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

Simplified use of debug

Browse Source 
Signed-off-by: Jim Martens <github@2martens.de>
This commit is contained in:
Jim Martens
2019-04-17 14:04:23 +02:00
parent be40a52dde
commit bc07fda615
3 changed files with 24 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
src/twomartens/masterthesis/aae/run.py
View File

@ -42,8 +42,7 @@ def run_simple(dataset: tf.data.Dataset,
channels: int = 3, channels: int = 3,
zsize: int = 64, zsize: int = 64,
batch_size: int = 16, batch_size: int = 16,
verbose: bool = False, verbose: bool = False) -> None:
debug: bool = False) -> None:
""" """
Runs the trained auto-encoder for given data set. Runs the trained auto-encoder for given data set.
@ -57,7 +56,6 @@ def run_simple(dataset: tf.data.Dataset,
zsize: size of the intermediary z (default: 64) zsize: size of the intermediary z (default: 64)
batch_size: size of each batch (default: 16) batch_size: size of each batch (default: 16)
verbose: if True training progress is printed to console (default: False) verbose: if True training progress is printed to console (default: False)
debug: if True summaries are collected (default: False)
""" """
# checkpointed tensors and variables # checkpointed tensors and variables
@ -79,7 +77,6 @@ def run_simple(dataset: tf.data.Dataset,
outputs = _run_one_epoch_simple(dataset, outputs = _run_one_epoch_simple(dataset,
batch_size=batch_size, batch_size=batch_size,
global_step=global_step, global_step=global_step,
debug=debug,
**checkpointables) **checkpointables)
if verbose: if verbose:
@ -91,7 +88,6 @@ def run_simple(dataset: tf.data.Dataset,
def _run_one_epoch_simple(dataset: tf.data.Dataset, def _run_one_epoch_simple(dataset: tf.data.Dataset,
batch_size: int, batch_size: int,
debug: bool,
encoder: model.Encoder, encoder: model.Encoder,
decoder: model.Decoder, decoder: model.Decoder,
global_step: tf.Variable) -> Dict[str, float]: global_step: tf.Variable) -> Dict[str, float]:
@ -104,11 +100,10 @@ def _run_one_epoch_simple(dataset: tf.data.Dataset,
reconstruction_loss, x_decoded, z = _run_enc_dec_step_simple(encoder=encoder, reconstruction_loss, x_decoded, z = _run_enc_dec_step_simple(encoder=encoder,
decoder=decoder, decoder=decoder,
inputs=x, inputs=x,
global_step=global_step, global_step=global_step)
debug=debug)
enc_dec_loss_avg(reconstruction_loss) enc_dec_loss_avg(reconstruction_loss)
if int(global_step % train.LOG_FREQUENCY) == 0 and debug: if int(global_step % train.LOG_FREQUENCY) == 0:
comparison = K.concatenate([x[:int(batch_size / 2)], x_decoded[:int(batch_size / 2)], comparison = K.concatenate([x[:int(batch_size / 2)], x_decoded[:int(batch_size / 2)],
z[:int(batch_size / 2)]], axis=0) z[:int(batch_size / 2)]], axis=0)
grid = util.prepare_image(comparison.cpu(), nrow=int(batch_size / 2)) grid = util.prepare_image(comparison.cpu(), nrow=int(batch_size / 2))
@ -131,8 +126,7 @@ def _run_one_epoch_simple(dataset: tf.data.Dataset,
def _run_enc_dec_step_simple(encoder: model.Encoder, decoder: model.Decoder, def _run_enc_dec_step_simple(encoder: model.Encoder, decoder: model.Decoder,
inputs: tf.Tensor, inputs: tf.Tensor,
global_step: tf.Variable, global_step: tf.Variable) -> Tuple[tf.Tensor, tf.Tensor, tf.Tensor]:
debug: bool) -> Tuple[tf.Tensor, tf.Tensor, tf.Tensor]:
""" """
Runs the encoder and decoder jointly for one step (one batch). Runs the encoder and decoder jointly for one step (one batch).
@ -141,7 +135,6 @@ def _run_enc_dec_step_simple(encoder: model.Encoder, decoder: model.Decoder,
decoder: instance of decoder model decoder: instance of decoder model
inputs: inputs from data set inputs: inputs from data set
global_step: the global step variable global_step: the global step variable
debug: if True summaries are collected
Returns: Returns:
tuple of reconstruction loss, reconstructed input, latent space value tuple of reconstruction loss, reconstructed input, latent space value
@ -151,7 +144,7 @@ def _run_enc_dec_step_simple(encoder: model.Encoder, decoder: model.Decoder,
reconstruction_loss = tf.losses.log_loss(inputs, x_decoded) reconstruction_loss = tf.losses.log_loss(inputs, x_decoded)
if int(global_step % train.LOG_FREQUENCY) == 0 and debug: if int(global_step % train.LOG_FREQUENCY) == 0:
summary_ops_v2.scalar(name='reconstruction_loss', tensor=reconstruction_loss, summary_ops_v2.scalar(name='reconstruction_loss', tensor=reconstruction_loss,
step=global_step) step=global_step)

14
src/twomartens/masterthesis/aae/train.py
View File

@ -52,8 +52,7 @@ def train_simple(dataset: tf.data.Dataset,
lr: float = 0.0001, lr: float = 0.0001,
train_epoch: int = 1, train_epoch: int = 1,
batch_size: int = 16, batch_size: int = 16,
verbose: bool = False, verbose: bool = False) -> None:
debug: bool = False) -> None:
""" """
Trains auto-encoder for given data set. Trains auto-encoder for given data set.
@ -74,7 +73,6 @@ def train_simple(dataset: tf.data.Dataset,
train_epoch: number of epochs to train (default: 1) train_epoch: number of epochs to train (default: 1)
batch_size: size of each batch (default: 16) batch_size: size of each batch (default: 16)
verbose: if True training progress is printed to console (default: False) verbose: if True training progress is printed to console (default: False)
debug: if True summaries are collected (default: False)
""" """
# checkpointed tensors and variables # checkpointed tensors and variables
@ -117,7 +115,6 @@ def train_simple(dataset: tf.data.Dataset,
_epoch = epoch + previous_epochs _epoch = epoch + previous_epochs
outputs = _train_one_epoch_simple(_epoch, dataset, outputs = _train_one_epoch_simple(_epoch, dataset,
verbose=verbose, verbose=verbose,
debug=debug,
batch_size=batch_size, batch_size=batch_size,
**checkpointables) **checkpointables)
@ -141,7 +138,6 @@ def train_simple(dataset: tf.data.Dataset,
def _train_one_epoch_simple(epoch: int, def _train_one_epoch_simple(epoch: int,
dataset: tf.data.Dataset, dataset: tf.data.Dataset,
verbose: bool, verbose: bool,
debug: bool,
batch_size: int, batch_size: int,
learning_rate_var: tf.Variable, learning_rate_var: tf.Variable,
decoder: model.Decoder, decoder: model.Decoder,
@ -170,11 +166,10 @@ def _train_one_epoch_simple(epoch: int,
optimizer=enc_dec_optimizer, optimizer=enc_dec_optimizer,
inputs=x, inputs=x,
global_step_enc_dec=global_step_enc_dec, global_step_enc_dec=global_step_enc_dec,
global_step=global_step, global_step=global_step)
debug=debug)
enc_dec_loss_avg(reconstruction_loss) enc_dec_loss_avg(reconstruction_loss)
if int(global_step % LOG_FREQUENCY) == 0 and debug: if int(global_step % LOG_FREQUENCY) == 0:
comparison = K.concatenate([x[:int(batch_size / 2)], x_decoded[:int(batch_size / 2)], comparison = K.concatenate([x[:int(batch_size / 2)], x_decoded[:int(batch_size / 2)],
z[:int(batch_size/2)]], axis=0) z[:int(batch_size/2)]], axis=0)
grid = util.prepare_image(comparison.cpu(), nrow=int(batch_size/2)) grid = util.prepare_image(comparison.cpu(), nrow=int(batch_size/2))
@ -211,7 +206,6 @@ def _train_enc_dec_step_simple(encoder: model.Encoder, decoder: model.Decoder,
inputs: inputs from data set inputs: inputs from data set
global_step: the global step variable global_step: the global step variable
global_step_enc_dec: global step variable for enc_dec global_step_enc_dec: global step variable for enc_dec
debug: if True summaries are collected
Returns: Returns:
tuple of reconstruction loss, reconstructed input, z value tuple of reconstruction loss, reconstructed input, z value
@ -224,7 +218,7 @@ def _train_enc_dec_step_simple(encoder: model.Encoder, decoder: model.Decoder,
enc_dec_grads = tape.gradient(reconstruction_loss, enc_dec_grads = tape.gradient(reconstruction_loss,
encoder.trainable_variables + decoder.trainable_variables) encoder.trainable_variables + decoder.trainable_variables)
if int(global_step % LOG_FREQUENCY) == 0 and debug: if int(global_step % LOG_FREQUENCY) == 0:
summary_ops_v2.scalar(name='reconstruction_loss', tensor=reconstruction_loss, summary_ops_v2.scalar(name='reconstruction_loss', tensor=reconstruction_loss,
step=global_step) step=global_step)
for grad, variable in zip(enc_dec_grads, encoder.trainable_variables + decoder.trainable_variables): for grad, variable in zip(enc_dec_grads, encoder.trainable_variables + decoder.trainable_variables):

19
src/twomartens/masterthesis/main.py
View File

@ -128,8 +128,13 @@ def _val(args: argparse.Namespace) -> None:
use_summary_writer = summary_ops_v2.create_file_writer( use_summary_writer = summary_ops_v2.create_file_writer(
f"{args.summary_path}/val/category-{category}/{args.iteration}" f"{args.summary_path}/val/category-{category}/{args.iteration}"
) )
with use_summary_writer.as_default(): if args.debug:
run.run_simple(coco_data, iteration=args.iteration_trained, debug=args.debug, with use_summary_writer.as_default():
run.run_simple(coco_data, iteration=args.iteration_trained,
weights_prefix=f"{args.weights_path}/category-{category_trained}",
zsize=64, verbose=args.verbose, channels=3, batch_size=batch_size)
else:
run.run_simple(coco_data, iteration=args.iteration_trained,
weights_prefix=f"{args.weights_path}/category-{category_trained}", weights_prefix=f"{args.weights_path}/category-{category_trained}",
zsize=64, verbose=args.verbose, channels=3, batch_size=batch_size) zsize=64, verbose=args.verbose, channels=3, batch_size=batch_size)
@ -149,10 +154,16 @@ def _auto_encoder_train(args: argparse.Namespace) -> None:
train_summary_writer = summary_ops_v2.create_file_writer( train_summary_writer = summary_ops_v2.create_file_writer(
f"{args.summary_path}/train/category-{category}/{args.iteration}" f"{args.summary_path}/train/category-{category}/{args.iteration}"
) )
with train_summary_writer.as_default(): if args.debug:
with train_summary_writer.as_default():
train.train_simple(coco_data, iteration=args.iteration,
weights_prefix=f"{args.weights_path}/category-{category}",
zsize=64, lr=0.0001, verbose=args.verbose,
channels=3, train_epoch=args.num_epochs, batch_size=batch_size)
else:
train.train_simple(coco_data, iteration=args.iteration, train.train_simple(coco_data, iteration=args.iteration,
weights_prefix=f"{args.weights_path}/category-{category}", weights_prefix=f"{args.weights_path}/category-{category}",
zsize=64, lr=0.0001, verbose=args.verbose, debug=args.debug, zsize=64, lr=0.0001, verbose=args.verbose,
channels=3, train_epoch=args.num_epochs, batch_size=batch_size) channels=3, train_epoch=args.num_epochs, batch_size=batch_size)