QoI Configuration

Source code:

• bff/workflows/configs.py
• bff/workflows/qoi.py
• bff/qoi/routines.py

Purpose

bff qoi computes quantities of interest for:

• the sampled trainset produced by bff trainset
• the reference trajectories provided by the user

Minimal Example

trainset:
  dir: ../02-training-data/trainset

refset:
  systems:
    - coordinates: ../01-prepare/colvars/reference/system-000/system.gro
      topology: ../01-prepare/colvars/reference/system-000/system.top
      trajectory: ../02-reference-data/trajectories/pos-000.xtc
      routines:
        - routine: rdf
          mol_resname: ACE
        - routine: ./restraint.py:distance_distribution
          atom_pair: [C2, CAL]

run:
  in_memory: true
  gc_collect: false
  maxtasksperchild: 100

output:
  path: ./qoi
  log: ./out.log
  write_raw: true

Top-Level Keys

• trainset Trainset slicing and worker settings.
• refset Reference systems, slicing, and routines.
• run Runtime settings for multiprocessing and memory handling.
• output Output prefix and logging settings.

trainset Keys

• dir Trainset directory produced by bff trainset.
• start First frame index to analyze.
• stop Final frame index to analyze.
• step Frame stride.
• workers Worker count for trainset analysis. -1 means the implementation default.
• progress_stride Progress update interval for trainset processing.

refset Keys

• start First frame index for reference trajectories.
• stop Final frame index for reference trajectories.
• step Frame stride for reference trajectories.
• systems Non-empty list of reference systems.

refset.systems[] Keys

• coordinates Coordinate file for the reference system.
• topology Topology file for the reference system.
• trajectory Reference trajectory file.
• routines Non-empty list of QoI routines to evaluate for that system.

Routine Keys

• routine Either a builtin routine name such as rdf or hb, or a custom routine string in the form path.py:function.
• any additional keys Passed directly to the routine as keyword arguments.

Builtin routines currently live in:

• bff/qoi/rdf.py
• bff/qoi/hbonds.py

run Keys

• in_memory Load trainset QoI results into memory before reduction.
• gc_collect Run explicit garbage collection between tasks.
• maxtasksperchild Worker recycling interval for multiprocessing.

output Keys

• path Output prefix for the written QoI dataset files.
• log Log file path.
• write_raw If true, also write the raw per-block QoI data as JSON.