Got nan value in output

[jtyang-chem]

I made all data by my scripts and give them into training, then I got a lcurve.out file with nan values.

• To make sure it's not a problem of version or input file, I also tested an example file of water box, everything goes well.
• my deepmd-kit version is v1.3.3 cpu.

I can't see the differences of which I made from the example data, I've already upload them to my github, could someone help me?

[njzjz]

I can reproduce using the latest development version.

[njzjz]

You have atoms overlapping, which is the reason why you got nan values.

[jtyang-chem]

Thanks for the reply. Is it possible to add this situ to some error report?

[njzjz]

#914 will raise an error if some atoms are overlapping.