[BUG] Atom vs Atomic problem marked as solved in #4536 still a problem

[MajGroup]

Bug summary

Training atomic dipoles returns nan values unless the data is labelled as atom_.npy instead of atomic_.npy.

DeePMD-kit Version

DeePMD-kit v3.0.1

Backend and its version

TF

How did you download the software?

conda

Input Files, Running Commands, Error Log, etc.

{
"_comment": "that's all",
"model"
: {
"type_map"
: [
"C",
"N",
"O",
"H",
"X",
"Z"
],
"descriptor": {
"type" : "se_e2_a",
"sel": "auto",
"rcut_smth": 0.5,
"rcut" : 6.0,
"neuron": [
16,
32,
48],
"resnet_dt": false,
"seed": 1,
"_comment": " that's all" },
"fitting_net": {
"type": "dipole",
"sel_type": [2],
"neuron"
: [
150,
150,
150
],
"resnet_dt": true,
"seed": 1222,
"activation_function": "tanh",
"_comment": " that's all"
},
"_comment": " that's all"
},
"learning_rate"
: {
"type": "exp",
"decay_steps": 1000,
"start_lr": 5e-4,
"stop_lr": 1e-8,
"_comment": "that's all"
},
"loss"
: {
"type": "tensor",
"pref": 0.0,
"pref_atomic": 1.0,
"_comment": " that's all"
},
"training"
: {
"training_data"
: {
"systems"
: [
"./training_data"
],
"batch_size": "auto",
"_comment": "that's all"
},
"validation_data":{
"systems": ["./validation_data"],
"batch_size": "auto",
"_comment": "that's all"
},

"numb_steps": 325000,
"seed":7515,
"disp_file": "lcurve.out",
"disp_freq": 100,
"save_freq": 100,
"_comment": "that's all"

},

"_comment": "that's all"
}

Steps to Reproduce

Training atomic dipole model from wannier data

Further Information, Files, and Links

No response

[anyangml]

The fix provided by #4538 is after the release of v3.0.1. Can you try the devel branch?

[MajGroup]

Indeed, I missed that it is a part of the 3.0.2 milestone. Thanks for clarifying. My apologies for missing this. Closing the thread