dpdata doesn't read the CP2K trajectory

[hanao2]

Hello,

I want to convert my CP2K trajectory to DeePMD raw data using dpdata. However, when using t=dpdata.LabeledSystem("output", fmt="cp2k/aimd_output"), I get the following:

Data Summary
Labeled System
-------------------
Frame Numbers      : 0
Atom Numbers       : 0
Including Virials  : No
Element List       :
-------------------

My trajectory contains 4 frames; therefore, there is some problem going around with my dpdata processing. I have 4 files named "cp2k.log", "IL-1.ener", "IL-frc-1.xyz" and "IL-pos-1.xyz" in the "output" directory and I tried to replicate the necessary keys from dpdata's tests/examples in my input file ("input.txt" which is attached here). The versions of dpdata and cp2k are 0.2.14 and 2022.1, respectively. Could you please help me with fixing the issue?

Thanks!
input.txt

[robinzyb]

try installing the cp2kdata package and use fmt='cp2kdata/md'. see #439 for detail

[hanao2]

Hello,

Thanks for your response. Upon using the 'cp2kdata/md' format, I get the following error:

import dpdata
import cp2kdata
t=dpdata.LabeledSystem('output', 'cp2kdata/md')
--- You are parsing data using package Cp2kData ---
Obtian Energies From output/IL-1.ener
Obtian Structures From output/IL-pos-1.xyz
Obtian Froces From output/IL-frc-1.xyz
Missing the atomic kind informations, atom names are true chemical symbols.
No cell information, please check if your inputs are correct.
--- You are parsing data using package Cp2kData ---
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/hana/.local/lib/python3.8/site-packages/dpdata/system.py", line 283, in __init__
    self.from_fmt(
  File "/home/hana/.local/lib/python3.8/site-packages/dpdata/system.py", line 320, in from_fmt
    return self.from_fmt_obj(load_format(fmt), file_name, **kwargs)
  File "/home/hana/.local/lib/python3.8/site-packages/dpdata/system.py", line 1190, in from_fmt_obj
    self.check_data()
  File "/home/hana/.local/lib/python3.8/site-packages/dpdata/system.py", line 307, in check_data
    dd.check(self)
  File "/home/hana/.local/lib/python3.8/site-packages/dpdata/system.py", line 118, in check
    raise DataError(
dpdata.system.DataError: Type of cells is NoneType, but expected ndarray

How can I fix it?

Thanks!

[robinzyb]

hello,
dpdata.LabeledSystem('output', 'cp2kdata/md') should be

import dpdata
cp2kmd_dir = "."
cp2kmd_output_name = "output"
dp = dpdata.LabeledSystem(cp2kmd_dir, cp2k_output_name=cp2kmd_output_name, fmt="cp2kdata/md")
print(dp)

https://github.com/robinzyb/cp2kdata please check out the readme in this repository

[robinzyb]

DeepmdData resets the natom if type_sel is not None.
https://github.com/deepmodeling/deepmd-kit/blob/45a8d1e3a11f5c4621254ea9552015f724e5593c/deepmd/utils/data.py#L537-L540
and See the tensor model for the type_sel:
https://github.com/deepmodeling/deepmd-kit/blob/3ac8c4c77b61ee4c645cfb59eaefb104d73c0309/deepmd/loss/tensor.py#L59
and the type_sel reads from https://github.com/deepmodeling/deepmd-kit/blob/3ac8c4c77b61ee4c645cfb59eaefb104d73c0309/examples/water/dplr/train/dw.json#L28

It seems to me that if we store the atomic_dipole as natoms=192 for 64 H2O to file atomic_dipole.npy. and actually the natoms reads 64 if we set sel_type=[0] in input.json. Then this line may cause errors. https://github.com/deepmodeling/deepmd-kit/blob/45a8d1e3a11f5c4621254ea9552015f724e5593c/deepmd/utils/data.py#LL556C1-L556C1

[njzjz]

Interesting to know this behavior, which is a bit complicated for dpdata.

[hanao2]

Hello all,

Thanks for the ongoing discussion. Are you trying to add this feature (conversion of atomic dipoles) to the package?

Thanks!

[hanao2]

Hello,

I managed to write a python script to convert the WC output into molecular dipoles. However, in future development, it would be nice to include this feature in the dpdata package. Thanks for your time and help!

[robinzyb]

Hello,
It is great that you have figured it out. I will try to include it in the near future!

[hanao2]

Hello,

Thanks again for your help with my previous issue on converting CP2K AIMD data. Currently, I want to convert a single frame using the 'cp2k/output' format. However, I get the following error: "Shape of forces is (1, 34, 3), but expected (1, 33, 3)"

Do you think if this is caused by cp2k version inconsistency? If not, how can I solve it? I attach the log file for your notice.

Thank you!

cp2k.log

[njzjz]

Should be fixed by #466

[hanao2]

Thanks! it solved the issue. However, I don't see the virial file at the output. any suggestions on how to get it?

[robinzyb]

I tested the log file with cp2kdata.

x=dpdata.LabeledSystem("./cp2k.log", fmt="cp2kdata/e_f")
x.to_deepmd_npy("cp2k_dpmd")

It works and has an virial file

[robinzyb]

Can We close this discussion and open a new one, I think this discussion has been mixed with different problems. @njzjz