Mingjie Sun

llm-idiosyncrasies. Code release for "Idiosyncrasies in Large Language Models"

gradient-checkpointing. Make huge neural nets fit in memory

lingua. Meta Lingua: a lean, efficient, and easy-to-hack codebase to research LLMs.

nGPT-pytorch. Quick implementation of nGPT, learning entirely on the hypersphere, from NvidiaAI

unsloth. Unsloth Studio is a web UI for training and running open models like Gemma 4, Qwen3.6, DeepSeek, gpt-oss locally.

llm-past-tense. Does Refusal Training in LLMs Generalize to the Past Tense? [ICLR 2025]

admm-pruning. Python

bias.

massive-activations. Code accompanying the paper "Massive Activations in Large Language Models"

diffusion-classifier. Diffusion Classifier leverages pretrained diffusion models to perform zero-shot classification without additional training

echo-embeddings. Python

open-unlearning. [NeurIPS D&B '25] The one-stop repository for LLM unlearning

ml-aim. This repository provides the code and model checkpoints for AIMv1 and AIMv2 research projects.

weight-selection. Python

dinov2. PyTorch code and models for the DINOv2 self-supervised learning method.

GPU-Puzzles. Solve puzzles. Learn CUDA.

OWL. Official Pytorch Implementation of "Outlier Weighed Layerwise Sparsity (OWL): A Missing Secret Sauce for Pruning LLMs to High Sparsity"

nanoGPT. The simplest, fastest repository for training/finetuning medium-sized GPTs.

EasyLM. Large language models (LLMs) made easy, EasyLM is a one stop solution for pre-training, finetuning, evaluating and serving LLMs in JAX/Flax.

wanda. A simple and effective LLM pruning approach.

Faithful-COT. Code and data accompanying our paper on arXiv "Faithful Chain-of-Thought Reasoning".

the-pile. Python

chain-of-thought-hub. Benchmarking large language models' complex reasoning ability with chain-of-thought prompting

tree-of-thought-llm. [NeurIPS 2023] Tree of Thoughts: Deliberate Problem Solving with Large Language Models

tta_conjugate. Test-Time Adaptation via Conjugate Pseudo-Labels

tuned-lens. Tools for understanding how transformer predictions are built layer-by-layer

RedPajama-Data. The RedPajama-Data repository contains code for preparing large datasets for training large language models.

prize. A prize for finding tasks that cause large language models to show inverse scaling

DeblurSR. DeblurSR: Event-Based Motion Deblurring Under the Spiking Representation (AAAI 2024)

DeepSpeed. DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.

alpaca-lora. Instruct-tune LLaMA on consumer hardware

AlpacaDataCleaned. Alpaca dataset from Stanford, cleaned and curated

FlexLLMGen. Running large language models on a single GPU for throughput-oriented scenarios.

stanford_alpaca. Code and documentation to train Stanford's Alpaca models, and generate the data.

FastChat. An open platform for training, serving, and evaluating large language models. Release repo for Vicuna and Chatbot Arena.

open_flamingo. An open-source framework for training large multimodal models.

GPTQ-for-LLaMa. 4 bits quantization of LLaMA using GPTQ

gptq. Code for the ICLR 2023 paper "GPTQ: Accurate Post-training Quantization of Generative Pretrained Transformers".

functorch. functorch is JAX-like composable function transforms for PyTorch.

dropout. Code release for "Dropout Reduces Underfitting"

vision-transformers-cifar10. Let's train vision transformers (ViT) for cifar 10 / cifar 100!

smoothinv. Single Image Backdoor Inversion via Robust Smoothed Classifiers

openai-cookbook. Examples and guides for using the OpenAI API

temperature_scaling. A simple way to calibrate your neural network.

EVA. EVA Series: Visual Representation Fantasies from BAAI

edm. Elucidating the Design Space of Diffusion-Based Generative Models (EDM)

ImageNetV2_pytorch. ImageNetV2 Pytorch Dataset

clip-italian. CLIP (Contrastive Language–Image Pre-training) for Italian

open_clip. An open source implementation of CLIP.

img2dataset. Easily turn large sets of image urls to an image dataset. Can download, resize and package 100M urls in 20h on one machine.

diffusion_denoised_smoothing. Certified robustness "for free" using off-the-shelf diffusion models and classifiers

breaking-poisoned-classifier. Code for paper "Poisoned classifiers are not only backdoored, they are fundamentally broken"

CFA. Python

arxiv-utils. Meaningful titles for tabs and PDF downloads! Also supports tab search.

guided-diffusion. Python

Training-Code-of-STM. Training code of Spatial Time Memory Network. Semi-supervised video object segmentation.

pytracking. Visual tracking library based on PyTorch.

dino. PyTorch code for Vision Transformers training with the Self-Supervised learning method DINO

3DCommonCorruptions. 3D Common Corruptions and Data Augmentation, CVPR 2022 [Oral]

PyContrast. PyTorch implementation of Contrastive Learning methods

robustness. Robustness and adaptation of ImageNet scale models. Pre-Release, stay tuned for updates.

awesome-semantic-segmentation-pytorch. Semantic Segmentation on PyTorch (include FCN, PSPNet, Deeplabv3, Deeplabv3+, DANet, DenseASPP, BiSeNet, EncNet, DUNet, ICNet, ENet, OCNet, CCNet, PSANet, CGNet, ESPNet, LEDNet, DFANet)

anytime. Anytime Dense Prediction with Confidence Adaptivity (ICLR 2022)

svm-pytorch. Linear SVM with PyTorch

augmix. AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty

ConvNeXt. Code release for ConvNeXt model

huginn. Create agents that monitor and act on your behalf. Your agents are standing by!

ffcv. FFCV: Fast Forward Computer Vision (and other ML workloads!)

reinforcement-learning. Minimal and Clean Reinforcement Learning Examples

learn2learn. A PyTorch Library for Meta-learning Research

bolei_awesome_posters. CVPR and NeurIPS poster examples and templates

backdoors101. Backdoors Framework for Deep Learning and Federated Learning. A light-weight tool to conduct your research on backdoors.

vision_transformer. Jupyter Notebook

pytorch-image-models. The largest collection of PyTorch image encoders / backbones. Including train, eval, inference, export scripts, and pretrained weights -- ResNet, ResNeXT, EfficientNet, NFNet, Vision Transformer (ViT), MobileNetV4, MobileNet-V3 & V2, RegNet, DPN, CSPNet, Swin Transformer, MaxViT, CoAtNet, ConvNeXt, and more

mlp-mixer-pytorch. An All-MLP solution for Vision, from Google AI

convmixer. Implementation of ConvMixer for "Patches Are All You Need? 🤷"

PBT_MARL_watered_down. My attempt to reproduce a water down version of PBT (Population based training) for MARL (Multi-agent reinforcement learning) using DDPPO (Decentralized & distributed proximal policy optimization) from ray[rllib].

PPO-for-Beginners. A simple and well styled PPO implementation. Based on my Medium series: https://medium.com/@eyyu/coding-ppo-from-scratch-with-pytorch-part-1-4-613dfc1b14c8.

Highlight-Indentation-for-Emacs. Minor modes to highlight indentation guides in emacs.

rllab-adv. Code to train RL agents along with Adversarial distrubance agents

pybullet-gym. Open-source implementations of OpenAI Gym MuJoCo environments for use with the OpenAI Gym Reinforcement Learning Research Platform.

brax. Massively parallel rigidbody physics simulation on accelerator hardware.

DeepRL. Modularized Implementation of Deep RL Algorithms in PyTorch

pytorch-a2c-ppo-acktr-gail. PyTorch implementation of Advantage Actor Critic (A2C), Proximal Policy Optimization (PPO), Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation (ACKTR) and Generative Adversarial Imitation Learning (GAIL).

Deep-Reinforcement-Learning-Algorithms-with-PyTorch. PyTorch implementations of deep reinforcement learning algorithms and environments

Deep-reinforcement-learning-with-pytorch. PyTorch implementation of DQN, AC, ACER, A2C, A3C, PG, DDPG, TRPO, PPO, SAC, TD3 and ....

StateAdvDRL. [NeurIPS 2020, Spotlight] Code for "Robust Deep Reinforcement Learning against Adversarial Perturbations on Observations"

instagram-filters. Instagram-like image filters

train-procgen. Code for the paper "Leveraging Procedural Generation to Benchmark Reinforcement Learning"

rad_procgen. RAD: Reinforcement Learning with Augmented Data (code for procgen experiments)

rad. RAD: Reinforcement Learning with Augmented Data

textgenrnn. Easily train your own text-generating neural network of any size and complexity on any text dataset with a few lines of code.

awesome-macos-command-line-zh. 用你的 macOS 终端搞事情。(Use your macOS terminal shell to do awesome things. )

robustbench. RobustBench: a standardized adversarial robustness benchmark [NeurIPS 2021 Benchmarks and Datasets Track]

vit-pytorch. Implementation of Vision Transformer, a simple way to achieve SOTA in vision classification with only a single transformer encoder, in Pytorch

imaginaire. NVIDIA's Deep Imagination Team's PyTorch Library

lucid. A collection of infrastructure and tools for research in neural network interpretability.

gdown.pl. Google Drive direct download of big files

auto-attack. Code relative to "Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks"

rs4a. Randomized Smoothing of All Shapes and Sizes (ICML 2020).

denoised-smoothing. Provably defending pretrained classifiers including the Azure, Google, AWS, and Clarifai APIs

celluloid. :movie_camera: Matplotlib animations made easy

robustness_applications. Notebooks for reproducing the paper "Computer Vision with a Single (Robust) Classifier"

blackbox-bandits. Code for "Prior Convictions: Black-Box Adversarial Attacks with Bandits and Priors"

fabric. Simple, Pythonic remote execution and deployment.

sampler. Tool for shell commands execution, visualization and alerting. Configured with a simple YAML file.

MultiRobustness. Code for the paper "Adversarial Training and Robustness for Multiple Perturbations", NeurIPS 2019

RIDNet. Pytorch code for "Real image denoising with feature attention", ICCV (Oral), 2019.

wtfpython. What the f*ck Python? 😱

transformers. 🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.

PySnooper. Never use print for debugging again

ntfy. 🖥️📱🔔 A utility for sending notifications, on demand and when commands finish.

Git-Beautify-For-MacOS-Terminal. An easy-to-use set of config files to beautify Git in your MacOS or OS X terminal. If you find it hard to parse a jumble of mono-colored type on your command line, this set-up can help you tame the ugly bash beast.

limited-blackbox-attacks. Code for "Black-box Adversarial Attacks with Limited Queries and Information" (http://arxiv.org/abs/1804.08598)

cloudy_vision. Cloudy Vision is an open source tool to test the image labeling capabilities of different computer vision API vendors.

captum. Model interpretability and understanding for PyTorch

FreeAdversarialTraining. PyTorch Implementation of Adversarial Training for Free!

RNAN. PyTorch code for our ICLR 2019 paper "Residual Non-local Attention Networks for Image Restoration"

vim-as-an-ide. Workshop on how to use Vim Plugins.

mixup-cifar10. mixup: Beyond Empirical Risk Minimization

FFDNet. FFDNet: Toward a Fast and Flexible Solution for CNN based Image Denoising (TIP, 2018)

reproducible-image-denoising-state-of-the-art. Collection of popular and reproducible image denoising works.

Pytorch-HarDNet. 35% faster than ResNet: Harmonic DenseNet, A low memory traffic network

BinaryNet.pytorch. Binarized Neural Network (BNN) for pytorch

pytorch-gradual-warmup-lr. Gradually-Warmup Learning Rate Scheduler for PyTorch

interval-bound-propagation. This repository contains a simple implementation of Interval Bound Propagation (IBP) using TensorFlow: https://arxiv.org/abs/1810.12715

EfficientNet-PyTorch. A PyTorch implementation of EfficientNet

python-is-cool. Cool Python features for machine learning that I used to be too afraid to use. Will be updated as I have more time / learn more.

pycls. Codebase for Image Classification Research, written in PyTorch.

robustness. A library for experimenting with, training and evaluating neural networks, with a focus on adversarial robustness.

smoothing-adversarial. Code for our NeurIPS 2019 *spotlight* "Provably Robust Deep Learning via Adversarially Trained Smoothed Classifiers"

pykalman. Kalman Filter, Smoother, and EM Algorithm for Python

provable-robustness-max-linear-regions. Provable Robustness of ReLU networks via Maximization of Linear Regions [AISTATS 2019]

relu_stable. Python

detectron2. Detectron2 is a platform for object detection, segmentation and other visual recognition tasks.

XNOR-Net. ImageNet classification using binary Convolutional Neural Networks

quantized.pytorch. Python

advertorch. A Toolbox for Adversarial Robustness Research

smsop. Code for Statistically-motivated Second-order Pooling, ECCV2018

TRADES. TRADES (TRadeoff-inspired Adversarial DEfense via Surrogate-loss minimization)

JigenDG. Repository for the CVPR19 oral paper "Domain Generalization by Solving Jigsaw Puzzles"

Sobel-Feldman. Sobel–Feldman, Prewitt, Canny filter

cnnimageretrieval. CNN Image Retrieval in MatConvNet: Training and evaluating CNNs for Image Retrieval in MatConvNet

PAR. Implementation of Patch-wise Adversarial Regularization from "Learning Robust Global Representations by Penalizing Local Predictive Power"

Meta-Learning-Domain-Generalization. Python

MLDG. The demo code for the MLDG paper "Learning to Generalize: Meta-Learning for Domain Generalization", https://arxiv.org/abs/1710.03463, https://arxiv.org/pdf/1710.03463.pdf

HEX_experiments. For replication of the experiments in the paper Learning Robust Representations by Projecting Superficial Statistics Out

pytorch_fft. PyTorch wrapper for FFTs

setGPU. Small Python library to automatically set CUDA_VISIBLE_DEVICES to the least loaded device on multi-GPU systems.

robust-features-code. Code for "Robustness May Be at Odds with Accuracy"

pix2pixHD. Synthesizing and manipulating 2048x1024 images with conditional GANs

smoothing. Provable adversarial robustness at ImageNet scale

CannyEdgePytorch. Python

bsuite. bsuite is a collection of carefully-designed experiments that investigate core capabilities of a reinforcement learning (RL) agent

rlpyt. Reinforcement Learning in PyTorch

ImageNet_Utils. :arrow_double_down: Utils to help download images by id, crop bounding box, label images, etc.

gpt-2. Code for the paper "Language Models are Unsupervised Multitask Learners"

ai-deadlines. :alarm_clock: AI conference deadline countdowns

mnist-svhn-transfer. PyTorch Implementation of CycleGAN and SSGAN for Domain Transfer (Minimal)

constructed-datasets. Datasets for the paper "Adversarial Examples are not Bugs, They Are Features"

PyTorch-GAN. PyTorch implementations of Generative Adversarial Networks.

votenet. Deep Hough Voting for 3D Object Detection in Point Clouds

node_embedding_attack. Adversarial Attacks on Node Embeddings via Graph Poisoning

my-git. A practical Git & GitHub handbook for modern engineering teams and AI-native development workflows. | 现代工程团队与 AI 编程时代的 Git / GitHub 实战手册

DSGAN. Implementation of Diversity-Sensitive Conditional Generative Adversarial Networks (ICLR 2019)

MUNIT. Multimodal Unsupervised Image-to-Image Translation

pytorch-lightning. Pretrain, finetune ANY AI model of ANY size on 1 or 10,000+ GPUs with zero code changes.

coding-interview-university. A complete computer science study plan to become a software engineer.

Awesome-MXNet. A curated list of MXNet examples, tutorials and blogs.

arxiv-latex-cleaner. arXiv LaTeX Cleaner: Easily clean the LaTeX code of your paper to submit to arXiv

shisi.eth-in-web3. 十四君公众号文章的相关代码资源开源

Anti-Anti-Spider. 越来越多的网站具有反爬虫特性，有的用图片隐藏关键数据，有的使用反人类的验证码，建立反反爬虫的代码仓库，通过与不同特性的网站做斗争（无恶意）提高技术。（欢迎提交难以采集的网站）（因工作原因，项目暂停）

stylize-datasets. A script that applies the AdaIN style transfer method to arbitrary datasets

ResNet. Reproduce ResNet-v2(Identity Mappings in Deep Residual Networks) with MXNet

Stylized-ImageNet. Code to create Stylized-ImageNet, a stylized version of standard ImageNet (ICLR 2019 Oral)

awesome-consensus. Awesome list for Paxos and friends

paxos. Plain Paxos Implementations in Python & Java

thuthesis. LaTeX Thesis Template for Tsinghua University

pytorch_GAN_zoo. A mix of GAN implementations including progressive growing

semseg. Semantic Segmentation in Pytorch

minimalRL. Implementations of basic RL algorithms with minimal lines of codes! (pytorch based)

pytorch-cnn-visualizations. Pytorch implementation of convolutional neural network visualization techniques

mmf. A modular framework for vision & language multimodal research from Facebook AI Research (FAIR)

tensorflow-GANs. 各类GAN综合在一起，借鉴了hwalsuklee大神的

fashion-mnist. A MNIST-like fashion product database. Benchmark :point_down:

MeshCNN. Convolutional Neural Network for 3D meshes in PyTorch

pix2pix. Image-to-image translation with conditional adversarial nets

BicycleGAN. Toward Multimodal Image-to-Image Translation

fast-MPN-COV. @CVPR2018: Efficient unrolling iterative matrix square-root normalized ConvNets, implemented by PyTorch (and code of B-CNN,Compact bilinear pooling etc.) for training from scratch & finetuning.

macos-terminal-themes. Color schemes for default macOS Terminal.app

thesis. My PhD thesis, titled "Reasonably Programmable Syntax"

pytorch-adversarial_box. PyTorch library for adversarial attack and training

the-coding-interview. Programming exercises, code katas and puzzles for your job interview training - or just for fun.

mxnet_mtcnn_face_detection. MTCNN face detection

programmer-job-blacklist. :see_no_evil:程序员找工作黑名单，换工作和当技术合伙人需谨慎啊 更新有赞

996.ICU. Repo for counting stars and contributing. Press F to pay respect to glorious developers.

955.WLB. 955 不加班的公司名单 - 工作 955，work–life balance (工作与生活的平衡)

PRNet. Joint 3D Face Reconstruction and Dense Alignment with Position Map Regression Network (ECCV 2018)

GENet. C++

senet.pytorch. PyTorch implementation of SENet