Weekly Progress of DeePMD-kit 20221024

[HuangJiameng]

What's new

• fix Findtensorflow when LD_LIBRARY_PATH is empty #2010
• migrate from setup.cfg to pyproject.toml #2011
• fix OP_CXX_ABI for TF>=2.9 #2013

Discussion

• Running Lammps parallel in conda environment installation (easy installation method) #1978
• Does deepmd have support ROCM GPU (MI series chip) version? #1999
• The performance gap between Install from source code and conda installation is very large (training speed performance) #2003
• CPU-GPU collaboration problems during dp train #2004
  If the CPU behaved strangely, one can list the CPU usage by ps -L -eo pcpu,psr,pid,lwp,user,args. Also, to enforce the CPU to balance all the threads, a possible method is to manually set an environment variable export KMP_AFFINITY=scatter according to Intel's official manual.
• grasp structure from raw/file #2006
• [LAMMPS] Benchmarking Same As NVIDIA #2007
• Would it be more accurate to use the energy of each atom to calculate the force than to use the average of the total energy to calculate the force? #2016