shawn-shan / eru

High Level Framework for PyTorch

ERU: One Framework for all

You have just found ERU.

High Level Framework for PyTorch

Eru is designed to help with Deep Learning developer to try out advanced models efficiently, so we recommend users to have some understanding of deep learning before using. For this reason, Eru is only implement on GPUs, but we will build CPU configuration soon.

Quick Start

Comming Soon

We are writing a detailed version of the documentation and tutorial...

Examples

We have several illustrative examples ranging from basic MLP to sentiment classifiers in the example folder

Want to contribute?

We welcome anyone who have some python or pytorch experience to join us. More information in CONTRIBUTING.md

Join our Slack channel first

Installation

Before installing Eru, please install Pytorch following the instruction on http://pytorch.org/

• Install Eru from PyPI (recommended):

pip3 install eru

eru's Issues

Syntax suggestion

I've been coding in python for quite a while now, but I've never come across the following syntax:

x = GRU(10, 128, return_sequence=False)(input) 
x = Dense(128, 10)(x)
x = Activation("softmax")(x)

What does it do?

By the way, I really like the idea of this project: PyTorch needs some abstraction to make it as usable as keras. I'd like to contribute at some stage. Just FYI, there's a similar sounding project called (torchsample)[https://github.com/ncullen93/torchsample] remove the need for a lot of PyTorch boilerplate.

(nice Silmarillion reference)