morphoclass.training.training_config module

Implementation of the training config.

class morphoclass.training.training_config.TrainingConfig(model_class: str, model_params: dict, splitter_class: str, splitter_params: dict, dataset_name: str, features_dir: pathlib.Path | None, optimizer_class: str, optimizer_params: dict, n_epochs: int, batch_size: int, seed: int, input_csv: pathlib.Path | None = None, oversampling: bool = False, neurite_type: str | None = None, train_all_samples: bool = False, checkpoint_path_pretrained: pathlib.Path | None = None, frozen_backbone: bool = False)

Bases: object

A training configuration.

batch_size: int

checkpoint_path_pretrained: pathlib.Path | None = None

dataset_name: str

features_dir: pathlib.Path | None

classmethod from_dict(data: dict, workdir: pathlib.Path | None = None) → TrainingConfig

Construct a training config from a dictionary.

classmethod from_file(path: pathlib.Path, workdir: pathlib.Path | None = None) → TrainingConfig

Read the training config from a YAML file.

classmethod from_separate_configs(conf_model: dict, conf_splitter: dict, features_dir: StrPath, workdir: pathlib.Path | None = None) → TrainingConfig

Construct a training config from separate configs.

frozen_backbone: bool = False

static import_obj(obj_full_name: str) → object

Import an object given its full module and class path.

input_csv: pathlib.Path | None = None

model_class: str

property model_cls

Get the model class.

model_params: dict

n_epochs: int

neurite_type: str | None = None

optimizer_class: str

property optimizer_cls

Get the optimizer class.

optimizer_params: dict

oversampling: bool = False

resolve_paths(workdir: pathlib.Path) → None

Resolve relative internal paths.

seed: int

splitter_class: str

property splitter_cls

Get the splitter class.

splitter_params: dict

train_all_samples: bool = False