Deploying ML models on GCP MLE using Tensorflow Estimators

Table of Contents

1. Project Motivation
2. Dependencies
3. Datasets
4. Content
5. Instructions
6. Licensing
7. Acknowledgements

Project Motivation:

This sample creates and trains MNIST model in Jupyter Notrbook using Tensorflow and then deploys the trained model to Google Cloud MLE to serve the model, and accept prediction requests by REST API/Google MLE Client API.

Dependencies

refer to requirements.txt

Datasets

Tensorflow MNIST dataset

Content

1. main.py: uses Google Cloud MLE Client API to make prediction calls
2. MNIST Notebook.ipynb: A jupyter notebook file used to create and train the model then use gsutil tool to upload the trained model to Google Cloud Storage (Bucket)

Instructions

1. Clone the repository: git clone https://github.com/m0hamdan/Google-CloudML-MNIST-Digits.git
2. Run the following command in the repo root directory: python main.py

Licensing

None

Acknowledgements

1. Tensorflow webiste as I used the model created here https://www.tensorflow.org/tutorials/estimators/cnn