How autoatml/autoplex works
Autoplex automates construction and benchmarking of machine-learning interatomic potentials. Users configure and launch either random-structure-search (RSS) or phonon workflows; these orchestrate structure generation or phonon data production, reference calculations, data preparation, and potential fitting. The fitting layer supports multiple MLIP families, with shared configuration and regularization. The README documents these workflow capabilities; some subsystem links below rely on the SOURCE INDEX rather than inspected implementations, so those links are marked unverified.
Main entry points: rss/jobs.py RSS iteration jobs · common/flows.py Fitting workflows · phonons/flows.py Phonon workflows · common/jobs.py Data collection
Components 20
Workflow orchestration
- RSS workflows
src/autoplex/auto/rss/flows.py - RSS iteration jobs
src/autoplex/auto/rss/jobs.py - Phonon workflows
src/autoplex/auto/phonons/flows.py - Phonon workflow jobs
src/autoplex/auto/phonons/jobs.py
Training data
- RSS structure generation
src/autoplex/data/rss/jobs.py - RSS data jobs
src/autoplex/data/rss/flows.py - RSS data processing
src/autoplex/data/rss/jobs.py - Phonon data flows
src/autoplex/data/phonons/flows.py - Phonon data jobs
src/autoplex/data/phonons/jobs.py - Shared data flows
src/autoplex/data/common/flows.py - Data collection
src/autoplex/data/common/jobs.py - NMR data jobs
src/autoplex/data/nmr/jobs.py
Potential fitting
- Fitting workflows
src/autoplex/fitting/common/flows.py - MLIP fitting
src/autoplex/fitting/common/jobs.py - Data regularization
src/autoplex/fitting/common/regularization.py - MLIP hyperparameters
src/autoplex/fitting/mlip_hypers/_gap_hypers.py
Compute integration
- Workflow settings
src/autoplex/settings.py - CASTEP jobs
src/autoplex/misc/castep/jobs.py - CASTEP flows
src/autoplex/misc/castep/flows.py
Other components
- Researcher
Connections 23
- Researcher → RSS workflows: configures and launches
- Researcher → Phonon workflows: configures and launches
- RSS workflows → RSS iteration jobs: dispatches
src/autoplex/auto/rss/flows.py - RSS iteration jobs → RSS structure generation: generates structures
src/autoplex/auto/rss/jobs.py - RSS iteration jobs → RSS data jobs: collects data
src/autoplex/auto/rss/jobs.py - RSS data jobs → RSS data processing: uses data jobs
src/autoplex/data/rss/flows.py - RSS iteration jobs → Fitting workflows: requests fitting
src/autoplex/auto/rss/jobs.py - Phonon workflows → Phonon workflow jobs: dispatches
src/autoplex/auto/phonons/flows.py - Phonon workflows → Phonon data flows: requests data generation
src/autoplex/auto/phonons/flows.py - Phonon data flows → Phonon data jobs: uses phonon jobs
src/autoplex/data/phonons/flows.py - Phonon data flows → Data collection: uses shared jobs
src/autoplex/data/phonons/flows.py - Phonon workflows → Fitting workflows: requests fitting
src/autoplex/auto/phonons/flows.py - Shared data flows → Data collection: dispatches
src/autoplex/data/common/flows.py - Data collection → Data regularization: uses regularization
src/autoplex/data/common/jobs.py - NMR data jobs → Data collection: uses shared jobs
src/autoplex/data/nmr/jobs.py - Fitting workflows → MLIP fitting: dispatches fitting
src/autoplex/fitting/common/flows.py - Fitting workflows → Data regularization: applies regularization
src/autoplex/fitting/common/flows.py - Fitting workflows → MLIP hyperparameters: selects hyperparameters
- MLIP fitting → MLIP hyperparameters: loads model settings
src/autoplex/fitting/common/jobs.py - RSS workflows → Workflow settings: reads settings
src/autoplex/auto/rss/flows.py - CASTEP flows → CASTEP jobs: dispatches jobs
src/autoplex/misc/castep/flows.py - RSS iteration jobs → CASTEP jobs: uses CASTEP jobs
src/autoplex/auto/rss/jobs.py - CASTEP jobs → Workflow settings: reads configuration
src/autoplex/misc/castep/jobs.py