deepmodeling/deepmd-kit

View on GitHub

A deep learning package for many-body potential energy representation and molecular dynamics

Stars: 2k
Forks: 649
Open beginner issues: 0
Indexed issues: 162
Avg merge: 6d 18h
Merged PRs (30d): 15

Dominant language: Python
License: LGPL-3.0
Last GitHub push: Sep 14, 2026
Latest indexed: Sep 15, 2026
Contributing guide: Contributing guide
Code of conduct: No code of conduct
Beginner labels: good first issue

Open All

162 open issues indexed

feat(pt_expt): honor training.save_dir through a checkpoint manager Open

enhancement

deepmodeling/deepmd-kit#5817 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(pt_expt): add compile-aware EMA lifecycle and checkpoint retention Open

enhancement

deepmodeling/deepmd-kit#5818 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(train): support interval metric aggregation for disp_avg Open

enhancement

deepmodeling/deepmd-kit#5819 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(pt_expt): add torch profiler lifecycle and trace export Open

enhancement

deepmodeling/deepmd-kit#5820 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(pt_expt): add rank-aware TensorBoard observation Open

enhancement

deepmodeling/deepmd-kit#5821 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(pt_expt): decouple DPA1 graph normalization and statistics from dense sel Open

enhancement

deepmodeling/deepmd-kit#5824 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
refactor(pt_expt): make graph-native preprocessing and export independent of get_sel Open

enhancement

deepmodeling/deepmd-kit#5825 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
perf: NeighborGraph attention is 4-9x slower than dense and OOM-prone -- move to block-dense attention via the destination CSR Open

enhancement

deepmodeling/deepmd-kit#5861 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
DPA1-L0-M PT-EXPT models fail in LAMMPS/Kokkos water MD Open

bug

deepmodeling/deepmd-kit#5864 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(jax): train non-energy task models with matching losses Open

enhancement

deepmodeling/deepmd-kit#5865 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(jax): support spin models or add backend-aware validation Open

enhancement

deepmodeling/deepmd-kit#5866 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(jax): consume the common optimizer configuration Open

enhancement

deepmodeling/deepmd-kit#5867 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(jax): support or reject common composite model variants Open

enhancement

deepmodeling/deepmd-kit#5868 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(jax): support epoch-based and multi-task training horizons Open

enhancement

deepmodeling/deepmd-kit#5869 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
fix(jax): honor training.stat_file for model statistics Open

deepmodeling/deepmd-kit#5870 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(jax): implement profiler lifecycle and trace output Open

enhancement

deepmodeling/deepmd-kit#5871 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(jax): implement TensorBoard training output Open

enhancement

deepmodeling/deepmd-kit#5872 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(jax): apply training.mixed_precision to the JAX train step Open

enhancement

deepmodeling/deepmd-kit#5873 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(tf2): support or reject legacy TensorFlow-only descriptors Open

enhancement

deepmodeling/deepmd-kit#5874 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(tf2): support or reject TensorFlow-only model features and variants Open

enhancement

deepmodeling/deepmd-kit#5875 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(tf2): support spin models or add backend-aware validation Open

enhancement

deepmodeling/deepmd-kit#5876 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(tf2): support or reject dens, ener_spin, and population losses Open

enhancement

deepmodeling/deepmd-kit#5877 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(tf2): align optimizer dispatch with the common schema Open

enhancement

deepmodeling/deepmd-kit#5878 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(tf2): support epoch-based and multi-task training horizons Open

enhancement

deepmodeling/deepmd-kit#5879 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(tf2): implement training profiler support Open

enhancement

deepmodeling/deepmd-kit#5880 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
feat(tf2): implement training.mixed_precision Open

enhancement

deepmodeling/deepmd-kit#5881 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
[Code scan] SE-Attention ops can read and write past mismatched tensor shapes Open

bug

deepmodeling/deepmd-kit#5895 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Generic LinearAtomicModel: combine all output-def-matched variables of submodels, not just energy Open

enhancement

deepmodeling/deepmd-kit#5901 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Auto-select the optimal NeighborGraph builder (nv/vesin/dense) by device and availability Open

enhancement

deepmodeling/deepmd-kit#5902 · 2 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
refactor(dpmodel): use tracing-safe device helper in change_type_map Open

enhancement

deepmodeling/deepmd-kit#5907 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
docs: document features added since v3.1.3 Open

Docs

deepmodeling/deepmd-kit#5917 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
docs(cli): document dp test stress output and dp show serialization-tree Open

Docs

deepmodeling/deepmd-kit#5921 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
docs(inference): document vesin neighbor list and nlist_backend for Python/ASE inference Open

Docs

deepmodeling/deepmd-kit#5922 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
docs(model): document charge-spin config, so3_readout, distinguish_types; fix population toctree Open

Docs

deepmodeling/deepmd-kit#5924 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
docs(train): document stat_file, EMA shadow model, and LMDB datasets Open

Docs

deepmodeling/deepmd-kit#5925 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
docs(loss): document loss_func/f_use_norm and three-value huber_delta Open

Docs

deepmodeling/deepmd-kit#5926 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
ci(pt_expt): cover torch.compile on a supported GPU release Open

enhancement

deepmodeling/deepmd-kit#5935 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
[Feature Request] Neighbor-graph builders hardcode n_node = nloc, so the ragged node axis is never exercised Open

enhancement

deepmodeling/deepmd-kit#5938 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
[BUG] DPA4 native spin model raises IndexError on mixed-type frames with -1 padding Open

bug

deepmodeling/deepmd-kit#5940 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
[BUG] neighbor_list behaves three different ways across backends (implemented / silently ignored / raising) Open

bug

deepmodeling/deepmd-kit#5941 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
[Feature Request] Add cutoff-based compact evaluation to LAMMPS pair_style deepmd for DPRc Open

enhancement

deepmodeling/deepmd-kit#5942 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Deprecate descriptor-scoped exclude_types in favour of the model-level build-seam owner Open

enhancement

deepmodeling/deepmd-kit#5947 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
fix(pt-expt): preserve DOS width in exported models Open

bug

deepmodeling/deepmd-kit#5949 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
DeepMD training problem Open

wontfix

deepmodeling/deepmd-kit#5950 · 11 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
ci: build and test Paddle custom operator with CUDA Open

enhancement

deepmodeling/deepmd-kit#5952 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
bug(pt-expt): compiled DPA2 lower miscompiles descriptors that depend on atype_ext Open

bug

deepmodeling/deepmd-kit#5953 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Record exact label availability when building LMDB datasets Open

enhancement

deepmodeling/deepmd-kit#5954 · 1 comment · 0 reactions · 0 assignees · Indexed Sep 15, 2026
pt_expt change-bias does not preserve the lower-input ABI of .pt2 models Open

bug

deepmodeling/deepmd-kit#5959 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
[BUG] Multi-frame DeepPot.eval produces NaNs and corrupted outputs for frozen native-spin DPA4 graph models Open

bug

deepmodeling/deepmd-kit#5969 · 3 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026
Make BackendModelFactory class injection complete and required (no silent dpmodel fallback) Open

enhancement

deepmodeling/deepmd-kit#5977 · 0 comments · 0 reactions · 0 assignees · Indexed Sep 15, 2026

deepmodeling/deepmd-kit

Get new issues in your inbox