Pokémon Name Generator

The code for generating pokemon names was forked from this repository. The generated model is used to expose its functionality as part of a REST API.

The model generates new unique Pokemon names with Keras using a recurrent neural network (LSTM). Written as a generic text generator that can be used to generate lines of poetry, other names, or just text in general. It all depends on what it gets input.

Training the model

Train the model with the jupyter notebook in ./training/name_generator.ipynb. It will produce a file which stores the model used to generate new names in ./training/model.h5.

To start the jupyter notebook:

jupyter notebook

Notebook with code and explainations

Sample Pokémon Names

Purndew	Chingoos	Nodow	Fregaycha
Magmagly	Cteenidel	Browodon	Noinga
Ferfeon	Midgeos	Deowwar	Harouthal
Spera	Kleffas	Picorno	Suorthe
Ponytau	Jellpid	Mewable	Meetty
Phound	Passir	Golduzon	Frislask

API Installation

Copy the files that are used for training into the api application:

cp training/input/names.txt api/static
cp training/model.h5 api/static/

Without Docker

Create a virtualenv and activate it:

python3 -m venv venv
. venv/bin/activate

Or windows:

py -3 -m venv venv
venv\Scripts\activate.bat

Install all the dependencies.

pip3 install -r requirements.txt

Start the API

python3 api/main.py

With Docker

Build the Docker container.

docker build -t name-generator .

Run the docker container and map the internal port to external port-

docker run -p 5000:5000/tcp name-generator

Run tests

coverage run -m pytest

API

/names?amount=<amount>

GET

Parameter	Type	Required	Default Value	Description
amount	int	no	1	amount of Pokemon names

Example

http://localhost:5000/?amount=5

["Goigon", "Tangros", "Kalmiphan", "Camermosh", "Wincino"]

Swagger

Endpoints are documented using Swagger. To access the interactive documentation got to:

/swagger