A Flexible and Energy Efficient Accelerator For Sparse Convolution Neural Network
Implementation of convolution layer in different flavors
Convolutions and more as einsum for PyTorch
Accelerating convolution using numba, cupy and xnor in python
Source code of "BP-Im2col: An implicit Im2col framework for efficient CNN backpropagation", accepted at ICCD 2022 and "CRSPU: Exploit Commonality of Regular Sparsity to Support Various Convolutions on Systolic Arrays", accepted at DATE 2023.
Implement convolution in SystemVerilog using Image-to-Column and Winograd transformations.
Fast bare-bones implementation of convolutional layers, residual blocks, Adam optimizer, backpropagation and custom accuracy and loss functions (F1 score per pixel and binary-crossentropy)
Implementation of the im2col function for 2D images (e.g. gray-scale).
Neural network from scratch using cupy with experiments on a ton of things, including my own optimizer and activation function
An independent tensor library for deep learning. Features reverse-mode automatic differentiation, topological sorting, and im2col optimized Conv2d layers, built entirely in NumPy.
Conv Forge · 卷积神经网络锻造炉 — 单文件零依赖，手写 2D 卷积(im2col+GEMM)/最大池化/反向传播，数值梯度检验对拍到机器精度，端到端训练真实收敛，10 项不变量全绿。作者：晨星 (CJX0712)
파이토치(PyTorch) 마스터하기: 기초 텐서 조작, Autograd, CNN, RNN 및 하드웨어(im2col) 가속 원리 학습 코드
10,000-image LeNet-5 forward pass in ~28 ms on a single A40 via fused convolution and Tensor Cores (TF32).
Backpropagation derived by hand and gradient-checked, then a convolutional network trained end to end in NumPy. PyTorch is used only as an oracle, never to do the learning.
CNN MNIST / CIFAR10 Classifier - CNN Implemented From Scratch in CuPy/NumPy
Pure-NumPy causal attention and 2D convolution, written to be read: full vs. blocked kernels, im2col/einsum/FFT dispatch, benchmark harnesses.
An implementation of Dense layer and CNN layer using only numpy
To associate your repository with the im2col topic, visit your repo's landing page and select "manage topics."