https://github.com/nicklhy/MemN2N

MXNet implementation of End-To-End Memory Networks for language modelling.
https://github.com/nicklhy/MemN2N

Last synced: 2 months ago
JSON representation

MXNet implementation of End-To-End Memory Networks for language modelling.

• Host: GitHub
• URL: https://github.com/nicklhy/MemN2N
• Owner: nicklhy
• Created: 2016-12-22T08:35:47.000Z (about 8 years ago)
• Default Branch: master
• Last Pushed: 2016-12-26T08:58:36.000Z (about 8 years ago)
• Last Synced: 2024-08-01T22:41:45.564Z (5 months ago)
• Language: Python
• Size: 1.97 MB
• Stars: 11
• Watchers: 2
• Forks: 2
• Open Issues: 1
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• Awesome-MXNet - [sym - Mxnet-Gluon) (<a name="NLP"></a>3. NLP / 2.14 Misc)

README

        
End-To-End Memory Networks in MemN2N

========================================

MXNet implementation of [End-To-End Memory Networks](http://arxiv.org/abs/1503.08895v4) for language modelling. The original Tensorflow code from [carpedm20](https://github.com/carpedm20) can be found [here](https://github.com/carpedm20/MemN2N-tensorflow).

![alt tag](http://i.imgur.com/nv89JLc.png)

**Known issue: SGD does not converge, ADAM converges but is not able to reach a good result([details](https://github.com/nicklhy/MemN2N/issues/1)).**

Setup

--------------

This code requires [MXNet](https://github.com/dmlc/mxnet). Also, it uses CUDA to run on GPU for faster training. There is a set of sample Penn Tree Bank (PTB) corpus in `data` directory, which is a popular benchmark for measuring quality of these models. But you can use your own text data set which should be formated like [this](data/).

Usage

--------------

To train a model with 6 hops and memory size of 100, run the following command:

    $ python train.py --nhop 6 --mem_size 100

To see all training options, run:

    $ python train.py --help

To test a model, run the script file test.py like:

    $ python test.py --network checkpoint/memnn-symbol.json --params checkpoint/memnn-0100.params --gpus 0