Japanese version is here
Chainer has worked with Amazon Web Services (AWS) to provide access to the Chainer deep learning framework as a listed choice across many of AWS applications. Chainer provides straightforward calculation of deep neural networks in Python. The combination with AWS leverages Chainer’s exceptional abilities in multi-GPU and multi-server scaling, as demonstrated when PFN trained ResNet50 on ImageNet-1K using Chainer in 15 minutes, four times faster than the previous record held by Facebook.
New ChainerMN functions for improved performance in cloud environments and performance testing results on AWS
ChainerMN is a package that adds multi-node distributed learning functionality to Chainer. We have added the following two new functions to v1.2.0 and v1.3.0 of ChainerMN, which are intended to improve the performance on systems whose inter-node communication bandwidth is low.
- Double buffering to conceal communication time
- All-Reduce function in half-precision floats (FP16)
Kubernetes is today the most popular open-source system for automating deployment, scaling, and management of containerized applications. As the rise of Kubernetes, bunch of companies are running Kubernetes as a platform for various workloads including web applications, databases, cronjobs and so on. Machine Learning workloads, including Deep Learning workloads, are not an exception even though such workloads require special hardwares like GPUs.
ONNX support by Chainer
As we mentioned on our blog, Theano will stop development in a few weeks. Many aspects of Chainer were inspired by Theano’s clean interface design, so we would like to introduce Chainer to users of Theano. We hope this article assists interested Theano users to move to Chainer easily.
The Chainer team is saddened to hear about the end of Theano development. Some of us used Theano when we first started studying deep learning and many aspects of Chainer were inspired by Theano’s clean interface design.
We compare the performance of an LSTM network both with and without cuDNN in Chainer. The NVIDIA CUDA® Deep Neural Network library (cuDNN) is a GPU-accelerated library of primitives for deep neural networks. cuDNN provides highly tuned implementations for standard routines such as LSTM, CNN.
Chainer-based deep reinforcement learning library, ChainerRL has been released. https://github.com/pfnet/chainerrl
At Deep Learning Summit 2017 in San Francisco on this January, PFN announced advancements on distributed deep learning using Chainer in multi-node environment. In this post, I would like to explain the detail of the announcement.
Recently we found some great research projects that are using Chainer for their algorithm implementations and experiments. We searched for such publicly available projects on arXiv and summarized them here as a table that lists papers along with their URL links: Research projects using Chainer.
This is the official blog of Chainer, a framework for neural networks. In this blog, we will provide information about Chainer and its development, including:
Chainer is a Python-based, standalone open source framework for deep learning models. Chainer provides a flexible, intuitive, and high performance means of implementing a full range of deep learning models, including state-of-the-art models such as recurrent neural networks and variational autoencoders.
- Released Chainer/CuPy v6.0.0
- ChainerX Beta Release
- Released Chainer/CuPy v5.0.0
- ChainerMN on AWS with CloudFormation
- Open source deep learning framework Chainer officially supported by Amazon Web Services
- New ChainerMN functions for improved performance in cloud environments and performance testing results on AWS
- ChainerMN on Kubernetes with GPUs
- Released Chainer/CuPy v4.0.0