tensorflow-single-node(Python)

Loading...

TensorFlow tutorial - MNIST For ML Beginners

This notebook demonstrates how to use TensorFlow on the Spark driver node to fit a neural network on MNIST handwritten digit recognition data.

Prerequisites:

  • A GPU-enabled cluster.
  • TensorFlow 1.15 or 2.x with GPU support installed manually.

The content of this notebook is adapted from TensorFlow project under Apache 2.0 license with slight modification to run on Databricks. Thanks to the developers of TensorFlow for this example!

Load the data (this step may take a while)

Define the model

Define loss and optimizer

Start TensorBoard so you can monitor training progress.

Train the model in batches