🚧 Early release · Under active development 🚧

`model.runtime.workflows.base_workflow`¶

Base workflow abstractions for ADELM calibration workflows.

Module Contents¶

Classes¶

Shared calibration-workflow scaffold.

API¶

class model.runtime.workflows.base_workflow.BaseWorkflow(config, config_path=None, config_overrides=None)¶

Shared calibration-workflow scaffold.

Initialization

workflow_name¶: ‘base’

classmethod from_config(config_path, config_overrides=None)¶

build_runner(show_logs=True, show_driver_diagnostics=True, runtime_output_dir=None, disable_runtime_output=False, disable_nn_init=False, workflow_name=None, fold_id=None, seed=None)¶

static log_stage(title, detail=None)¶

create_optimizer(runner, lr)¶

create_scheduler(optimizer)¶

resolve_output_dir(output_dir=None)¶

build_cross_validation(runner)¶

static set_random_seed(seed)¶

static select_folds(cv, fold_id=None)¶

initialize_results(cv, runner)¶

initialize_single_run_results(runner)¶

static summarize_history(history)¶

static save_nn_weights(runner, output_dir)¶

save_config_snapshot(output_dir, runtime_overrides=None)¶

save_train_log_csv(history, output_dir)¶

save_target_csvs(runner, predictions, targets, test_targets, output_dir)¶

save_metrics_csv(metrics, output_dir)¶

save_full_inference_netcdf(runner, output_dir)¶

save_final_inference(runner, evaluation, output_dir)¶

fold_output_dir(root_dir, fold, seed)¶

no_cv_output_dir(root_dir, seed)¶

summarize_evaluation(evaluation)¶

target_metrics(predictions, targets)¶

print_dataset_summary(runner, train_targets=None, val_targets=None, label=None, train_site_count=None, val_site_count=None)¶

optimize_runner(runner, train_targets, valid_targets=None, epochs=None, lr=None, train_chunk_size=None, max_grad_norm=None, show_progress=True, debug=False)¶

abstract run(epochs=10, lr=0.001, train_chunk_size=None, max_grad_norm=1.0, skip_nan_grads=True, show_progress=True, debug=False, output_dir=None, fold_id=0)¶