Twister2

Twister2

  • Getting Started
  • Docs
  • Tutorial
  • AI
  • Examples
  • Contribute
  • Download
  • Configurations
  • Java Docs
  • GitHub
  • Blog

›Deep Learning

AI

  • Overview

Twister2Flow

  • Overview on Twister2Flow

Machine Learning

  • Overview on Machine Learning
  • K-Means
  • SVM

Deep Learning

  • Overview on Deep Learning
  • Pytorch Integration

Deep Learning

Twister2 Deep Learning is currently working on supporting Pytorch, Tensorflow and MXNet. With our 0.5.0 release we have introduced an experimental version of a deep learning with distributed data parallel Pytorch implementation. We use an MPI backend to support distributed training and currently we are working on providing better performance with GPU.

Twister2:DeepNet

Twister2:DeepNet is the distributed runtime for deep learning algorithm training and inference. Currently we support an experimental version of a distributed training for Pytorch. Our initial implementation uses Apache Arrow and Parquet to move data from our data pre-processing pipeline to the data training framework. We are actively working on transforming this process into a seamless in-memory computation model.

← SVMPytorch Integration →
Twister2
Docs
Getting Started (Quickstart)Guides (Programming Guides)
Community
Stack OverflowProject Chat
More
BlogGitHubStar
Copyright © 2020 Indiana University