Helper Module for Deep Learning.

Core classes.

class pynet.core.Base(optimizer_name='Adam', learning_rate=0.001, loss_name='NLLLoss', metrics=None, use_cuda=False, pretrained=None, resume=False, add_labels=False, **kwargs)[source]¶

Base class for perform Deep Learning training.

__init__(optimizer_name='Adam', learning_rate=0.001, loss_name='NLLLoss', metrics=None, use_cuda=False, pretrained=None, resume=False, add_labels=False, **kwargs)[source]¶

Class instantiation.

Observers will be notified, allowed signals are: - ‘before_epoch’ - ‘after_epoch’ - ‘kernel_regularizer’

Parameters

optimizer_name: str, default ‘Adam’

the name of the optimizer: see ‘torch.optim’ for a description of available optimizer.

learning_rate: float, default 1e-3

the optimizer learning rate.

loss_name: str, default ‘NLLLoss’

the name of the loss: see ‘torch.nn’ for a description of available loss.

metrics: list of str

a list of extra metrics that will be computed.

use_cuda: bool, default False

wether to use GPU or CPU.

pretrained: path, default None

path to the pretrained model or weights.

resume: bool, default False

if set to true, the code will restore the weights of the model but also restore the optimizer’s state, as well as the hyperparameters used, and the scheduler.

add_labels: bool, default False

if set and labels are specified in the data manager, add the labels to the forward function parameters.

kwargs: dict

specify directly a custom ‘model’, ‘optimizer’ or ‘loss’. Can also be used to set specific optimizer parameters.

test(loader, with_logit=False, logit_function='softmax', predict=False, concat_layer_outputs=None, is_validation=False)[source]¶

Evaluate the model on the test or validation data.

Parameters

loader: a pytorch Dataset

the data laoder.

with_logit: bool, default False

apply the logit function to the result.

logit_funtction: str, default ‘softmax’

choose the logit function.

predict: bool, default False

take the argmax over the channels.

concat_layer_outputs: list of str, default None

the outputs of the intermediate layers to be merged with the predicted data (must be the same size).

is_validation: bool default False

specify if we are in the validation phase.

Returns

y: array-like

the predicted data.

loss: float

the value of the loss function.

values: dict

the values of the metrics.

testing(manager, with_logit=False, logit_function='softmax', predict=False, concat_layer_outputs=None)[source]¶

Evaluate the model.

Parameters

manager: a pynet DataManager

a manager containing the test data.

with_logit: bool, default False

apply the logit function to the result.

logit_function: str, default ‘softmax’

choose the logit function.

predict: bool, default False

take the argmax over the channels.

concat_layer_outputs: list of str, default None

the outputs of the intermediate layers to be merged with the predicted data (must be the same size).

Returns

y: array-like

the predicted data.

X: array-like

the input data.

y_true: array-like

the true data if available.

loss: float

the value of the loss function if true data availble.

values: dict

the values of the metrics if true data availble.

train(loader)[source]¶

Train the model on the trained data.

Parameters

loader: a pytorch Dataset

the data laoder.

Returns

loss: float

the value of the loss function.

values: dict

the values of the metrics.

training(manager, nb_epochs, checkpointdir=None, fold_index=None, scheduler=None, with_validation=True, save_after_epochs=1, add_labels=False)[source]¶

Train the model.

Parameters

manager: a pynet DataManager

a manager containing the train and validation data.

nb_epochs: int, default 100

the number of epochs.

checkpointdir: str, default None

a destination folder where intermediate models/histories will be saved.

fold_index: int, default None

the index of the fold to use for the training, default use all the available folds.

scheduler: torch.optim.lr_scheduler, default None

a scheduler used to reduce the learning rate.

with_validation: bool, default True

if set use the validation dataset.

save_after_epochs: int, default 1

determines when the model is saved and represents the number of epochs before saving.

Returns

train_history, valid_history: History

the train/validation history.