Encountered difficulties while converting traj. files

[Jx1-true]

I'm trying to convert a .traj file into .npy format for DeepMD training using dpdata, but I’m facing some difficulties. The file contains many frames, and it seems that some frames have different numbers of atoms, which might be causing issues during the conversion. I’m not sure if dpdata can handle this directly or if I need to use ase in combination with it. Does dpdata automatically filter out inconsistent frames, or do I need to preprocess them manually? I'd appreciate any advice on the best way to proceed.

I've attached the .traj file for reference. Thanks in advance for your help!

train.traj.zip