deep learning paper implementations / tutorials
A collection of 60+ annotated PyTorch implementations and tutorials of deep learning papers, including transformers, optimizers, GANs, and reinforcement learning, with side-by-side explanatory notes.
Same problem, same approach. Swapping one for another is a config change, not a rewrite.
choose it for a similarly annotated, actively maintained Jupyter-notebook collection of deep learning model implementations.
Repository of Jupyter notebook tutorials for teaching the Deep Learning Course at the University of Amsterdam (MSc AI), Fall 2023
Solves the same problem with a different architecture or at a different layer. Expect to rewrite the integration.
choose it for a structured fastai-based book/course rather than paper-by-paper implementations.
The fastai book, published as Jupyter Notebooks
choose it for a broad collection of ML and DL examples across many frameworks rather than paper-specific notes.
A collection of machine learning examples and tutorials.
choose it for a university-course lab structure with hands-on assignments instead of paper implementations.
Lab Materials for MIT 6.S191: Introduction to Deep Learning
choose it to stay in Keras rather than PyTorch for GAN implementations.
Keras implementations of Generative Adversarial Networks.
choose it for a friendly Chinese-language beginner PyTorch tutorial instead of paper implementations.
Build your neural network easy and fast, 莫烦Python中文教学
choose it when you need a maintained GAN training library with configs rather than annotated educational code.
StudioGAN is a Pytorch library providing implementations of representative Generative Adversarial Networks (GANs) for conditional/unconditional image generation.
choose it for a video/notes style set of deep learning experiments rather than paper-specific implementations.
Videos, notes and experiments to understand deep learning
choose it for a Chinese-language deep learning textbook/course based on Hung-yi Lee's lectures.
《李宏毅深度学习教程》（李宏毅老师推荐👍，苹果书🍎），PDF下载地址：https://github.com/datawhalechina/leedl-tutorial/releases
choose it for basic TensorFlow tutorials instead of PyTorch paper implementations.
Tensorflow tutorial from basic to hard, 莫烦Python 中文AI教学
choose it for NumPy from-scratch implementations of classic ML algorithms, not paper-specific code.
Machine Learning From Scratch. Bare bones NumPy implementations of machine learning models and algorithms with a focus on accessibility. Aims to cover everything from linear regression to deep learning.
choose it for NumPy implementations of a broad ML toolkit, more classic ML than paper notes.
Machine learning, in numpy
choose it for ready-to-run TensorFlow tutorials aimed at beginners.
:satellite: Simple and ready-to-use tutorials for TensorFlow
choose it if you are working in PaddlePaddle and need pretrained vision transformer models rather than PyTorch notes.
:robot: PaddleViT: State-of-the-art Visual Transformer and MLP Models for PaddlePaddle 2.0+
choose it for a large set of TensorFlow examples for beginners rather than paper-specific implementations.
TensorFlow Tutorial and Examples for Beginners (support TF v1 & v2)
An earlier or more limited way to do the job. Still the right call when you need something small, proven, or CPU-only.
choose it only for a focused, older survey of self-attention mechanisms for computer vision.
Implementation of various self-attention mechanisms focused on computer vision. Ongoing repository.
choose it for a simple, self-contained implementation of only the original transformer.
Attention is all you need implementation
choose it for a dedicated and broader set of PyTorch GAN implementations than the seed's GAN section.
PyTorch implementations of Generative Adversarial Networks.
choose it for the canonical official implementation of CycleGAN/pix2pix if that specific task is your focus.
Image-to-Image Translation in PyTorch
choose it for older, simple Chinese-language NLP model implementations, narrower than the seed.
Simple implementations of NLP models. Tutorials are written in Chinese on my website https://mofanpy.com
choose it as a companion to the GANs in Action book, with narrower GAN-only notebooks.
Companion repository to GANs in Action: Deep learning with Generative Adversarial Networks
choose it for a smaller, older collection of generative model implementations in PyTorch/TensorFlow.
Collection of generative models, e.g. GAN, VAE in Pytorch and Tensorflow.
choose it for a single-paper implementation of ALAE with code for that specific model.
[CVPR2020] Adversarial Latent Autoencoders
choose it for a dedicated implementation of Attention Is All You Need if you only need that architecture.
Transformer: PyTorch Implementation of "Attention Is All You Need"
These projects were analysed and named annotated_deep_learning_paper_implementations among their alternatives. The relationship is not symmetric — how annotated_deep_learning_paper_implementations rates them is a separate judgement, made when annotated_deep_learning_paper_implementations is analysed in its own right.
calls annotated_deep_learning_paper_implementations “Maintained PyTorch implementations of deep learning research papers with notes; choose for modern DL code instead of classic NumPy algorithms.”
Machine learning, in numpy
calls annotated_deep_learning_paper_implementations “choose for in-depth annotated implementations of recent papers to learn advanced deep learning by reading code.”
Tutorials, assignments, and competitions for MIT Deep Learning related courses.
calls annotated_deep_learning_paper_implementations “When you want to study and implement state-of-the-art deep learning papers with detailed annotations, more advanced than the seed's basics.”
A collection of machine learning examples and tutorials.