Accumulate a confusion matrix for a classification task. ignore_index only supports index <0, or > num_classes

Computes and stores the average and current value

Static graph convolution layer

Computes and stores the average and current value

Fast BachNorm1d for input with shape [B, N, C], where the feature dimension is at last. Borrowed from torch-points3d: https://github.com/torch-po

Find the neighbors' indices based on dilated knn

Dynamic graph convolution layer

The Feature Propogation module in PointNet++

LARS for PyTorch Paper: `Large batch training of Convolutional Networks` - https://arxiv.org/pdf/1708.03888.pdf Args: params (i

Local aggregation layer for a set Set abstraction layer abstracts features from a larger set to a smaller set Local aggregation layer aggrega

A registry to map strings to classes. Args: name (str): Registry name.

f''' Chamder Distance L1

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

Subclass torch's LayerNorm to handle fp16.

Point cloud to subsampled groups

r"""Implements AdaBelief algorithm. Modified from Adam in PyTorch Arguments: params (iterable): iterable of parameters to optimize or dic

Get the distances and indices to a fixed number of neighbors Reference: https://gist.github.com/ModarTensai/60fe0d0e3536adc28778448419908f47

Implements a pure pytorch variant of FuseLAMB (NvLamb variant) optimizer from apex.optimizers.FusedLAMB reference: https://github.com/NVIDIA/DeepL

MADGRAD_: A Momentumized, Adaptive, Dual Averaged Gradient Method for Stochastic Optimization. .. _MADGRAD: https://arxiv.org/abs/2101.1

r"""Feature Propagation module in PointNet++. Propagates the features of one set to another

Point Vision Transformer ++: with early convolutions

Residual Dynamic graph convolution block

Transformer Encoder without hierarchical structure

Implements Adafactor algorithm. This implementation is based on: `Adafactor: Adaptive Learning Rates with Sublinear Memory Cost` (see https://

Implements the AdaHessian algorithm from "ADAHESSIAN: An Adaptive Second OrderOptimizer for Machine Learning" Arguments: params (ite

Used in attention U-Net.

This is a template for defining your customized 3D backbone and use it for pre-training in ULIP framework. The expected input is Batch_size x

Cosine decay with restarts. This is described in the paper https://arxiv.org/abs/1608.03983. Inspiration from https://github.com/all