Attentive Pooler

Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).

Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).

Layer to pad and convolve input

Add sinusoidal position embeddings to the patch embeddings.

computes and stores the average and current value

Attentive Pooler

Transformer encoder consisting of `config.num_hidden_layers` self attention layers. Each layer is a [`CLIPEncoderLayer`]. Args:

Layer to perform a convolution followed by ELU

Track a series of values and provide access to smoothed values over a window or the global series average.

Layer to transform a depth image into a point cloud

Kozachenko-Leonenko entropic loss regularizer from Sablayrolles et al. - 2018 - Spreading vectors for similarity search

2D video to patch embedding: (B, T, C, H, W) -> (B, N, D)

Layer which projects 3D points into a camera with intrinsics K and at position T

Constructs a resnet model with varying number of input images. Adapted from https://github.com/pytorch/vision/blob/master/torchvision/models/resne

Pytorch module for a resnet encoder

Layer to compute the SSIM loss between a pair of images

Wraps TactileVideoMAE with a fixed sensor type so downstream tasks can call model_encoder(x) without providing sensor_type explicitly.

Attentive Classifier