Enums

ActivationFunctions

class prescyent.utils.enums.activation_functions.ActivationFunctions(value)

Bases: str, Enum

Map to a given activation function

GELU = 'gelu'

torch.nn.GELU

RELU = 'relu'

torch.nn.ReLU

SIGMOID = 'sigmoid'

torch.nn.Sigmoid

---

LearningTypes

class prescyent.utils.enums.learning_types.LearningTypes(value)

Bases: str, Enum

Method used to generate the MotionDataSample in the dataloaders

AUTOREG = 'auto_regressive'

Generate pairs with the following behavior, with the expected in points and features and out points and features : - given a time step T, history_size H and future_size F we have: x a sequence of lenght H with the frames = [T-H, .. T] y a sequence of lenght H with the frames = [T-H+1, … T+1]

SEQ2ONE = 'sequence_2_one'

Generate pairs with the following behavior, with the expected in points and features and out points and features : - given a time step T, history_size H and future_size F we have: x a sequence of lenght H with the frames = [T-H, .. T] y a sequence of lenght 1 with the frames = [T+F]

SEQ2SEQ = 'sequence_2_sequence'

Generate pairs with the following behavior, with the expected in points and features and out points and features : - given a time step T, history_size H and future_size F we have: x a sequence of lenght H with the frames = [T-H, .. T] y a sequence of lenght F with the frames = [T+1, … T+F]

---

LossFunctions

class prescyent.utils.enums.loss_functions.LossFunctions(value)

Bases: str, Enum

Map to the required loss function in the torch_module class

CROSSENTROPYLOSS = 'crossentropyloss'

torch.nn.CrossEntropyLoss

HINGEEMBEDDINGLOSS = 'hingeembeddingloss'

torch.nn.HingeEmbeddingLoss

KLDIVLOSS = 'kldivloss'

torch.nn.KLDivLoss

L1LOSS = 'l1loss'

torch.nn.L1Loss

MARGINRANKINGLOSS = 'marginrankingloss'

torch.nn.MarginRankingLoss

MFDLOSS = 'mfdloss'

MeanFinalDistanceLoss

MSELOSS = 'mseloss'

torch.nn.MSELoss

MTDLOSS = 'mtdloss'

MeanTotalDistanceLoss

MTDVLOSS = 'mtdvloss'

MeanTotalDistanceAndVelocityLoss

NLLLOSS = 'nllloss'

torch.nn.NLLLoss

TRIPLETMARGINLOSS = 'tripletmarginloss'

torch.nn.TripletMarginLoss

---

Profilers

class prescyent.utils.enums.profilers.Profilers(value)

Bases: str, Enum

Map to a pytorch_lightning.profilers in the lightnin predictor

ADVANCED = 'advanced'

pytorch_lightning.profilers.AdvancedProfiler

SIMPLE = 'simple'

pytorch_lightning.profilers.SimpleProfiler

TORCH = 'torch'

pytorch_lightning.profilers.PyTorchProfiler

---

Scalers

class prescyent.utils.enums.scalers.Scalers(value)

Bases: str, Enum

Map to a given scaling method

NORMALIZATION = 'normalization'

STANDARDIZATION = 'standardization'

TrajectoryDimensions

class prescyent.utils.enums.trajectory_dimensions.TrajectoryDimensions(value)

Bases: str, Enum

Dimensions ids of batched trajectories

ALL = '[0, 1, 2, 3]'

BATCH = '[0]'

FEATURE = '[3]'

POINT = '[2]'

SPATIAL = '[2, 3]'

TEMPORAL = '[1]'