feat(pt_expt): implement .pte inference pipeline with dynamic shapes - #5284

Merged

wanghan-iapcm merged 16 commits into

deepmodeling:masterfrom

wanghan-iapcm:feat-pt-expt-infer

Mar 5, 2026

Commits on Mar 3, 2026

feat(pt_expt): implement .pte inference pipeline with dynamic shapes
Han Wang
committed
a31fd18

Commits on Mar 4, 2026

support ase nlist
Han Wang
committed
ebf90e3
update doc
Han Wang
committed
eed7c25
fix device conversion
Han Wang
committed
ae86b11
Resolve Ruff ARG001 in _build_dynamic_shapes
Han Wang
committed
8fd1153
Updated the backend to use the GPU-aware NeighborStat from deepmd.pt_expt.utils.neighbor_stat (ported from PR #5270)
Han Wang
committed
e065c22
use PT's AutoBatchSize to catch OOM
Han Wang
committed
8da4064
build and extend coords on cuda device
Han Wang
committed
291a14b
fix
Han Wang
committed
3304fe8
assign dtype for torch tensors input to neighbor stat
Han Wang
committed
57d2c8a
provide options for make_fx rather than hard coding
Han Wang
committed
1a2624e
align the implementation of env_mat with the training branch
Han Wang
committed
c415ec3
borrow conftest.py
Han Wang
committed
40de8ca

Commits on Mar 5, 2026

fix : test device
Han Wang
committed
497f26a
Merge branch 'master' into feat-pt-expt-infer
Han Wang
committed
f545479
revert change on inputs
Han Wang
committed
143d5c4