Weekly Progress of DeePMD-kit 20220919

[HuangJiameng]

What's new

• highlight LAMMPS codes in doc #1921
• handle float point error in sys_probs #1919
• Use GeluCustom as operator name #1918
• LAMMPS: Use of “override” instead of “virtual”  #1915
• find protobuf headers from extra paths #1910

Discussion

• Include more frames when training dp potential model? #1912
• DeePMD potential for thermal conductivity calculation #1914
• How to use the wildcard in input.json #1911
• disabling DeePMD TF32 calculation in LAMMPS #1920
• error on LAMMPS with DPLR #1927
• Does mixed precision only benefit double-precision models? #1925
• Inference with C++ interface under OpenMP #1924
• How to reduce the RMSE in energy without losing accuracy in force #1929