Workflows API

Workflow public exports.

This module provides package-level re-exports for workflow classes. CLI entry dispatch imports concrete workflow modules directly.

class dpeva.workflows.AnalysisWorkflow(config: AnalysisConfig | Dict[str, Any], config_path: str | None = None)

基类：object

Workflow for analyzing DeepMD test results. Refactored using UnifiedAnalysisManager.

run(*, emit_completion_marker: bool = True)

Run analysis; nested callers leave completion to their parent.

class dpeva.workflows.CollectionWorkflow(config: Dict | CollectionConfig, config_path=None)

基类：object

Orchestrates the Collection pipeline (Refactored). Delegates responsibilities to dedicated Managers: - CollectionIOManager: I/O & Data Loading - UQManager: UQ Calculation & Filtering - SamplingManager: DIRECT Sampling

run()

class dpeva.workflows.DataCleaningWorkflow(config: Dict | DataCleaningConfig)

基类：object

run()

class dpeva.workflows.FeatureWorkflow(config: Dict | FeatureConfig, *, original_config: dict[str, Any] | None = None, config_metadata: dict[str, Any] | None = None, run_options: RunOptions | None = None)

基类：object

Workflow for generating atomic/structural descriptors using DeepPot. Refactored using DDD Managers.

property output_dir_path: Path

run()

Executes the feature generation workflow.

1. Configures workflow logging.

2. Validates data paths.

3. Generates descriptors via CLI (dp eval-desc) or Python API (DeepPot).

4. Handles local or Slurm execution.

class dpeva.workflows.InferenceWorkflow(config: Dict | InferenceConfig, config_path: str | None = None, *, original_config: dict[str, Any] | None = None, config_metadata: dict[str, Any] | None = None, run_options: RunOptions | None = None)

基类：object

Workflow for running inference using an ensemble of DPA models. Supports both Local and Slurm backends via JobManager. Refactored using DDD Managers.

analyze_results()

Analyze results for all models.

run()

Executes the inference workflow.

1. Configures workflow logging.

2. Discovers trained models.

3. Submits parallel inference jobs (dp test) via Slurm or Local backend.

4. Optionally triggers analysis based on explicit config.

class dpeva.workflows.TrainingWorkflow(config: Dict | TrainingConfig)

基类：object

Workflow for parallel fine-tuning of DeepMD models. Supports both local multiprocessing and Slurm submission. Refactored using DDD Managers.

run()

Executes the training workflow.

1. Configures workflow logging.

2. Generates task-specific configurations and seeds.

3. Creates task directories and copies base models.

4. Generates submission scripts (Slurm or Local).

5. Submits jobs to the execution backend.