Input Data Extraction from different datasets

[AnuragKr]

Hello @njzjz,

Q1. I am trying to reproduce MD17 and QM9 result. I want to know how you have extracted data to provide input to DeePMD-kit for producing results for paper.
If you have used dpdata which format you have used?

Q2. Can DeePMD accept equilibrium molecules as input rather than Molecular dynamics trajectory data ?

[njzjz]

I am not the author. Ping @wanghan-iapcm @jameswind for comments.

[AnuragKr]

@wanghan-iapcm @jameswind Can you please help me with a question mentioned above?

[wanghan-iapcm]

1. the deepmd-kit data format is explained in https://docs.deepmodeling.com/projects/deepmd/en/master/data/index.html .
2. yes.

[O-Nay]

Hello, dear @AnuragKr. I'm quite new to molecular dynamics, can you please show detailed instructions on how to train QM9 on DeepMD and output the results?

[AnuragKr]

@O-Nay Thanks for your question, I have done training with MD-17, not QM-9.
Steps --

1. Download the Numpy format for every molecule from this link
2. Create a Data folder that will contain type. raw and type_map.raw , nopbc(for organic molecule) file and set.000 folder.
3. Manually filled in the details of the type.raw and type_map.raw as per molecules.
4. Wrote a script to read data from numpy file and create a coord.npy, force.npy and energy.npy file inside data/set.000 folder as per DeePMD-data-format Read the descriptor of the data format before writing a script -- link
   For QM-9 you can refer to this script and then modify it to create data in DeePMD format -- https://github.com/atomistic-machine-learning/SchNet/blob/master/scripts/load_qm9.py

[O-Nay]

@AnuragKr Please tell me, could you provide data for ethanol? Preferably a folder with set.000 and type.raw. Look at the file structure and the script itself. Forgive me for my impudence, it's hard to figure it out yet)

[AnuragKr]

@O-Nay I am sharing a link to a few molecules for which I have created data. Please let me know if you find any difficulty in downloading.