nninit

Weight initializers

Provides weight initialization schemes for PyTorch neural networks

Weight initialization schemes for PyTorch nn.Modules

GitHub

70 stars
4 watching
8 forks
Language: Python
last commit: over 9 years ago
pytorchweight-initialization-schemes

Related projects:

RepositoryDescriptionStars
kaixhin/nninitProvides parameter initialisation schemes for neural network modules in Torch7100
fgxaos/pytorch-innvestigatePyTorch implementation of an explainability technique for deep neural networks9
akanimax/pro_gan_pytorchImplementation of a deep learning model for generating high-quality images with improved stability and variation.538
philkr/magic_initThis code provides an initialization method for convolutional neural networks based on data-dependent parameters.138
chenxi116/pnasnet.pytorchPyTorch implementation of PNASNet-5 architecture317
ericsun99/shufflenet-v2-pytorchA PyTorch implementation of a ShuffleNet-v2 neural network architecture for image classification.431
jhjacobsen/pytorch-i-revnetDeep invertible neural network implementation using PyTorch for image recognition and reconstruction tasks.390
hassony2/kinetics_i3d_pytorchTransfers pre-trained I3D network weights from TensorFlow to PyTorch532
graal-research/poutyneA PyTorch framework simplifying neural network training with automated boilerplate code and callback utilities572
harshtrivedi/nmn-pytorchA PyTorch implementation of Neural Module Networks for Visual Question Answering108
wlwkgus/noisynaturalgradientPyTorch implementation of Noisy Natural Gradient as Variational Inference for Bayesian Neural Networks121
blue-season/pywarmA lightweight neural network construction API for PyTorch.184
randl/shufflenetv2-pytorchAn implementation of a lightweight convolutional neural network architecture for mobile devices191
tobiascz/mnist_pytorch_python_and_capiAn example project demonstrating how to train and deploy a neural network in Python and C++ using PyTorch 1.096
michaelklachko/pnn.pytorchA PyTorch implementation of a neural network architecture that modifies the input to its layers by applying noise masks.57