# key-value-memory-networks

**Repository Path**: greitzmann/key-value-memory-networks

## Basic Information

- **Project Name**: key-value-memory-networks
- **Description**: Key-Value Memory Networks for Directly Reading Documents, Alexander Miller, Adam Fisch, Jesse Dodge, Amir-Hossein Karimi, Antoine Bordes, Jason Weston https://arxiv.org/abs/1606.03126
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2019-11-27
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Key Value Memory Networks by Keras

![Key Value Memory Networks](https://raw.githubusercontent.com/jojonki/key-value-memory-networks/images/kvmemnns.png)

See original thesis.
> Key-Value Memory Networks for Directly Reading Documents, Alexander Miller, Adam Fisch, Jesse Dodge, Amir-Hossein Karimi, Antoine Bordes, Jason Weston
https://arxiv.org/abs/1606.03126

There are still some bugs in my code, pull-requests will be appreciated! :beer:

## Setup

First of all, you need to download dataset.
```
$ ./download.sh
```



## Train

You can use presaved pickle files or build data dictionaries with `process_data.py`.

```
$ python train.py
```

## Evaluate
```
$ python evaluate.py -m saved_keras_model_path
```

## Interactive
```
$ python interactive.py -m saved_keras_model_path
```