aux_loss: True if auxiliary intermediate losses (losses at each encoder/decoder layer)

Args: thres (list[str] or None): threshold for outlier only_valid (bool): whether invalid pixels are excluded from ev

Init. Args: features (int): number of features

Init. Args: features (int): number of features

Alternate self and cross attention with gradient checkpointing to save memory :param feat: image feature concatenated from left and

Run forward function and calculate loss for auxiliary head in training.

Args: cls_pred (Tensor): The prediction with shape (num_query, 1, *) or (num_query, *). gt_labels (Te

Find target locations using ground truth disparity. Find ground truth response at those locations using attention weight. :p

Compute low res disparity using the attention weight by finding the most attended pixel and regress within the 3px window :param pos

Compute low res occlusion by using inverse of the matched values :param matched_attn: updated attention weight without dustbins, [N,

Compute relative difference between each pixel location from left image to right image, to be used to calculate disparity :p

Configure a logger. Adapted from Detectron2. Args: name: The name of the logger to configure. level: The logging level

Run forward function and calculate loss for decode head in inference.

Run forward function and calculate loss for decode head in inference.

Run forward function and calculate loss for decode head in inference.

Run forward function and calculate loss for decode head in training.

Run forward function and calculate loss for decode head in training.

Run forward function and calculate loss for decode head in training.

disable Batchnorm tracking stats to reduce dependency on dataset (this acts as InstanceNorm with affine when batch size is 1)

Expand onehot labels to match the size of prediction.

Returns: str: module name of the caller tuple: a hashable key to be used to identify different callers

Forward function for feature maps before classifying each pixel with ``self.cls_seg`` fc. Args: inputs (list[Tensor]): Lis

Forward function for feature maps before classifying each pixel with ``self.cls_seg`` fc. Args: inputs (list[Tensor]): Li

Forward function for feature maps before classifying each pixel with ``self.cls_seg`` fc. Args: inputs (list[Tensor]): Lis

Estimate disparity between pair of frames

Generate the indices of a random permutation.

Generate a mask which is upper triangular :param sz: square matrix size :return: diagonal binary mask [sz,sz]

Return a process group based on gloo backend, containing all the ranks The result in cached.

Initialize ``auxiliary_head``

Initialize ``decode_head``

Initialize ``decode_head``

Initialize slurm distributed training environment. If argument ``port`` is not specified, them the master port will be system environment vari