chainer

Version:

7.7.0, 7.8.0

Category:

ai

Cluster:

Loki

Author / Distributor

https://chainer.org/ https://github.com/chainer/chainer

Description

Chainer is a flexible, intuitive, and powerful deep learning framework written in Python. It supports define-by-run (dynamic computation graphs), making it highly suitable for advanced research and development in neural networks.

Chainer supports: - Feedforward neural networks (FNNs) - Convolutional neural networks (CNNs) - Recurrent neural networks (RNNs), including LSTM - CUDA acceleration via CuPy - Training on multiple GPUs

It integrates well with other scientific libraries such as NumPy, CuPy, and Matplotlib, and is compatible with tools like TensorBoard.

Documentation

Official documentation: https://docs.chainer.org/

GitHub repository: https://github.com/chainer/chainer

Examples/Usage

  • Load the Chainer module:

$ module load chainer/python/3.8/7.8.1

  • Run a Python script using Chainer:

$ python3 train_model.py

  • Check GPU availability (if installed with CuPy):

import cupy
print(cupy.cuda.runtime.getDeviceCount())

  • Unload the Chainer module:

$ module unload chainer/python/3.8/7.8.1

Installation

Source code is obtained from Chainer