deepmodeling / deepmodeling/deepmd-kit

[mixed_batch=True]: Replace SameNlocBatchSampler with

Open
#5,386 0 comments 0 reactions 0 assignees View on GitHub
branch: master lang: py todo comment type: todo
Dominant language
Python
Stars
2k
Forks
649
Avg merge
6d 18h
Merged PRs (30d)
15

Description

RandomSampler(replacement=False) + padding collate_fn.
Changes needed:
1. _collate_lmdb_batch: pad coord/force/atype to max_nloc,
add "atom_mask" bool tensor (nframes, max_nloc)
2. Use RandomSampler(_data, replacement=False) as sampler
3. Use fixed batch_size in DataLoader (not batch_sampler)
4. Model forward: apply atom_mask to descriptor/fitting
5. Loss: mask out padded atoms in force loss

Line: 240
https://github.com/deepmodeling/deepmd-kit/blob/565f4beec534576ec27ea32446b0a8e000d989e7/deepmd/pt/train/training.py#L237-L247

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.