deeppavlov.core.models

Abstract model classes and interfaces.

class deeppavlov.core.models.component.Component

Abstract class for all callables that could be used in Chainer’s pipe.

class deeppavlov.core.models.serializable.Serializable(save_path: Optional[Union[pathlib.Path, str]], load_path: Optional[Union[pathlib.Path, str]] = None, mode: str = 'infer', *args, **kwargs)

Abstract base class that expresses the interface for all models that can serialize data to a path.

class deeppavlov.core.models.estimator.Estimator(save_path: Optional[Union[pathlib.Path, str]], load_path: Optional[Union[pathlib.Path, str]] = None, mode: str = 'infer', *args, **kwargs)

Abstract class for components that could be fitted on the data as a whole.

class deeppavlov.core.models.nn_model.NNModel(save_path: Optional[Union[pathlib.Path, str]], load_path: Optional[Union[pathlib.Path, str]] = None, mode: str = 'infer', *args, **kwargs)

Abstract class for deep learning components.

class deeppavlov.core.models.torch_model.TorchModel(model: torch.nn.Module, device: Union[torch.device, str] = 'cuda', optimizer: str = 'AdamW', optimizer_parameters: Optional[dict] = None, learning_rate_drop_patience: Optional[int] = None, learning_rate_drop_div: Optional[float] = None, load_before_drop: bool = True, min_learning_rate: float = 1e-07, clip_norm: Optional[float] = None, *args, **kwargs)

Class implements torch model’s main methods.

Parameters

• model – torch.nn.Model-based neural network model

• device – device to use

• optimizer – name of torch.optim optimizer

• optimizer_parameters – dictionary with optimizer parameters

• learning_rate_drop_patience – how many validations with no improvements to wait

• learning_rate_drop_div – the divider of the learning rate after learning_rate_drop_patience unsuccessful validations

• load_before_drop – whether to load best model before dropping learning rate or not

• min_learning_rate – min value of learning rate if learning rate decay is used

• args –

• kwargs – dictionary with other model parameters

device

cpu or cuda device to use

opt

dictionary with all model parameters

model

torch model

epochs_done

number of epochs that were done

optimizer

torch.optim instance

learning_rate_drop_patience

how many validations with no improvements to wait

learning_rate_drop_div

the divider of the learning rate after learning_rate_drop_patience unsuccessful validations

load_before_drop

whether to load best model before dropping learning rate or not

min_learning_rate

min value of learning rate if learning rate decay is used

clip_norm

clip gradients by norm coefficient