nnpd.core.runner¶

Plan first; train and cache models; analyze saved models with separate analysis settings.

Attributes¶

`TRAINING_FILE`
`ANALYSIS_FILE`
`KEEP_SETTINGS`
`STAGES`

Classes¶

Job

Functions¶

`expand_jobs`(→ list[Job])	Validate every training alternative, and the analysis of each, before anything runs.
`infeasible`(→ dict[str, list[str]])	Configurations whose analysis would exceed the configured limits, with the reasons.
`check_feasible`(→ None)	Refuse the whole launch before anything is computed if any configuration cannot be analyzed.
`plan`(→ list[Job])	What run would train, with each model's analysis workload; computes nothing.
`analyze_run`(→ dict)	Compute the selected metrics and plots of one run, replacing earlier ones. Never trains.
`save_references`(→ None)
`execute`(→ list[pathlib.Path])	Train, analyze, or both, from the settings files in settings_dir; the CLI does the same.
`restore`(→ nnpd.core.api.Context)	Open a saved run, with the analysis settings it was analyzed with, on any supported device.
`verify_store`(→ int)	Recompute all saved artifact checksums; raises on the first damaged file.

Module Contents¶

nnpd.core.runner.TRAINING_FILE = 'settings_training.py'¶

nnpd.core.runner.ANALYSIS_FILE = 'settings_analysis.py'¶

nnpd.core.runner.KEEP_SETTINGS = 3¶

nnpd.core.runner.STAGES = ('train', 'run', 'analyze')¶

class nnpd.core.runner.Job¶

settings: dict¶

member: dict¶

changed: str | None¶

workload: dict¶

property id: str¶

as_dict() → dict¶

nnpd.core.runner.expand_jobs(training_settings: dict, experiment: nnpd.core.api.Experiment, analysis_settings: dict | None = None, analysis: nnpd.core.api.Analysis | None = None) → list[Job]¶: Validate every training alternative, and the analysis of each, before anything runs.

nnpd.core.runner.infeasible(jobs: list[Job]) → dict[str, list[str]]¶: Configurations whose analysis would exceed the configured limits, with the reasons.

nnpd.core.runner.check_feasible(jobs: list[Job]) → None¶: Refuse the whole launch before anything is computed if any configuration cannot be analyzed.

nnpd.core.runner.plan(profile: str | None = None, settings_dir: str | pathlib.Path = '.') → list[Job]¶: What run would train, with each model’s analysis workload; computes nothing.

nnpd.core.runner.analyze_run(context: nnpd.core.api.Context) → dict¶: Compute the selected metrics and plots of one run, replacing earlier ones. Never trains.

nnpd.core.runner.save_references(context: nnpd.core.api.Context) → None¶

nnpd.core.runner.execute(stage: str = 'run', *, profile: str | None = None, settings_dir: str | pathlib.Path = '.') → list[pathlib.Path]¶

Train, analyze, or both, from the settings files in settings_dir; the CLI does the same.

‘train’ trains the models of settings_training.py and resets their analysis. ‘analyze’ analyzes the latest trained models with settings_analysis.py; it never trains. ‘run’ does both. ‘jobs’: torchrun workers process disjoint runs; ‘ddp’: workers train each model together and only rank zero analyzes. After success, runs of earlier settings are removed, the cache keeps what the last KEEP_SETTINGS trainings used, and the settings files are copied into the output folder.

nnpd.core.runner.restore(run_dir: str | pathlib.Path, *, device: str = 'cpu') → nnpd.core.api.Context¶: Open a saved run, with the analysis settings it was analyzed with, on any supported device.

nnpd.core.runner.verify_store(root: str | pathlib.Path) → int¶: Recompute all saved artifact checksums; raises on the first damaged file.