ML Flow

Voici la liste complète des méthodes de la classe MLflowClient dans la version 2.17.2 :

Méthodes de gestion des expériences :

1. create_experiment(name, artifact_location=None, tags=None)
2. get_experiment(experiment_id)
3. get_experiment_by_name(name)
4. list_experiments(view_type=ViewType.ACTIVE_ONLY, max_results=None, page_token=None)
5. delete_experiment(experiment_id)
6. restore_experiment(experiment_id)
7. rename_experiment(experiment_id, new_name)
8. set_experiment_tag(experiment_id, key, value)
9. delete_experiment_tag(experiment_id, key)

Méthodes de gestion des runs :

1. create_run(experiment_id, start_time=None, tags=None, run_name=None)
2. get_run(run_id)
3. delete_run(run_id)
4. restore_run(run_id)
5. list_run_infos(experiment_id, run_view_type=ViewType.ACTIVE_ONLY, max_results=None, page_token=None)
6. search_runs(experiment_ids=None, filter_string='', run_view_type=ViewType.ACTIVE_ONLY, max_results=100, order_by=None, page_token=None)
7. set_terminated(run_id, status=None, end_time=None)
8. log_param(run_id, key, value)
9. log_params(run_id, params)
10. log_metric(run_id, key, value, timestamp=None, step=None)
11. log_metrics(run_id, metrics, timestamp=None, step=None)
12. set_tag(run_id, key, value)
13. set_tags(run_id, tags)
14. delete_tag(run_id, key)
15. log_batch(run_id, metrics=None, params=None, tags=None, dataset_inputs=None)
16. get_metric_history(run_id, key)
17. get_metric_history_bulk(run_id, metric_keys)

Méthodes de gestion des artefacts :

1. log_artifact(run_id, local_path, artifact_path=None)
2. log_artifacts(run_id, local_dir, artifact_path=None)
3. list_artifacts(run_id, path=None)
4. download_artifacts(run_id, path, dst_path=None)
5. upload_artifact(run_id, artifact_file, artifact_path=None)
6. upload_artifacts(run_id, artifact_dir, artifact_path=None)
7. get_artifact_uri(run_id, artifact_path=None)
8. download_artifact_from_uri(artifact_uri, output_path=None)

Méthodes du registre de modèles :

1. create_registered_model(name, tags=None, description=None)
2. get_registered_model(name)
3. search_registered_models(filter_string='', max_results=None, order_by=None, page_token=None)
4. delete_registered_model(name)
5. rename_registered_model(name, new_name)
6. update_registered_model(name, description=None)
7. set_registered_model_tag(name, key, value)
8. delete_registered_model_tag(name, key)
9. get_latest_versions(name, stages=None)
10. create_model_version(name, source, run_id=None, tags=None, run_link=None, description=None)
11. get_model_version(name, version)
12. delete_model_version(name, version)
13. update_model_version(name, version, description=None)
14. transition_model_version_stage(name, version, stage, archive_existing_versions=False)
15. set_model_version_tag(name, version, key, value)
16. delete_model_version_tag(name, version, key)
17. get_model_version_download_uri(name, version)
18. search_model_versions(filter_string='', max_results=None, order_by=None, page_token=None)

Méthodes de gestion des datasets :

1. create_dataset(name, source=None, schema=None, profile=None, tags=None, description=None)
2. get_dataset(name)
3. delete_dataset(name)
4. search_datasets(filter_string='', max_results=None, order_by=None, page_token=None)
5. set_dataset_tag(name, key, value)
6. delete_dataset_tag(name, key)
7. create_dataset_version(dataset_name, source, schema=None, profile=None, description=None, tags=None)
8. get_dataset_version(dataset_name, version)
9. delete_dataset_version(dataset_name, version)
10. search_dataset_versions(filter_string='', max_results=None, order_by=None, page_token=None)
11. set_dataset_version_tag(dataset_name, version, key, value)
12. delete_dataset_version_tag(dataset_name, version, key)

Méthodes pour les entrées de run :

1. log_inputs(run_id, datasets=None, tags=None)

Méthodes diverses :

1. set_tracking_uri(uri)
2. get_tracking_uri()
3. set_registry_uri(uri)
4. get_registry_uri()

Cette liste couvre toutes les méthodes publiques disponibles dans la classe MLflowClient pour la version 2.17.2, conformément au code source disponible sur le dépôt GitHub de MLflow.