This repository provides PyTorch implementations of 1D, 2D, and 3D convolutional layers using Kolmogorov-Arnold network theory, including variants such as KAN, Fast KAN, ChebyKAN, WavKAN, JacobiKAN, and BernsteinKAN. It offers complete model architectures like ResNet-like, DenseNet-like, VGG-like, and U-Net variants, along with training scripts using the accelerate library and pretrained weights for ImageNet1k. Benchmarks are provided on MNIST, CIFAR-10, CIFAR-100, TinyImageNet, and ImageNet1k datasets.