vivekpd15 / iris_keras_tensorflow_serving

Train a model for classification on the iris dataset, and export it for tensorflow serving

Train and Save Model on Iris data set using Keras

Exported Model configured for Tensorflow Serving

There is a lack of simple documentation on to train a model for classification and save it using keras for Tensorflow Serving. Hopefully this example will lighten the way to productionize AI.

Getting Environment Set Up

Clone Repo

git clone git clone https://github.com/brianalois/iris_keras_tensorflow_serving.git
cd iris_keras_tensorflow_serving

PIPENV

Use Pipenv because we are awesome develoeprs

I am using pipenv in order to standardize environments, kind or like the famous NPM for node

Install Pipenv

https://docs.pipenv.org/

pip install pipenv

or if you are using mac install with homebrew

brew install pipenv

Install Dependencies

run this in the repo directory, installs files from Pipfile

pipenv install

Run it using pipenv

pipenv run python index.py

No Pipenv

Don't want to use Pipenv because I am not awesome

If you do not want to use pipenv then you must install these dependencies You must have tensorflow keras, and numpy installed(obviously)

pip install numpy
pip install tensorflow keras pandas sklearn

run the file to export the trained model

python index.py

Variables

There are 2 variables starting at line 18

model_version: change this to change the name of the folder of the specific model version

model_version = "1"

epoch: the higher this number is the more accurate the model, but the longer it will take to train. 5000 is good, but may take a while

epoch = 100