2martens/masterthesis
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Added summaries for grads and learning rate

Browse Source 
Signed-off-by: Jim Martens <github@2martens.de>
This commit is contained in:
Jim Martens
2019-02-08 20:13:21 +01:00
parent e85b850c63
commit a319cf988e
1 changed files with 21 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
src/twomartens/masterthesis/aae/train.py
View File

@ -217,6 +217,10 @@ def train(dataset: tf.data.Dataset, iteration: int,
if last_epoch != -1: if last_epoch != -1:
previous_epochs = last_epoch + 1 previous_epochs = last_epoch + 1
with summary_ops_v2.always_record_summaries():
summary_ops_v2.scalar(name='learning_rate', tensor=checkpointables['learning_rate_var'],
step=checkpointables['global_step'])
for epoch in range(train_epoch - previous_epochs): for epoch in range(train_epoch - previous_epochs):
_epoch = epoch + previous_epochs _epoch = epoch + previous_epochs
outputs = _train_one_epoch(_epoch, dataset, targets_real=y_real, outputs = _train_one_epoch(_epoch, dataset, targets_real=y_real,
@ -327,6 +331,8 @@ def _train_one_epoch(epoch: int, dataset: tf.data.Dataset, targets_real: tf.Tens
# update learning rate # update learning rate
if (epoch + 1) % 30 == 0: if (epoch + 1) % 30 == 0:
learning_rate_var.assign(learning_rate_var.value() / 4) learning_rate_var.assign(learning_rate_var.value() / 4)
summary_ops_v2.scalar(name='learning_rate', tensor=learning_rate_var,
step=global_step)
if verbose: if verbose:
print("learning rate change!") print("learning rate change!")
@ -435,6 +441,7 @@ def _train_xdiscriminator_step(x_discriminator: XDiscriminator, decoder: Decoder
_xd_train_loss = xd_real_loss + xd_fake_loss _xd_train_loss = xd_real_loss + xd_fake_loss
xd_grads = tape.gradient(_xd_train_loss, x_discriminator.trainable_variables)
if int(global_step % LOG_FREQUENCY) == 0: if int(global_step % LOG_FREQUENCY) == 0:
summary_ops_v2.scalar(name='x_discriminator_real_loss', tensor=xd_real_loss, summary_ops_v2.scalar(name='x_discriminator_real_loss', tensor=xd_real_loss,
step=global_step) step=global_step)
@ -442,7 +449,8 @@ def _train_xdiscriminator_step(x_discriminator: XDiscriminator, decoder: Decoder
step=global_step) step=global_step)
summary_ops_v2.scalar(name='x_discriminator_loss', tensor=_xd_train_loss, summary_ops_v2.scalar(name='x_discriminator_loss', tensor=_xd_train_loss,
step=global_step) step=global_step)
xd_grads = tape.gradient(_xd_train_loss, x_discriminator.trainable_variables) summary_ops_v2.histogram(name='x_discriminator_grads', tensor=xd_grads,
step=global_step)
optimizer.apply_gradients(zip(xd_grads, x_discriminator.trainable_variables), optimizer.apply_gradients(zip(xd_grads, x_discriminator.trainable_variables),
global_step=global_step_xd) global_step=global_step_xd)
@ -473,10 +481,12 @@ def _train_decoder_step(decoder: Decoder, x_discriminator: XDiscriminator,
xd_result = tf.squeeze(x_discriminator(x_fake)) xd_result = tf.squeeze(x_discriminator(x_fake))
_decoder_train_loss = binary_crossentropy(targets, xd_result) _decoder_train_loss = binary_crossentropy(targets, xd_result)
grads = tape.gradient(_decoder_train_loss, decoder.trainable_variables)
if int(global_step % LOG_FREQUENCY) == 0: if int(global_step % LOG_FREQUENCY) == 0:
summary_ops_v2.scalar(name='decoder_loss', tensor=_decoder_train_loss, summary_ops_v2.scalar(name='decoder_loss', tensor=_decoder_train_loss,
step=global_step) step=global_step)
grads = tape.gradient(_decoder_train_loss, decoder.trainable_variables) summary_ops_v2.histogram(name='decoder_grads', tensor=grads,
step=global_step)
optimizer.apply_gradients(zip(grads, decoder.trainable_variables), optimizer.apply_gradients(zip(grads, decoder.trainable_variables),
global_step=global_step_decoder) global_step=global_step_decoder)
@ -515,6 +525,7 @@ def _train_zdiscriminator_step(z_discriminator: ZDiscriminator, encoder: Encoder
_zd_train_loss = zd_real_loss + zd_fake_loss _zd_train_loss = zd_real_loss + zd_fake_loss
zd_grads = tape.gradient(_zd_train_loss, z_discriminator.trainable_variables)
if int(global_step % LOG_FREQUENCY) == 0: if int(global_step % LOG_FREQUENCY) == 0:
summary_ops_v2.scalar(name='z_discriminator_real_loss', tensor=zd_real_loss, summary_ops_v2.scalar(name='z_discriminator_real_loss', tensor=zd_real_loss,
step=global_step) step=global_step)
@ -522,7 +533,8 @@ def _train_zdiscriminator_step(z_discriminator: ZDiscriminator, encoder: Encoder
step=global_step) step=global_step)
summary_ops_v2.scalar(name='z_discriminator_loss', tensor=_zd_train_loss, summary_ops_v2.scalar(name='z_discriminator_loss', tensor=_zd_train_loss,
step=global_step) step=global_step)
zd_grads = tape.gradient(_zd_train_loss, z_discriminator.trainable_variables) summary_ops_v2.histogram(name='z_discriminator_grads', tensor=zd_grads,
step=global_step)
optimizer.apply_gradients(zip(zd_grads, z_discriminator.trainable_variables), optimizer.apply_gradients(zip(zd_grads, z_discriminator.trainable_variables),
global_step=global_step_zd) global_step=global_step_zd)
@ -555,6 +567,8 @@ def _train_enc_dec_step(encoder: Encoder, decoder: Decoder, z_discriminator: ZDi
reconstruction_loss = binary_crossentropy(inputs, x_decoded) reconstruction_loss = binary_crossentropy(inputs, x_decoded)
_enc_dec_train_loss = encoder_loss + reconstruction_loss _enc_dec_train_loss = encoder_loss + reconstruction_loss
enc_dec_grads = tape.gradient(_enc_dec_train_loss,
encoder.trainable_variables + decoder.trainable_variables)
if int(global_step % LOG_FREQUENCY) == 0: if int(global_step % LOG_FREQUENCY) == 0:
summary_ops_v2.scalar(name='encoder_loss', tensor=encoder_loss, summary_ops_v2.scalar(name='encoder_loss', tensor=encoder_loss,
step=global_step) step=global_step)
@ -562,8 +576,8 @@ def _train_enc_dec_step(encoder: Encoder, decoder: Decoder, z_discriminator: ZDi
step=global_step) step=global_step)
summary_ops_v2.scalar(name='encoder_decoder_loss', tensor=_enc_dec_train_loss, summary_ops_v2.scalar(name='encoder_decoder_loss', tensor=_enc_dec_train_loss,
step=global_step) step=global_step)
enc_dec_grads = tape.gradient(_enc_dec_train_loss, summary_ops_v2.histogram(name='encoder_decoder_grads', tensor=enc_dec_grads,
encoder.trainable_variables + decoder.trainable_variables) step=global_step)
optimizer.apply_gradients(zip(enc_dec_grads, optimizer.apply_gradients(zip(enc_dec_grads,
encoder.trainable_variables + decoder.trainable_variables), encoder.trainable_variables + decoder.trainable_variables),
global_step=global_step_enc_dec) global_step=global_step_enc_dec)