Evaluates the loss function of the predicted logits w.r.t. the targets :param logits: predicted logits for the first augmentation of

Updates self.queue with the features of the novel points in the current batch :param feats: the features for the novel points in the

Args: voxel_size: side length of a voxel clip_bound: boundary of the voxelizer. Points outside the bound will be deleted

Args: optimizer (Optimizer): Wrapped optimizer. warmup_epochs (int): Maximum number of iterations for linear warmup

Load data from given dataset directory.

Load data from given dataset directory.

Load data from given dataset directory.

Load data from given dataset directory.

Called when epoch is passed as a param to the `step` function of the scheduler.

Compute learning rate using chainable form of the scheduler.

returns a boolean mask selecting the p-th percentile of the predictions with highest confidence for each class :param preds:

Creates the set of unknown labels in a way that all the classes in the same splits have roughly the same number of points