Some useful tools related to Amber and DP.
pip install dpamber
corr
tool generates DeePMD-kit training data for DPRc from AMBER sander low-level QM/MM data and high-level data. For details of DPRc, read the DPRc paper.
Before using this tool, one need to prepare low-level and high-level QM/MM data:
Low-level and high-level data should use the same coordinate and the same MM method, but different QM methods. So, the correction energy for training will be
An example of the command is
dpamber corr --cutoff 6. --qm_region ":1" --parm7_file some_param.param7 --nc some_coord.nc --hl high_level --ll low_level --out dataset
where --cutoff
takes cutoff radius of the QM/MM interaction for training. --qm_region
takes AMBER mask format for the QM region. --parm7_file
and --nc
take the PARM7 file and the trajectory (NetCDF) file, respectively. --ll
and --hl
are the prefixes of low-level and high-level files, including the mdout file (.mdout
), the mden file (.mden
) and the mdfrc file (.mdfrc
). The output dataset directory should be put in --out
.
See details from dpamber corr -h
.
devi
can be used to calculate the model deviation of a given trajectory.
You need to install DeePMD-kit using
pip install dpamber[dpgpu]
See dpamber devi -h
for details.