yamle.data.regression module#

class yamle.data.regression.ToyRegressionDataModule(num_samples, test_samples, *args, **kwargs)[source]#

Bases: BaseDataModule

Data module for a toy regression 1D dataset.

Implements a toy regression dataset with 1D inputs and 1D outputs.

$ y = sin(3x) + x^2 - 0.3 + mathcal{N}(0, 0.2) + mathcal{n}(0, gaussian_noise_sigma) $ on the interval $ [-1, 1] $. but with a gap in the middle from $ [-0.2, 0.2] $.

The validation and test data is on intervals $ [-2, 2] $ and witout $mathcal{N}(0, 1)$.

Parameters:

  • num_samples (int) – Number of samples to generate.

  • test_samples (int) – Number of samples to generate for the test set.

outputs_dim = 2#
targets_dim = 1#
outputs_dtype = torch.float32#
inputs_dim = (1,)#
task = 'regression'#
prepare_data()[source]#

Generate the toy regression dataset.

Return type:

None

plot(tester, save_path, specific_name='')[source]#

Plot the data and the uncertainty of the model.

Return type:

None

static add_specific_args(parent_parser)[source]#

Add dataset specific arguments to the parser.

Return type:

ArgumentParser

available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.UCIRegressionDataModule(dataset, *args, **kwargs)[source]#

Bases: RealWorldRegressionDataModule

Data module for the UCI regression datasets.

Currently supports the following datasets:

  • Concrete

  • Energy

  • Boston

  • Wine

  • Yacht

  • Abalone

  • Telemonitoring

Parameters:

dataset (str) – Name of the dataset to use.

task = 'regression'#
outputs_dim = 2#
outputs_dtype = torch.float32#
inputs_dtype = torch.float32#
plot(tester, save_path, specific_name='')[source]#

Plots the dataset.

Return type:

None

available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.MedMNISTRegressionDataModule(dataset, pad_to_32=True, *args, **kwargs)[source]#

Bases: VisionRegressionDataModule

Data module for the MedMNIST dataset.

Parameters:

  • dataset (str) – Name of the dataset to use.

  • pad_to_32 (bool) – Whether to pad the images to 32x32. Defaults to True.

outputs_dtype = torch.float32#
prepare_data()[source]#

Download and prepare the data, the data is stored in self._train_dataset, self._validation_dataset and self._test_dataset.

Return type:

None

static add_specific_args(parent_parser)[source]#

Add dataset specific arguments to the parser.

Return type:

ArgumentParser

available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.RetinaMNISTDataModule(*args, **kwargs)[source]#

Bases: MedMNISTRegressionDataModule

Data module for the RetinaMNIST dataset.

inputs_dim = (3, 28, 28)#
outputs_dim = 2#
targets_dim = 1#
mean: Tuple[float, ...] = (0.3974, 0.2446, 0.1554)#
std: Tuple[float, ...] = (0.2977, 0.2001, 0.1503)#
available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.ConcreteUCIRegressionDataModule(*args, **kwargs)[source]#

Bases: UCIRegressionDataModule

Data module for the Concrete UCI regression dataset.

inputs_dim = (8,)#
available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.EnergyUCIRegressionDataModule(*args, **kwargs)[source]#

Bases: UCIRegressionDataModule

Data module for the Energy UCI regression dataset.

inputs_dim = (9,)#
available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.BostonUCIRegressionDataModule(*args, **kwargs)[source]#

Bases: UCIRegressionDataModule

Data module for the Boston UCI regression dataset.

inputs_dim = (13,)#
available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.WineQualityUCIRegressionDataModule(*args, **kwargs)[source]#

Bases: UCIRegressionDataModule

Data module for the Wine Quality UCI regression dataset.

inputs_dim = (11,)#
available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.YachtUCIRegressionDataModule(*args, **kwargs)[source]#

Bases: UCIRegressionDataModule

Data module for the Yacht Hydrodynamics UCI regression dataset.

inputs_dim = (6,)#
available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.AbaloneUCIRegressionDataModule(*args, **kwargs)[source]#

Bases: UCIRegressionDataModule

Data module for the Abalone UCI regression dataset.

inputs_dim = (10,)#
available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.TelemonitoringUCIRegressionDataModule(*args, **kwargs)[source]#

Bases: UCIRegressionDataModule

Data module for the Telemonitoring UCI regression dataset.

We use the subject# column to split the data manually between train, validation, calibration and test sets to ensure that a patient does not appear in more than one set.

We predict the total_UPDRS column.

inputs_dim = (20,)#
prepare_data()[source]#

Prepares the data for training, validation, calibration and testing.

Return type:

None

available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.TimeSeriesRegressionDataModule(time_window, *args, **kwargs)[source]#

Bases: RealWorldRegressionDataModule

This is a wrapper class for time series regression datasets.

Parameters:

time_window (int) – The number of time steps to use for creating the time series.

plot(tester, save_path, specific_name='')[source]#

Plots the predictions of the model on the test set.

Parameters:

  • tester (LightningModule) – The model to test.

  • save_path (str) – The path to save the plot.

  • specific_name (str) – A specific name to add to the plot name.

Return type:

None

static add_specific_args(parent_parser)[source]#

Add dataset specific arguments to the parser.

Return type:

ArgumentParser

prepare_data()[source]#

Prepares the data for training, validation, and testing.

Return type:

None

available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.TemperatureTimeSeriesDataModule(time_window, *args, **kwargs)[source]#

Bases: TimeSeriesRegressionDataModule

Data module for temperature time series regression.

outputs_dim = 2#
inputs_dim = (1,)#
task = 'regression'#
available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.WikiFaceDataset(root_dir, seed=42, image_size=(64, 64), test_portion=0.1, transform=None, split='train')[source]#

Bases: Dataset

Dataset for the WikiFace age prediction task.

The dataset is the wikipedia portion of the dataset: https://data.vision.ee.ethz.ch/cvl/rrothe/imdb-wiki/

Parameters:

  • root_dir (str) – The root directory to store the dataset.

  • seed (int) – The seed for the random number generator.

  • image_size (Tuple[int, int]) – The size of the image.

  • test_portion (float) – The portion of the dataset to use for testing.

  • transform (Optional[Callable]) – The transform to apply to the image.

  • split (str) – The split of the dataset to use. Either ‘train’ or ‘test’.

md5sum = 'f536eb7f5eae229ae8f286184364b42b'#
class yamle.data.regression.TorchvisionRegressionDataModule(dataset, *args, **kwargs)[source]#

Bases: VisionRegressionDataModule

outputs_dtype = torch.float32#
prepare_data()[source]#

Download and prepare the data, the data is stored in self._train_dataset and self._test_dataset.

Return type:

None

available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.WikiFaceRegressionDataModule(*args, **kwargs)[source]#

Bases: TorchvisionRegressionDataModule

Data module for the WikiFace dataset.

inputs_dim = (3, 64, 64)#
outputs_dim = 2#
targets_dim = 1#
outputs_dtype = torch.float32#
mean: Tuple[float, ...] = (0.4802, 0.4481, 0.3975)#
std: Tuple[float, ...] = (0.2302, 0.2265, 0.2262)#
available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.TorchvisionRotationRegressionDataModule(dataset, pad_to_32=False, min_angle=0, max_angle=90, *args, **kwargs)[source]#

Bases: VisionRegressionDataModule

Data module for the rotation regression task.

It can load the most common torchvision datasets and apply a rotation to the images to create a rotation regression task. The task is to predict the rotation of the image.

Parameters:

  • dataset (str) – The dataset to load. Either ‘mnist’, ‘fashionmnist’, ‘cifar10’, ‘cifar100’, ‘svhn’, ‘tinyimagenet’.

  • pad_to_32 (bool) – Whether to pad the images to 32x32.

  • min_angle (float) – The minimum angle to rotate the image. Defaults to 0.

  • max_angle (float) – The maximum angle to rotate the image. Defaults to 90.

outputs_dim = 2#
targets_dim = 1#
outputs_dtype = torch.float32#
prepare_data()[source]#

Download and prepare the data, the data is stored in self._train_dataset, self._validation_dataset and self._test_dataset.

Return type:

None

static add_specific_args(parent_parser)[source]#

This method is used to add datamodel specific arguments to the general parser.

Return type:

ArgumentParser

available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.TinyImageNetRotationRegressionDataModule(*args, **kwargs)[source]#

Bases: TorchvisionRotationRegressionDataModule

Data module for the rotation regression task on TinyImageNet.

inputs_dim = (3, 64, 64)#
mean: Tuple[float, ...] = (0.4802, 0.4481, 0.3975)#
std: Tuple[float, ...] = (0.2302, 0.2265, 0.2262)#
available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.TorchvisionRotationRegressionDataModuleMNIST(*args, **kwargs)[source]#

Bases: TorchvisionRotationRegressionDataModule

Data module for the rotation regression task on MNIST.

inputs_dim = (1, 28, 28)#
mean: Tuple[float, ...] = (0.1307,)#
std: Tuple[float, ...] = (0.3081,)#
prepare_data()[source]#

Download and prepare the data, the data is stored in self._train_dataset, self._validation_dataset and self._test_dataset.

Return type:

None

available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.TorchvisionRotationRegressionDataModuleFashionMNIST(*args, **kwargs)[source]#

Bases: TorchvisionRotationRegressionDataModule

Data module for the rotation regression task on FashionMNIST.

inputs_dim = (1, 28, 28)#
mean: Tuple[float, ...] = (0.286,)#
std: Tuple[float, ...] = (0.353,)#
prepare_data()[source]#

Download and prepare the data, the data is stored in self._train_dataset, self._validation_dataset and self._test_dataset.

Return type:

None

available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.TorchvisionRotationRegressionDataModuleSVHN(*args, **kwargs)[source]#

Bases: TorchvisionRotationRegressionDataModule

Data module for the rotation regression task on SVHN.

inputs_dim = (3, 32, 32)#
mean: Tuple[float, ...] = (0.4377, 0.4438, 0.4728)#
std: Tuple[float, ...] = (0.198, 0.201, 0.197)#
available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.TorchvisionRotationRegressionDataModuleCIFAR10(*args, **kwargs)[source]#

Bases: TorchvisionRotationRegressionDataModule

Data module for the rotation regression task on CIFAR10.

inputs_dim = (3, 32, 32)#
mean: Tuple[float, ...] = (0.4914, 0.4822, 0.4465)#
std: Tuple[float, ...] = (0.2023, 0.1994, 0.201)#
available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#
class yamle.data.regression.TorchvisionRotationRegressionDataModuleCIFAR100(*args, **kwargs)[source]#

Bases: TorchvisionRotationRegressionDataModule

Data module for the rotation regression task on CIFAR100.

inputs_dim = (3, 32, 32)#
mean: Tuple[float, ...] = (0.4914, 0.4822, 0.4465)#
std: Tuple[float, ...] = (0.2023, 0.1994, 0.201)#
available_transforms: List[str]#
available_test_augmentations: List[str]#
test_augmentations: List[str]#