Stars
🐫 CAMEL: The first and the best multi-agent framework. Finding the Scaling Law of Agents. https://www.camel-ai.org
A fast, effective data attribution method for neural networks in PyTorch
A lightweight logger for machine learning teams to log images and predictions in production.
I will build Transformer from scratch
Easily turn large sets of image urls to an image dataset. Can download, resize and package 100M urls in 20h on one machine.
A software for automating materials science computations
Distilling Model Failures as Directions in Latent Space
Implementation of Vision Transformer, a simple way to achieve SOTA in vision classification with only a single transformer encoder, in Pytorch
FFCV: Fast Forward Computer Vision (and other ML workloads!)
User-friendly secure computation engine based on secure multi-party computation
Pretrain, finetune ANY AI model of ANY size on multiple GPUs, TPUs with zero code changes.
A framework for analyzing computer vision models with simulated data
Minimal implementation of Denoised Smoothing (https://arxiv.org/abs/2003.01908) in TensorFlow.
Python library for argument and configuration management
Code relative to "Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks"
RobustBench: a standardized adversarial robustness benchmark [NeurIPS 2021 Benchmarks and Datasets Track]
[CVPR 2021] VirTex: Learning Visual Representations from Textual Annotations
End-to-End Object Detection with Transformers
Code for our ICLR Trustworthy ML 2020 workshop paper "Improved Image Wasserstein Attacks and Defenses"
Implementations of several fast approximate algorithms for geometric optimal transport (OT)
Randomized Smoothing of All Shapes and Sizes (ICML 2020).
Statistical adaptive stochastic optimization methods
DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.
Python sample codes and textbook for robotics algorithms.
[ICLR 2020] A repository for extremely fast adversarial training using FGSM
Code for NeurIPS 2019 paper: "Tensor Programs I: Wide Feedforward or Recurrent Neural Networks of Any Architecture are Gaussian Processes"
Tools for accelerating safe exploration research.
Semisupervised learning for adversarial robustness https://arxiv.org/pdf/1905.13736.pdf