mmml active-learning¶
Sample structures for re-labeling.
Usage¶
mmml active-learning --help
Options¶
usage: mmml active-learning [-h] -i TRAJ [TRAJ ...] [-o OUTPUT] [--max-temp K]
[--stride STRIDE] [--max-frames N]
[--no-temp-filter]
Extract frames from MD trajectories for active learning (pyscf-evaluate input).
Input & configuration:
-i, --input TRAJ [TRAJ ...]
Trajectory file(s) (.traj, .xyz). Globs supported, e.g.
'out/*.traj'
Output & artifacts:
-o, --output OUTPUT Output NPZ path (default: md_sampled.npz)
Diagnostics & safety:
-h, --help show this help message and exit
Other options:
--max-temp K Keep only frames with T < max-temp K (default: 300).
Ignored if trajectories have no velocities.
--stride STRIDE Use every Nth frame (default: 1)
--max-frames N Maximum frames to extract (default: no limit)
--no-temp-filter Do not filter by temperature (keep all frames)
CLI to extract frames from MD trajectories for active learning. Filters frames
by temperature (e.g. T < 300 K) and saves to NPZ format compatible with mmml
pyscf-evaluate for extending the training set. Usage: mmml active-learning -i
out/physnet_md/physnet_ase.traj -o md_sampled.npz mmml active-learning -i
traj1.traj traj2.traj -o md_sampled.npz --max-temp 300 mmml active-learning -i
"out/*.traj" -o md_sampled.npz --stride 5