Chainer family

[Updated on May 31: Add detail description for ChainerCV & ChainerMN]

Recently several sub-libraries for Chainer are released,

ChainerRL

RL: Reinforcement Learning

Deep Reinforcement Learning library.

Recent state-of-the-art deep reinforcement algorithms are implemented, including

• A3C (Asynchronous Advantage Actor-Critic)
• ACER (Actor-Critic with Experience Replay) (only the discrete-action version for now)
• Asynchronous N-step Q-learning
• DQN (including Double DQN, Persistent Advantage Learning (PAL), Double PAL, Dynamic Policy Programming (DPP))
• DDPG (Deep Deterministic Poilcy Gradients) (including SVG(0))
• PGT (Policy Gradient Theorem) 

How to install

pip install chainerrl

• github repository
• ChainerRL – Deep Reinforcement Learning Library

ChainerCV

CV: Computer Vision

Image processing library for deep learning training. Common data-augmentation are implemented. Also the trained models for Bounding box detection and semantic segmentation are provided. 

How to install

pip install chainercv 

• github repository
• document

ChainerMN

MN: Multi Node

Distributed deep learning framework for Chainer.

It was announced at Deep Learning Summit 2017 that training time for ImageNet classification task took 4.4 hours (ResNet-50, 100 Epochs, 128 GPUs), which is fastest among other distributed deep learning frameworks known to date.

Reference

• Performance of Distributed Deep Learning using ChainerMN

How to install

You need to install CUDA-aware API and NCCL beforehand, and then,

pip install cython
pip install chainermn

• github repository
• document