73 projects
TensorFlow
TensorFlow is an open-source machine learning framework developed by Google that enables numerical computation and large-scale machine learning. It provides a flexible system for defining and executing computations involving tensors, which are multi-dimensional arrays. The framework supports deep learning and neural networks across multiple platforms and devices.
47,391
6,206
$203M
Transformers
Transformers is a state-of-the-art Natural Language Processing (NLP) library that provides thousands of pretrained models for text, vision, and audio tasks. It offers APIs to easily download and use these models, as well as to train new ones. The library supports multiple deep learning frameworks including PyTorch, TensorFlow, and JAX.
24,101
3,181
$52M
Keras
Keras is an open-source deep learning framework that provides a Python interface for artificial neural networks. It serves as a high-level API capable of running on top of multiple backend engines including TensorFlow, offering a user-friendly, modular, and extensible way to build and train deep learning models.
18,062
2,636
$8.2M
Ultralytics YOLO
Ultralytics YOLO11 π
12,697
836
$4.1M
llama cpp
A port of Facebook's LLaMA model in C/C++, focusing on efficient inference and deployment of large language models on consumer hardware. The project provides optimized implementations for running LLaMA models with minimal dependencies and memory requirements.
12,293
1,880
$30M
PaddlePaddle
PaddlePaddle is an open-source deep learning platform developed by Baidu that provides a comprehensive suite of tools for AI model development, training, and deployment. It features an easy-to-use API, high performance distributed training capabilities, and extensive support for various deep learning applications including computer vision, natural language processing, and speech recognition.
9,399
488
$106M
YOLOv5
YOLOv5 is a computer vision model and framework for real-time object detection, offering state-of-the-art performance, easy training and deployment capabilities, and extensive documentation. It implements the YOLO (You Only Look Once) architecture with improvements for speed and accuracy.
8,830
713
$802K
Diffusers
π€ Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch and FLAX.
6,877
915
$29M
PyTorch Geometric
Graph Neural Network Library for PyTorch
5,080
726
$5.3M
fairseq
Facebook AI Research Sequence-to-Sequence Toolkit written in Python.
4,651
718
$6.6M
Detectron2
Detectron2 is a computer vision library developed by Facebook AI Research (FAIR) that implements state-of-the-art object detection algorithms. It provides a modular, flexible platform for implementing and training computer vision models, with support for tasks like object detection, instance segmentation, keypoint detection, and panoptic segmentation.
4,304
603
$2.3M
ncnn
ncnn is a high-performance neural network inference framework optimized for the mobile platform
3,521
311
$28M
TRL - Transformer Reinforcement Learning
TRL (Transformer Reinforcement Learning) is a library for training transformer language models using reinforcement learning. It provides implementations of algorithms like PPO (Proximal Policy Optimization) adapted for language models, enabling fine-tuning of models based on reward signals and human feedback.
3,366
512
$2.8M
FlashAttention
FlashAttention is a high-performance implementation of attention for training deep learning models, offering faster and more memory-efficient attention computation compared to standard implementations. It achieves this through a novel algorithm that reduces memory access and increases hardware utilization.
3,225
517
$3.3M
OpenCV Contrib
Open Source Computer Vision Library
3,103
469
$17M
bitsandbytes
A Python library for efficient 8-bit quantization and matrix operations, optimized for machine learning and deep learning applications. It provides low-precision arithmetic and memory-efficient tensor operations.
2,771
416
$1M
Coqui TTS
πΈπ¬ - a deep learning toolkit for Text-to-Speech, battle-tested in research and production
2,623
363
$11M
PEFT
π€ PEFT: State-of-the-art Parameter-Efficient Fine-Tuning.
2,349
413
$9.1M
MMCV
MMCV is a foundational computer vision library that provides a wide range of utilities and operations for deep learning research and development. It includes efficient implementations of common vision operations, CUDA acceleration, and serves as a base dependency for the OpenMMLab computer vision projects.
2,319
261
$2.5M
Megatron-LM
Ongoing research training transformer models at scale
2,299
266
$23M
Nerfstudio
A collaboration friendly studio for NeRFs
2,187
271
$18M
Flair
A very simple framework for state-of-the-art Natural Language Processing (NLP)
2,179
357
$2M
tinygrad
tinygrad is a minimalist deep learning framework designed to be simple, lightweight, and easy to understand. It implements neural network operations and autograd functionality in pure Python, allowing users to build and train neural networks with minimal dependencies.
2,015
355
$15M
PyTorch3D
PyTorch3D is FAIR's library of reusable components for deep learning with 3D data
1,909
303
$3.9M
TensorFlow Probability
TensorFlow Probability is a Python library built on TensorFlow that combines probabilistic models and deep learning. It enables users to encode stochastic computation through probability distributions and probabilistic layers for building deep probabilistic models.
1,882
355
$20M
DeepLabCut
DeepLabCut is an open-source toolbox for markerless pose estimation of animals performing various behaviors using deep learning. It implements a transfer learning approach to adapt deep neural networks pre-trained on human pose estimation to animal pose tracking with minimal training data.
1,842
156
$3.4M
PaddleSeg
PaddleSeg is a deep learning image segmentation toolkit based on PaddlePaddle framework that provides high-performance semantic segmentation, interactive segmentation, and panoptic segmentation models along with training and deployment solutions
1,812
78
$6.9M
GGML
GGML is a tensor library for machine learning that enables efficient neural network inference on CPU. It provides low-level primitives for implementing deep learning models with a focus on performance and memory efficiency, particularly for running large language models on consumer hardware.
1,561
277
$8.3M
PyTorch Image Models (timm)
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
1,540
299
$13M
TT-Metal
TT-Metal is a software framework for programming Tenstorrent AI accelerator chips, providing low-level hardware access and control capabilities for machine learning workloads
1,515
88
$120M
SpeechBrain
SpeechBrain is an open-source speech toolkit built on PyTorch that provides state-of-the-art speech technologies, including speech recognition, speaker recognition, speech enhancement, multi-microphone signal processing and speech separation. It features a unified, flexible interface for speech research and applications.
1,425
181
$9.4M
CUTLASS
CUDA Templates for Linear Algebra Subroutines
1,419
235
$63M
Flax
Flax is a neural network library for JAX that is designed for flexibility.
1,349
315
$4.8M
OpenNMT-py
OpenNMT-py is an open-source neural machine translation framework built in PyTorch. It provides tools and libraries for training and deploying sequence-to-sequence models, particularly focused on neural machine translation tasks. The system supports various model architectures, training configurations, and features for both research and production use cases.
1,308
263
$4M
NVIDIA DALI
A GPU-accelerated library containing highly optimized building blocks and an execution engine for data processing to accelerate deep learning training and inference applications.
1,263
225
$15M
xFormers
Hackable and optimized Transformers building blocks, supporting a composable construction.
1,239
244
$1.9M
MMAction2
MMAction2 is an open-source toolbox for video understanding and action recognition tasks. It provides a comprehensive collection of deep learning models, training frameworks, and evaluation tools for video-based tasks like action recognition, temporal action detection, and skeleton-based action recognition.
1,201
102
$3.9M
FlashInfer
FlashInfer is a high-performance inference engine focused on optimizing attention computation for large language models (LLMs). It provides efficient CUDA kernels and implementations for accelerating attention mechanisms in transformer architectures.
1,007
161
$9M
PyTensor
PyTensor allows you to define, optimize, and efficiently evaluate mathematical expressions involving multi-dimensional arrays.
909
127
$7.7M
Torch Neural Network Package
A neural network package for the Torch scientific computing framework that provides a modular approach to building and training neural networks. It includes implementations of common neural network layers, loss functions, and optimization algorithms.
763
177
$1.5M
H2O
H2O is an Open Source, Distributed, Fast & Scalable Machine Learning Platform: Deep Learning, Gradient Boosting (GBM) & XGBoost, Random Forest, Generalized Linear Modeling (GLM with Elastic Net), K-Means, PCA, Generalized Additive Models (GAM), RuleFit, Support Vector Machine (SVM), Stacked Ensembles, Automatic Machine Learning (AutoML), etc.
759
89
$34M
DyNet
DyNet is a neural network library designed for dynamic neural networks where the network structure can change for every training example. It is written in C++ with bindings for Python and is particularly well-suited for natural language processing tasks and other applications with variable-sized inputs.
665
124
$11M
MMF (Multimodal Framework)
MMF (Multimodal Framework) is an open-source modular framework for vision and language multimodal research. It enables training, evaluation and deployment of various multimodal AI models, supporting tasks like visual question answering, image captioning, and visual reasoning.
659
91
$2.1M
gnina
A deep learning framework for molecular docking
545
77
$4.3M
Ascend Extension for PyTorch
An extension that enables PyTorch to run on Huawei Ascend AI processors/NPUs, providing hardware acceleration and optimization capabilities for deep learning models on Ascend platforms
486
14
$14M
Neural Network (NN) Streamer
? Neural Network (NN) Streamer, Stream Processing Paradigm for Neural Network Apps/Devices.
458
48
$34M
XNNPACK
High-efficiency floating-point neural network inference operators for mobile, server, and Web
427
98
$106M
PyTorch Metric Learning
PyTorch Metric Learning is a library that provides a comprehensive collection of metric learning losses, miners, samplers, and trainers designed to work with PyTorch. It enables users to train deep neural networks using various metric learning techniques to learn similarity between data points.
424
86
$1.2M
FedML
FedML is an open-source library and MLOps platform for federated learning research and production deployment. It provides a unified framework for distributed/federated training, serving, and mobile/IoT deployment with consistent APIs and modular architecture.
401
57
$7.9M
NetKet
Machine learning algorithms for many-body quantum systems
319
44
$2.9M
Thinc
Thinc is a lightweight deep learning library that offers a functional approach to neural network layers, designed to be composable and interoperable with PyTorch, TensorFlow and MXNet. It provides a clean, hackable machine learning framework used extensively in spaCy's natural language processing models.
300
107
$1.7M
einops
Einops is a Python library that provides a powerful and unified way to manipulate tensor dimensions in deep learning frameworks. It offers consistent and readable operations for reshaping, permuting, and splitting tensors across multiple ML frameworks like PyTorch, TensorFlow, and JAX.
299
111
$403K
RWKV
The mission of the Project is to develop a recurrent neural net language model with GPT-level LLM performance, which can also be directly trained like a GPT transformer.
283
43
$71M
Lux
Elegant and Performant Scientific Machine Learning in Julia
235
80
$1.7M
TorchMultimodal
TorchMultimodal is a framework for building and training multimodal AI models using PyTorch, developed by Meta/Facebook Research. It provides modular components and tools for combining different types of data (text, images, audio, etc.) in deep learning applications.
120
19
$1.5M
ShaderNN
The mission of the Project is to develop a lightweight deep learning inference framework optimized for Convolutional Neural Networks.
14
3
$35M
BeyondML
The mission of the Project is to advance the state of artificial intelligence by designing and implementing an open-source framework for developing sparse, optimized neural networks capable of efficiently performing multiple tasks across multiple data domains.
7
4
$6.4M
AO Sparsity
PyTorch native quantization and sparsity for training and inference
DNNLibrary.jl
Relax! Flux is the ML library that doesn't make you tensor
Gloo
Collective communications library with various primitives for multi-machine training.