HowTo: Use TensorFlow to train handwritten digit recognition model on MNIST dataset

The following instruction demonstrates how to build an Interactive Container in TWSC and using Jupyter Notebook's working environment to run MNIST (Handwritten digit recognition dataset) AI training

Step 1. Sign in TWSC

• If you don’t have an account yet, please refer to Sign up for TWSC.

Step 2. Create a TWSC Interactive Container

• Please refer to Interactive Container to create a Interactive Container.

Step 3. Connect to the container and download the training program

• Use Jupyter Notebook to connect to the container and open Terminal.

info

Refer to Using Jupyter Notebook.

• Enter the following command to download the NCHC_GitHub training program to the container.

git clone https://github.com/TW-NCHC/AI-Services.git

Step 4. Conduct AI module training

• Go back to the Jupyter Notebook homepage, click AI-Services/Tutorial_One, click New on the right and then click on the internal option Python3 to open the notebook.

• After opening the Notebook, copy the code in Keras_MNIST.txt under the original directory to the Notebook.

• After copying the code, click the Run button to start training.

• The training results will be displayed below the program.