Last active
August 15, 2020 13:04
-
-
Save NiloyPurkait/7ff24a80ef69c7e000888c27e9e18dc5 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
def train_step(inp, tar): | |
# targets shifted by 1 index position | |
tar_inp = tar[:, :-1] | |
tar_real = tar[:, 1:] | |
#Get encoding, combined and decoding masks | |
enc_padding_mask, combined_mask, dec_padding_mask = create_masks(inp, tar_inp) | |
# Initialize Generator and Discriminator gradient tapes | |
with tf.GradientTape() as gen_tape, tf.GradientTape() as disc_tape: | |
# Get the predicted probabilities from generator | |
predictions, _ = generator(inp, tar_inp, | |
True, | |
enc_padding_mask, | |
combined_mask, | |
dec_padding_mask) | |
# Get predicted sequences | |
batch_pred = tf.argmax(predictions, axis=-1) | |
# Pad predicted sequences | |
batch_pred = tf.keras.preprocessing.sequence.pad_sequences(batch_pred, padding='post', | |
value=0, maxlen=tar.shape[-1]) | |
# Get discriminator's predictions of real & generated output | |
disc_preds_real = discriminator([inp, tar], training=True) | |
disc_preds_fake = discriminator([inp, batch_pred], training=True) | |
# Calculate loss using discriminator and generator loss functions | |
d_loss = discriminator_loss(disc_preds_real, disc_preds_fake) | |
g_loss = generator_loss(disc_preds_fake) | |
# Get discriminator gradients and apply using optimizer | |
disc_grads = disc_tape.gradient(d_loss, discriminator.trainable_weights) | |
discriminator_optimizer.apply_gradients(zip(disc_grads, discriminator.trainable_weights)) | |
# Get generator gradients and apply using optimizer | |
gen_grads = gen_tape.gradient(g_loss, generator.trainable_weights) | |
generator_optimizer.apply_gradients(zip(gen_grads, generator.trainable_weights)) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment