/local/deepkit-2.2.7-gpu/lib/python3.10/typing.py", line 312, in inner
    return func(*args, **kwds)
  File "/public/home/mgi/local/deepkit-2.2.7-gpu/lib/python3.10/typing.py", line 1058, in __getitem__
    _check_generic(self, params, len(self.__parameters__))
  File "/public/home/mgi/local/deepkit-2.2.7-gpu/lib/python3.10/site-packages/typing_extensions.py", line 152, in _check_generic
    raise TypeError(f"{cls} is not a generic class")
TypeError: typing.List[str] is not a generic class

[njzjz]

The error message shows that your dargs package comes from a different directory, which might be incompatible.
The minimal dargs version is bumped to version 0.4.1 in #2945 (released in v2.2.7).

[919607]

Thank you very much for your suggestion. After checking, it was found that there are multiple versions of the dargs package. After 'pip uninstall' one of the packages, this issue was resolved