v2.1.0

[wanghan-iapcm]

New feature:

• Model compression for se_3, se_r descriptors. Energy and tensor models (Model compression for se-t descriptor #1225 enable model compression for dipole and polar type fitting net #1228 Finish model compression for se_r descriptor! #1361 )
• Add init-frz-model support for se-t type descriptor (add init-frz-model support for se-t type descriptor #1245)
• Added all activation functions for model compression. (Added all activation functions for model compression. #1283)
• Update guidelines for the number of threads (update guidelines for the number of threads #1291)
• Enable mixed precision support for deepmd-kit (Enable mixed precision support for deepmd-kit #1285 fix bug of mixed precision training #1471 )
• Unify C++ errors and pass message to LAMMPS (unify C++ errors and pass message to LAMMPS #1326)
• Optimize DPTabulate._build_lower method (optimize DPTabulate._build_lower method #1323)
• Calculate neighbor statistics from CLI (calculate neighbor statistics from CLI #1476)
• Add an interface to eval descriptors (add an interface to eval descriptors #1483)

Enhancement:

• deprecate numb_test in the training script (deprecate numb_test #1249)
• Accelerate model compression (Accelerate model compression #1274)
• Use c++14 for TF 2.7 (Use C++14 for TF 2.7 #1275)
• Add a citation badge (add a citation badge to README #1280)
• Add embedding network dimension check of model compression (Add embedding network dimension check of model compression #1303)
• Provide an option to skip neighbor stat (provide an option to skip neighbor stat #1313)
• Add an error message to compress/freeze (add an error message for copress/freeze #1319)
• Redirect print_summary to LAMMPS log (redirect print_summary to LAMMPS log #1324)
• Enable OpenMP for prod_force and prod_virial (enable OpenMP for prod_force and prod_virial #1360)
• Update issue templates (Update issue templates #1368)
• Bump LAMMPS version to stable_29Sep2021_update2 (bump LAMMPS version to stable_29Sep2021_update2 #1279)
• Remove api_cc/include/custom_op.h (remove api_cc/include/custom_op.h #1405)
• Introduce TensorFlow Profiler (introduce TensorFlow Profiler #1414)
• Only test/eval fitting properties during training (only test/eval fitting properties #1416 replace tf.no_op #1419 )
• Remove the dependency on inputs from inputs_zero (remove the dependency on inputs from inputs_zero #1417)
• Support recursive detection for the systems of model_devi (support recursive detection for the system of model_devi #1424)
• Enable TF remapper optimizer (enable TF remapper optimizer #1418)
• Dynamically load op library in C++ interface (dynamically load op library in C++ interface #1384)
• Dplr doc and examples (Dplr doc and examples #1458)
• Bump the Python version to 3.10 (bump the Python version to 3.10 #1465)
• Do some small optimization to ops (avoid concat or add in loops. Instead, append tensors to a list, and concat or accumulate_n after loops) (do some small optimization to ops #943)
• Optimizations related to data statistics
  • Skip data_stat in init_from_model and restart mode (skip data_stat in init_from_model and restart mode #1463)
  • Assign energy shift stats if atomic energies are assigned (assign energy shift stats if atomic energies are assigned #1477)
  • Recover input stats from frozen models (recover input stats from frozen models #1482)
• Test: move loading graphs to setUpClass to accelerate tests (test: move loading graphs to setUpClass #1484)
• Run test_python in the pre-built container (run test_python in the pre-built container #1487)

Bug fixings:

• Update and fix typos in doc (Fix typos in training.md #1238 #1239 fix a typo in install_lammps.md #1328 fix typo in doc/troubleshooting/howtoset_netsize.md #1300 fix a typo in document #1445 typo: 'should be in consistent' -> 'should be consistent' #1490 Fix typos in docs #1497 fix the example of inference with C++ interface #1504 typos: fix more small grammar issues on install pages #1503 y-axis should use log scale instead of symlog #1514 )
• Fix compress training bug within the dp train --init-frz-model interface (Fix compress training bug within the dp train --init-frz-model interface #1233)
• Fix Python bugs of loc_frame descriptor (fix Python bugs of loc_frame descriptor #1253)
• Fix bug of loc_frame descriptor when using lammps (fix bug of loc_frame descriptor when using lammps #1255)
• Fix single precision error (fix single precision model compression error #1212)
• Fix the np.frombuffer in dp transfer (Fix bug in transfer.py #1246)
• Fix SyntaxWarning in graph.py (fix SyntaxWarning in graph.py #1278)
• Change googletest from master to main (change googletest from master to main #1292)
• update_deepmd_input when compress (updata_deepmd_input when compress #1297 fix typo updata->update #1301 )
• Add importlib_metadata as dependency (add importlib_metadata as dependency #1308)
• Fix bugs about parameters of memset (Fix bug with cudamemset's parameter #1302)
• Fix model compression bug when fparam or aparam is not zero (Fix model compression bug when fparam or aparam is not zero #1306)
• Add space between words in messages (add space between words in messages #1312)
• Do not print virial error with nopbc data (do not print virial error with nopbc data #1314)
• Fix test errors with TensorFlow 2.7 (Fix test errors with TensorFlow 2.7 #1315)
• Fix bug of hip model compression (Fix rocm model compression bug #1325)
• Prevents rcut_smth larger than rcut (prevents rcut_smth larger than rcut #1354)
• Fix cell and virial transpose bug in dp_ipi (Fix cell and virial transpose bug in dp_ipi #1353)
• Fix bug in DipoleFittingSeA: (fix bug in DipoleFittingSeA: #1363)
• Fix cxx standard for LAMMPS (fix cxx standard for LAMMPS #1379)
• Explicitly set neighbor request to full in compute deeptensor/atom to fix bug [BUG] Incorrect dipole prediction with LAMMPS of new version #1381 (explicitly set neighbor request to full in compute deeptensor/atom to fix bug #1381 #1382)
• Fix NameError (fix NameError #1385)
• Fix network precision under specific situation (fix network precision under specific situation #1391 add unittest for deepmd.common.cast_precision #1394 )
• Initialize input virial vector to zero (initialize input virial vector to zero #1397)
• Make OpenMP an optional dependency (make OpenMP an optional dependency #1498)
• Fix nvcc warning when using cuda-11.x toolkit (Fix nvcc warning when using cuda-11.x toolkit #1401)
• Add UT for se_3 type descriptor (add UT for se_3 type descriptor #1404)
• Fix github git url (fix github git url #1409)
• Fix gelu grad multi definitions error (fix gelu grad multi definitions error #1406)
• Fix cast_precision breaking docstring (fix cast_precision breaking docstring #1437)
• Add image link of ROCm version. (Add image link of ROCm version. #1432)
• Pass integer zero to memset (pass integer zero to memset #1499)

---

This discussion was created from the release v2.1.0.

[likefallwind]

Cool！