ClassificationTracer Class

ml_debugger.training.classification.classification_torchtracer.ClassificationTrainingTorchTracer

Bases: TrainingTracer, ClassificationTorchTracer

TrainingTracer for classification tasks using PyTorch models.

__init__(model, model_name, version_name, target_layers=None, additional_fields=None, additional_label_fields=None, auto_sync=False, force_table_recreate=False, api_endpoint=None, api_key=None)

Initialize classification tracer.

Parameters:

Name	Type	Description	Default
model	Module	PyTorch model to trace.	required
model_name	str	Name of the ML model.	required
version_name	str	Version identifier for the ML model.	required
target_layers	Optional[Dict[str, str]]	Mapping of layer aliases to module paths.	None
additional_fields	Optional[List[dict]]	Extra fields for database schema.	None
additional_label_fields	Optional[List[dict]]	Extra fields for groundtruth database schema.	None
auto_sync	bool	Enable backgroup syncing of logged data.	False
force_table_recreate	bool	Whether to drop and recreate existing tables.	False
api_endpoint	Optional[str]	URL of the service API for data upload.	None
api_key	Optional[str]	API key for authenticating with the service.	None

__call__(model_input, ground_truth, input_ids, dataset_type='train', n_epoch=None, **kwargs)

Invoke the tracer on a single inference, recording I/O data.

Parameters:

Name	Type	Description	Default
model_input	Any	Input data for the model inference.	required
ground_truth	Optional[Any]	Optional ground truth corresponding to input.	required
input_ids	List[str]	Identifiers of each input data.	required
dataset_type	str	Identifier of input dataset. (e.g. 'train', 'val', 'test', 'deploying')	'train'
n_epoch	Optional[int]	Number identifier of training iteration.	None
**kwargs	Any	Additional keyword arguments for parsing and saving I/O data. (will be passed to self._parse_and_save_io_data)	{}

Returns:

Name	Type	Description
Any	Any	The raw output produced by the model.

Raises:

Type	Description
Any	Propagates exceptions from parsing and saving operations.

get_hooked_features(layer_name)

Retrieve the captured output for a given layer alias.

Parameters:

Name	Type	Description	Default
layer_name	str	Alias of the layer whose activation was captured.	required

Returns:

Name	Type	Description
Any	Any	Activation data stored for the specified layer.

Raises:

Type	Description
KeyError	If no activation has been captured for layer_name.

export(output_path=None, n_epoch=None)

Export extracted features and ground truth data into a ZIP archive.

Parameters:

Name	Type	Description	Default
output_path	Optional[str]	Path or directory for saving the ZIP file. If no .zip extension, the default filename is appended. Defaults to cwd.	None
n_epoch	Optional[int]	Target epoch number. If None, exports all epochs.	None

Returns:

Type	Description
Optional[Path]	Path to the created ZIP file, or None on non-primary distributed ranks.

wait_for_save(n_epoch=None, interval=3, *, force=False)

upload(n_epoch=None, *, force=False)

drop_epoch_data(n_epoch)

Delete cached rows scoped to a single epoch.

Parameters:

Name	Type	Description	Default
n_epoch	int	削除対象の epoch 番号 (int 必須)。"None = 全件削除" と 取り違えるリスクを避けるため、本 API は int のみ受け付ける (Optional[int] ではない)。	required

Returns:

Type	Description
Dict[str, int]	Dict[str, int]: {'extracted_feature': N, 'ground_truth': M} 形式の 削除行数。ground_truth が無い (logging_mode) 場合はキーが含まれない。

Raises:

Type	Description
TypeError	n_epoch が int でない場合。

drop_version_data()

Drop all tables for the current (model, version) and recreate them empty.

既存の force_table_recreate=True 経路と同じ挙動を明示メソッドとして 提供する。内部的には __initialize_session(force_table_recreate=True) を 呼び直し、新しい repository / async batch writer を起動する。

Note

re-init 前に async_batch_writer を flush() + stop() で完全停止 させてから session を再初期化する。これを怠ると daemon thread が旧 session を握ったまま残ってしまう。