Last active
December 7, 2019 15:27
-
-
Save FloydHsiu/ab33c7d98d78f9873757810e2f8db50d to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import tensorflow as tf | |
from tensorflow.keras import Model, layers | |
from spectral_normalization import SpectralNormalization | |
class Cnn(k.Model): | |
def __init__(self, name=None): | |
super(cnn, self).__init__(name=name) | |
self.conv1 = SpectralNormalization(layers.Conv2D(1, (3, 3))) | |
self.conv2 = SpectralNormalization(layers.Conv2D(64, (3, 3))) | |
def call(self, inputs, training=False): | |
x = self.conv1(inputs, training) | |
x = self.conv2(x, training) | |
return x | |
cnn = Cnn() | |
optimizers = tf.optimizers.Adam(learning_rate=1e-4, beta_1=0.5) | |
####### Training ####### | |
with tf.GradientTape() as g_tape: | |
outputs = cnn(inputs, training=True) # Set training to True is important | |
# if training is False, convolution kernel won't be updated by spectral normalization. | |
loss = tf.reduce_mean(outputs) | |
grad = g_tape.gradient(loss, cnn.trainable_variables) | |
optimizers.apply_gradient(zip(grad, cnn.trainable_varialbles)) | |
####### Inference ####### | |
outputs = cnn(inputs, training=False) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment