In the previous recipe, we focused on the loss and metric. However, with TensorBoard, you can also keep track of the weights. Taking a closer look at the weights can help in understanding how your model works and learns.
- We start by TensorFlow, as follows:
import tensorflow as tf from tensorflow.examples.tutorials.mnist import input_data
- Now, we will be able to load the Fashion-MNIST dataset with just one line of code:
mnist = input_data.read_data_sets('Data/fashion', one_hot=True)
- Before proceeding, we need to set the placeholders for our model:
n_classes = 10 input_size = 784 x = tf.placeholder(tf.float32, shape=[None, input_size]) y = tf.placeholder(tf.float32, shape=[None, n_classes]) keep_prob = tf.placeholder(tf.float32)
- We define four functions that will help us build our network architecture:
def weight_variable(shape): initial = tf.truncated_normal(shape, stddev=0.1) return tf.Variable(initial) def bias_variable(shape): ...