training module

class training.Training(client, uuid)

Bases: object

cancel()

cancel this training

property checkpoints

Get training checkpoints

Returns:

AttrDict: checkpoints

delete()

delete this training

Returns:

AttrDict: API reponse

property info

Get verbose info about the training

Returns:

AttrDict: The API response

property logs

Get training logs

Returns:

AttrDict: logs

monitor()

Plot an interactive graph monitoring training metrics

restart()

restart this training

Returns:

AttrDict: API reponse

save(checkpoint_id=None, name=None)

Save one of the training checkpoints (if any) as a trained model

Args:

checkpoint_id (str): id of the checkpoint to save, defaults to lates name (str): name of the saved model, defaults to auto generated

Returns:

upon success, a description of the newly created trained model.

stop()

stop this training

Returns:

AttrDict: API reponse

static submit(client, dataset_id, model_class='ncs.models.point_regressor.PointRegressor', network_complexity=0, log_transform_input_fields=False, normalize_input_fields=False, log_transform_input_scalars=False, normalize_input_scalars=False, log_transform_output_fields=False, normalize_output_fields=False, log_transform_output_scalars=False, normalize_output_scalars=False, restore_from=None, user_config=None)

Submit a new training to the API

Args:

client: API client dataset_id (str): uuid or name of the dataset to use for training. model_class (str): class of model to use. network_complexity (float): complexity of the model from 0 to 10 (based on EfficientNet framework). log_transform_input_fields (bool): whether to log transform the input fields in the dataset. normalize_input_fields (bool): whether to normalize the input fields in dataset. log_transform_input_scalars (bool): whether to log transform the input scalars in the dataset. normalize_input_scalars (bool): whether to normalize the input scalars in dataset. log_transform_output_fields (bool): whether to log transform the output fields in the dataset. normalize_output_fields (bool): whether to normalize the output fields in dataset. log_transform_output_scalars (bool): whether to log transform the output scalars in the dataset. normalize_output_scalars (bool): whether to normalize the output scalars in dataset. restore_from (str): uuid or name of a trained model to restore the training from. user_config (dict): custom config override.

Returns:

New Training