How to train the special element with different atoms and structrure with deepmd-kit?

[zcb-code]

I am training an element PD, I have many structures with different atom numbers. I have many OUTCARs. Is it possible to use these OUTCARS(only with one step self-consistent computing), to train a initial *pd.file ? After I run the script:

from dpdata import LabeledSystem,MultiSystems
from glob import glob
"""
process multi systems
"""
fs=glob('./*/OUTCAR') # remeber to change here !!!
ms=MultiSystems()
for f in fs:
try:
ls=LabeledSystem(f)
except:
print(f)
if len(ls)>0:
ms.append(ls)

ms.to_deepmd_raw('deepmd')
ms.to_deepmd_npy('deepmd')

---

$ cd deepmd;tree, I got:
.
|-- Li10
| |-- box.raw
| |-- coord.raw
| |-- energy.raw
| |-- force.raw
| |-- set.000
| | |-- box.npy
| | |-- coord.npy
| | |-- energy.npy
| | |-- force.npy
| | -- virial.npy | |-- type.raw | |-- type_map.raw | -- virial.raw
|-- Li4
| |-- box.raw
| |-- coord.raw
| |-- energy.raw
| |-- force.raw
| |-- set.000
| | |-- box.npy
| | |-- coord.npy
| | |-- energy.npy
| | |-- force.npy
| | -- virial.npy | |-- type.raw | |-- type_map.raw | -- virial.raw
|-- Li5
| |-- box.raw
| |-- coord.raw
| |-- energy.raw
| |-- force.raw
| |-- set.000
| | |-- box.npy
| | |-- coord.npy
| | |-- energy.npy
| | |-- force.npy
| | -- virial.npy | |-- type.raw | |-- type_map.raw | -- virial.raw
|-- Li6
| |-- box.raw
| |-- coord.raw
| |-- energy.raw
| |-- force.raw
| |-- set.000
| | |-- box.npy
| | |-- coord.npy
| | |-- energy.npy
| | |-- force.npy
| | -- virial.npy | |-- type.raw | |-- type_map.raw | -- virial.raw
|-- Li7
| |-- box.raw
| |-- coord.raw
| |-- energy.raw
| |-- force.raw
| |-- set.000
| | |-- box.npy
| | |-- coord.npy
| | |-- energy.npy
| | |-- force.npy
| | -- virial.npy | |-- type.raw | |-- type_map.raw | -- virial.raw
|-- Li8
| |-- box.raw
| |-- coord.raw
| |-- energy.raw
| |-- force.raw
| |-- set.000
| | |-- box.npy
| | |-- coord.npy
| | |-- energy.npy
| | |-- force.npy
| | -- virial.npy | |-- type.raw | |-- type_map.raw | -- virial.raw
|-- Li9
| |-- box.raw
| |-- coord.raw
| |-- energy.raw
| |-- force.raw
| |-- set.000
| | |-- box.npy
| | |-- coord.npy
| | |-- energy.npy
| | |-- force.npy
| | -- virial.npy | |-- type.raw | |-- type_map.raw | -- virial.raw
`-- out

---

My second question is that, if possible, how do I modify my Li.json file before train?
my Li.json :
{
"_comment": " model parameters",
"model": {
"type_map": ["Li"],
"descriptor" :{
"type": "se_a",
"sel": [],
"rcut_smth": 0.50,
"rcut": 6.00,
"neuron": [25, 50, 100],
"resnet_dt": false,
"axis_neuron": 16,
"seed": 1,
"_comment": " that's all"
},
"fitting_net" : {
"n_neuron": [240, 240, 240],
"resnet_dt": true,
"seed": 1,
"_comment": " that's all"
},
"_comment": " that's all"
},

"learning_rate" :{
    "type":         "exp",
    "start_lr":     0.001,
    "decay_steps":  500,
    "decay_rate":   0.95,
    "_comment":     "that's all"
},

"loss" :{
    "start_pref_e": 0.02,
    "limit_pref_e": 1,
    "start_pref_f": 1000,
    "limit_pref_f": 1,
    "start_pref_v": 0,
    "limit_pref_v": 0,
    "_comment":     " that's all"
},

"_comment": " traing controls",
"training" : {
    "systems":      ["./"],     
    "set_prefix":   "set",
    "stop_batch":   500000,
    "batch_size":   1,

    "seed":         1,

    "_comment": " display and restart",
    "_comment": " frequencies counted in batch",
    "disp_file":    "lcurve.out",
    "disp_freq":    100,
    "numb_test":    10,
    "save_freq":    1000,
    "save_ckpt":    "model.ckpt",
    "load_ckpt":    "model.ckpt",
    "disp_training":true,
    "time_training":true,
    "profiling":    false,
    "profiling_file":"timeline.json",
    "_comment":     "that's all"
},

"_comment":         "that's all"

---

how to modify the "systems": ["./"], part?

Thanks.

[amcadmus]

RTD https://deepmd.readthedocs.io/en/master/train-input.html

[zcb-code]

Thanks for your recommend.

[njzjz]

how to modify the "systems": ["./"], part?

Something like ["data/Li1", "data/Li2", "data/Li3"]. I suggest to write a script to generate it.

[zcb-code]

Thanks for your kind replay. I have another two question, how many sets are needed in files of Li2? And the set number in Li2,Li3 needs to equal or not?

[njzjz]

It depends on you. You may see https://doi.org/10.6084/m9.figshare.12973055 for an example.

[zcb-code]

Thanks for your useful reply.

[njzjz]

@zcb-code Note: if you are using either v1.3.0, v1.3.1, v1.3.2, or v1.3.3, the key model/fitting_net/n_neuron will not work due to a bug. Use model/fitting_net/neuron instead.