Wraps the iterable with tqdm for global rank zero. Args: iterable: the iterable to wrap with tqdm total: the total le

Multiply two quaternions. Usual torch rules for broadcasting apply. Args: a: Quaternions as tensor of shape (..., 4), real part

Perform per-sample random masking by per-sample shuffling. Per-sample shuffling is done by argsort random noise. Args:

Saves a checkpoint to the ``checkpoint_dir`` Args: state: A mapping containing model, optimizer and lr scheduler.

Saves a checkpoint to the ``checkpoint_dir`` Args: state: A mapping containing model, optimizer and lr scheduler.

This function disables printing when not in master process

Steps the learning rate scheduler if necessary. Args: scheduler_cfg: The learning rate scheduler configuration. H

A single training step, running forward and backward. The optimizer step is called separately, as this is given as a closure to the optimizer

Args: img_size (int, tuple): input image size patch_size (int, tuple): patch size in_chans (int): number

Returns list of tensors of shape [B, N, D], one for each mask-pred.

Collects the currently assigned signal handlers.

We use the slip horizon labels to debounce fast switches in slip events, which is likely noise if there is atleast one slip event in

Recursively parses the output of :meth:`lightning.pytorch.LightningModule.configure_optimizers`. Args: configure_optim_output: Th

Returns all valid signals supported on the current platform. Behaves identically to :func:`signals.valid_signals` in Python 3.8+ and implemen