nvflare.app_common.utils package

Submodules

• nvflare.app_common.utils.component_utils module
  • check_component_type()
• nvflare.app_common.utils.error_handling_utils module
  • get_error_handling_message()
  • should_ignore_result_error()
• nvflare.app_common.utils.export_utils module
  • update_export_props()
• nvflare.app_common.utils.file_utils module
  • get_ext_format()
  • get_file_ext()
  • get_file_format()
  • resolve_path_under_root()
• nvflare.app_common.utils.fl_component_wrapper module
  • FLComponentWrapper
    • FLComponentWrapper.STATE
    • FLComponentWrapper.abort()
    • FLComponentWrapper.critical()
    • FLComponentWrapper.debug()
    • FLComponentWrapper.error()
    • FLComponentWrapper.event()
    • FLComponentWrapper.exception()
    • FLComponentWrapper.finalize()
    • FLComponentWrapper.get_component()
    • FLComponentWrapper.get_context_prop()
    • FLComponentWrapper.get_shareable_header()
    • FLComponentWrapper.info()
    • FLComponentWrapper.initialize()
    • FLComponentWrapper.is_aborted()
    • FLComponentWrapper.panic()
    • FLComponentWrapper.stop_task()
    • FLComponentWrapper.warning()
• nvflare.app_common.utils.fl_model_utils module
  • FLModelUtils
    • FLModelUtils.from_dxo()
    • FLModelUtils.from_model_learnable()
    • FLModelUtils.from_shareable()
    • FLModelUtils.get_configs()
    • FLModelUtils.get_meta_prop()
    • FLModelUtils.set_meta_prop()
    • FLModelUtils.to_dxo()
    • FLModelUtils.to_model_learnable()
    • FLModelUtils.to_shareable()
    • FLModelUtils.update_model()
• nvflare.app_common.utils.json_utils module
  • ObjectEncoder
    • ObjectEncoder.default()
• nvflare.app_common.utils.math_utils module
  • parse_compare_criteria()
• nvflare.app_common.utils.tensor_disk_offload_context module
  • apply_enable_tensor_disk_offload()
  • restore_enable_tensor_disk_offload()

Module contents

get_error_handling_message(ignore_result_error: bool | None, client_name: str, error_code: Any, current_round: int | None, controller_name: str, failed_clients: Set[str], num_targets: int, min_responses: int) → str

Generate appropriate log message based on error handling mode.

Parameters:

• ignore_result_error – The error handling mode (None, False, or True).

• client_name – Name of the client with the error.

• error_code – The return code from the client result (ReturnCode constant or None).

• current_round – Current training round (may be None if not set in result).

• controller_name – Name of the controller class.

• failed_clients – Set of client names that have failed.

• num_targets – Total number of target clients.

• min_responses – Minimum number of responses required.

Returns:

Appropriate message string for logging.

should_ignore_result_error(ignore_result_error: bool | None, client_name: str, failed_clients: Set[str], num_targets: int, min_responses: int) → bool

Determine whether a client result error should be ignored or cause a panic.

This function implements the three-mode error handling policy: - None (Dynamic): Ignore errors if min_responses can still be reached, panic otherwise. - False (Strict): Never ignore errors, always panic. - True (Resilient): Always ignore errors, never panic.

Note: This function can be safely called multiple times for the same client error. The failed_clients set uses idempotent add() operations, so duplicate calls for the same client will not affect the remaining count calculation.

Parameters:

• ignore_result_error – The error handling mode. - None: Dynamic mode - ignore if min_responses still reachable. - False: Strict mode - always panic on error. - True: Resilient mode - always ignore errors.

• client_name – Name of the client with the error.

• failed_clients – Set of client names that have already failed (will be updated in dynamic mode only).

• num_targets – Total number of target clients for the current task.

• min_responses – Minimum number of responses required.

Returns:

True if the error should be ignored (no panic needed). False if a panic should be triggered.