tensorflow-mnist-tutorial

Tutorial code for training on MNIST dataset using high-level APIs in TensorFlow.

Prerequisites

• TensorFlow 1.4.1

Setup

Download the MNIST dataset and persist as binary files using the following command:

python3 experiment.py --download_dataset True --data_dir <data_dir>

• data_dir: Path where to save the MNIST dataset

Now, run the training and evaluation via the following command:

python3 experiment.py --batch_size <batch size> --epochs <epochs> --data_dir <data dir> --model_dir <model dir>

• data_dir: Path from where to load input data
• model_dir: Path where model params and summaries will be saved
• batch_size: Batch size
• epochs: Number of epochs