torchts package

Submodules

Base classes

class torchts.nn.model.TimeSeriesModel(optimizer, optimizer_args=None, criterion=<function mse_loss>, criterion_args=None, scheduler=None, scheduler_args=None, scaler=None)

Bases: pytorch_lightning.core.lightning.LightningModule

Base class for all TorchTS models.

Parameters

• optimizer (torch.optim.Optimizer) – Optimizer

• opimizer_args (dict) – Arguments for the optimizer

• criterion – Loss function

• criterion_args (dict) – Arguments for the loss function

• scheduler (torch.optim.lr_scheduler) – Learning rate scheduler

• scheduler_args (dict) – Arguments for the scheduler

• scaler (torchts.utils.scaler.Scaler) – Scaler

configure_optimizers()

Configure optimizer.

Returns

torch.optim.Optimizer – Optimizer

fit(x, y, max_epochs=10, batch_size=128)

Fits model to the given data.

Parameters

• x (torch.Tensor) – Input data

• y (torch.Tensor) – Output data

• max_epochs (int) – Number of training epochs

• batch_size (int) – Batch size for torch.utils.data.DataLoader

abstract forward(x, y=None, batches_seen=None)

Forward pass.

Parameters

x (torch.Tensor) – Input data

Returns

torch.Tensor – Predicted data

predict(x)

Runs model inference.

Parameters

x (torch.Tensor) – Input data

Returns

torch.Tensor – Predicted data

test_step(batch, batch_idx)

Tests model for one step.

Parameters

• batch (torch.Tensor) – Output of the torch.utils.data.DataLoader

• batch_idx (int) – Integer displaying index of this batch

training_step(batch, batch_idx)

Trains model for one step.

Parameters

• batch (torch.Tensor) – Output of the torch.utils.data.DataLoader

• batch_idx (int) – Integer displaying index of this batch

validation_step(batch, batch_idx)

Validates model for one step.

Parameters

• batch (torch.Tensor) – Output of the torch.utils.data.DataLoader

• batch_idx (int) – Integer displaying index of this batch

Module contents