Building Machine Learning Infrastructure!
算法/模型优化
- 深度学习中的Normalization模型
- How Does Batch Normalization Help Optimization?
- Non-convex Optimization for Machine Learning, 报道
工作流
TFX
TensorFlow Serving
可视化
TensorSpace
AutoML
- 谷歌AutoML创造者Quoc Le:未来最好的人工智能模型将由机器生成
- Taking Human out of Learning Applications: A Survey on Automated Machine Learning, 中文
- Neural Architecture Search: A Survey, 中文
- ENAS: 更有效地设计神经网络模型(AutoML)
- 中科院 & 图森提出通过稀疏优化进行一次神经架构搜索
- 视频:大规模机器学习与AutoML技术
AdaNet
- Introducing AdaNet: Fast and Flexible AutoML with Learning Guarantees
- 谷歌开源AdaNet:基于TensorFlow的AutoML框架
Google Vizier
Google Vizier: A Service for Black-Box Optimization 开源实现:
katib,
advisor
框架
OneFlow
TensorFlow
- TensorFlow内核剖析
- 源码分析:TensorFlow Ops, TensorFlow Kernels
- 基于Tensorflow高阶API构建大规模分布式深度学习模型: 开篇, 基于Dataset API处理Input pipeline, Estimator, 特征工程, CVR预估之ESMM
Mesh TensorFlow: Model Parallelism Made Easier
Uber Horovod
Horovod: fast and easy distributed deep learning in TensorFlow
视频:An Uber Journey in Distributed Deep Learning
PyTorch
Caffe: Convolutional Architecture for Fast Feature Embedding
Github
Ecosystem
MXNet
Mxnet: A flexible and efficient machine learning library for heterogeneous distributed systems
Ecosystem
Awesome-MXNet
GluonCV,
GluonNPL
- 基于系统和算法的协同设计的大规模分布式机器学习:论文及PPT, 视频, Yanring_笔记
- MXNet设计和实现简介 – 2015.12.03
- 2018年了,MXNet 发展的如何了
Petuum
- 视频:System and Algorithm Co-Design, Theory and Practice, for Distributed Machine Learning
- A New Look at the System, Algorithm and Theory Foundations of Distributed Machine Learning
- Petuum Poseidon:高效的分布式深度学习平台, PPT
- 分布式深度学习系统:上、下
- 通向分布式深度学习系统
Ray
Adam
Project Adam: Building an Efficient and Scalable Deep Learning Training System
调度方案
Kubernetes
- Deploy Your First Deep Learning Model On Kubernetes With Python, Keras, Flask, and Docker, 翻译
- 基于 K8S 构建 AI 平台方案解析, 视频
- 视频:基于 Kubernetes 的分布式 TensorFlow 系统
PaddlePaddle EDL
kubeflow
YARN
Submarine
Spark
Mesos
编译优化
Halide
XLA
TVM
- TVM: An Automated End-to-End Optimizing Compiler for Deep Learning, v1中文, 视频
- 如何评价陈天奇的模块化深度学习系统NNVM?
- 深度学习编译中间件之NNVM:1~17
- 如何评价陈天奇团队新开源的TVM?
- NNVM和TVM交互和功能分工:1, 2
- 如何学习TVM的代码?
- 视频:Efficient Deep Learning on Edge Device with IR Stack
VTA
Tensor Comprehensions
Github
Announcing Tensor Comprehensions
Tensor Comprehensions: Framework-Agnostic High-Performance Machine Learning Abstractions
Glow
Glow: Graph Lowering Compiler Techniques for Neural Networks