Skip to content

Latest commit

 

History

History
175 lines (116 loc) · 7.85 KB

DLFramework.md

File metadata and controls

175 lines (116 loc) · 7.85 KB

面向PyTorch/Tensorflow的神经网络图和训练度量的轻量库 https://github.com/waleedka/hiddenlayer pytorch结合友好，画网络图方便直观，可用于notebook实践演示

机器学习DevOps(部署/监控/版本管理与扩展)相关资源大列表 https://github.com/EthicalML/awesome-machine-learning-operations

用于查找数据集中标签错误、用含噪标签进行学习的机器学习包 https://github.com/cgnorthcutt/cleanlab

ONNX模型转换工具，目前已支持Keras, CoreML, LightGBM, Scikit-Learn https://github.com/onnx/onnxmltools

OpenCV的OpenVINO训练扩展，基于PyTorch的神经网络压缩(模型压缩)框架

https://github.com/opencv/openvino_training_extensions/tree/develop/pytorch_toolkit/nncf

【Cottonwood 神经网络框架(教学用)】’The Cottonwood Neural Network Framework - A flexible neural network framework for running experiments and trying ideas.'

https://github.com/brohrer/cottonwood

【Thinc：支持PyTorch、TensorFlow、 MXNet等各主流框架的轻量深度学习库，提供优雅的、类型检查的函数式编程 API】 https://thinc.ai/

FAIR Self-Supervised Learning Integrated Multi-modal Environment (SSLIME) https://github.com/facebookresearch/fair-sslime

【基于Scikit-Learn的Python高性能拓扑机器学习工具箱】 https://github.com/giotto-ai/giotto-tda

【DeepSpeed：微软的深度学习优化库，让分布式训练更简单、更高效、更有效】 https://github.com/microsoft/DeepSpeed

Catalyst：聚焦快速开发与复现的深度学习/强化学习高级工具集 https://github.com/catalyst-team/catalyst

【thor：C++深度学习工具库】’thor - thor: C++ helper library, for deep learning purpose' https://github.com/jinfagang/Thor

【Radish：C++深度学习模型训练/部署框架】'Radish - C++ model train&inference framework' https://github.com/LieLuoboai/radish

ImJoy：插件化的深度学习混合计算平台 https://github.com/oeway/ImJoy

BytePS：支持TensorFlow, Keras, PyTorch, MXNet的通用高性能训练框架 https://github.com/bytedance/byteps

Alink：基于Flink的通用算法平台，由阿里巴巴计算平台PAI团队研发 https://github.com/alibaba/Alink

机器学习模型管理框架概览：MLFlow/DVC/Sacred https://www.inovex.de/blog/machine-learning-model-management/

Reaction：快速便捷的深度学习模型服务框架 https://github.com/catalyst-team/reaction

深度学习监控自动化 https://medium.com/nanonets/how-to-automate-surveillance-easily-with-deep-learning-4eb4fa0cd68d

机器学习/AI模型的Panini产品化部署(Panini Beta测试阶段免费) https://towardsdatascience.com/deploy-ml-dl-models-to-production-via-panini-3e0a6e9ef14

ONNC：开放神经网络ONNX编译器 https://github.com/ONNC/onnc

Weight Watcher：深度网络泛化精度预测工具 https://github.com/CalculatedContent/WeightWatcher

神经网络模型可视化浏览器(支持ONNX/Keras/Core ML/TensorFlow Lite等，实验性支持PyTorch/Torch/CNTK/TensorFlow.js/TensorFlow等) https://github.com/lutzroeder/Netron

Xfer：MXNet迁移学习库 https://github.com/amzn/xfer/

【PyTorch项目模板】’pytorch-template - my pytorch project template (for kaggle and research)' by lyakaap https://github.com/lyakaap/pytorch-template

【PyTorch实现的XNOR-Net】’XNOR-Net-PyTorch - PyTorch Implementation of XNOR-Net' by jiecaoyu https://github.com/jiecaoyu/XNOR-Net-PyTorch

用PyTorch训练神经网络的简单工具 https://github.com/belskikh/kekas https://github.com/belskikh/kekas/blob/master/Tutorial.ipynb

Keras Tuner：(Google)以人文本的超参调试框架 https://elie.net/talk/cutting-edge-tensorflow-keras-tuner-hypertuning-for-humans/

Keras超参调试器 https://github.com/keras-team/keras-tuner

studio.ml：用来简化、加快模型构建过程的模型管理框架 https://github.com/studioml/studio

【Eisen：深度学习开发框架，模型训练、验证、测试和部署开源软件包】 https://github.com/eisen-ai

【Unity Barracuda：Unity轻量跨平台神经网络推理库】 https://github.com/Unity-Technologies/barracuda-release

Horovod：TensorFlow/Keras/PyTorch/MXNet深度学习分布式训练框架 https://github.com/horovod/horovod

DLPerf：深度学习框架性能分析工具包 https://github.com/Oneflow-Inc/DLPerf

AutoKernel - 简单易用，低门槛的自动算子优化工具，提高深度学习算法部署效率' https://github.com/OAID/AutoKernel

TensorRT Models Deploy from ONNX

https://github.com/linghu8812/tensorrt_inference

Libonnx：轻量、可移植的纯C99 ONNX推理引擎，用于支持硬件加速的嵌入式设备 https://github.com/xboot/libonnx

ONE (On-device Neural Engine)：三星开源的设备端高性能神经网络推理框架 https://github.com/Samsung/ONE

RTNeural：C++实现的轻量实时神经网络推理引擎 https://github.com/jatinchowdhury18/RTNeural

ivy：模板化深度学习框架，支持框架无关的函数、层和库 https://github.com/ivy-dl/ivy

'deepx_core：腾讯开源的专注于张量计算/深度学习的基础库(C++)’ github.com/Tencent/deepx_core

TinyMS：基于MindSpore AI框架开发，面向上层用户的一个高级API开发库 github.com/tinyms-ai/tinyms

Sabertooth：JAX+ Flax实现的的支持CPU/GPU/TPU的独立预训练包 github.com/nikitakit/sabertooth

ivadomed：深度学习医学图像分析集成框架 github.com/ivadomed/ivadomed

TFCC：C++深度学习推理框架 github.com/Tencent/WeChat-TFCC

VNN：高性能、轻量级神经网络部署框架 github.com/joyycom/VNN

Model Zoo for MindSpore github.com/mindspore-ai/models

wonnx - A GPU-accelerated ONNX inference run-time written 100% in Rust, ready for the web github.com/webonnx/wonnx

Easy Parallel Library：面向大模型的高效分布式深度学习训练框架 github.com/alibaba/EasyParallelLibrary

oneAPI Deep Neural Network Library (oneDNN)：oneAPI深度神经网络库 github.com/oneapi-src/oneDNN

Colossal-AI，可用于 AI 大规模并行训练，仅需一半数量的 GPU，便能完成相同效果的 GPT-3 训练工作，极大降低了项目研发成本！ GitHub：github.com/hpcaitech/ColossalAI 近期，该项目终于发布了正式版，重点优化了分布式训练性能、简化项目实用流程，并新增了中文教程，大幅降低开发者的使用成本。

ivy：机器学习统一框架，支持所有框架，目前支持 Jax，TensorFlow，PyTorch，MXNet 和 Numpy github.com/unifyai/ivy

【flashlight：快速、灵活的C++机器学习库，由Facebook AI研究语音团队及Torch和Deep Speech的创作者用C++编写】’flashlight - a fast, flexible machine learning library written entirely in C++ from the Facebook AI Research Speech team and the creators of Torch and Deep Speech' GitHub: https:// github.com/flashlight/flashlight

【Galileo：图深度学习框架，具备超大规模、易使用、易扩展、高性能、双后端等优点】'Galileo - library for large scale graph training by JD' by JD Galileo Team GitHub: github.com/JDGalileo/galileo

'Ecosystem：使用了OpenMMLab体系的开源项目’ by OpenMMLab GitHub: github.com/open-mmlab/ecosystem

【TinyMaix：面向单片机的超轻量级的神经网络推理库，即TinyML推理库】'TinyMaix - a tiny inference library for microcontrollers (TinyML).' by Sipeed GitHub: github.com/sipeed/TinyMaix

'sio4onnx - Simple tool to change the INPUT and OUTPUT shape of ONNX.' by Katsuya Hyodo GitHub: github.com/PINTO0309/sio4onnx

'MegCC - 运行时超轻量，高效，移植简单的深度学习模型编译器' by MegEngine GitHub: github.com/MegEngine/MegCC

Dmlc/dgl: 「Python package built to ease deep learning on graph, on top of existing DL frameworks.」 https://github.com/dmlc/dgl

【ivy：机器学习统一框架，支持所有框架，目前支持 Jax，TensorFlow，PyTorch，MXNet 和 Numpy】’ivy - The Unified Machine Learning Framework' GitHub: github.com/unifyai/ivy