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

Removed sample generation

Browse Source 
Signed-off-by: Jim Martens <github@2martens.de>
This commit is contained in:
Jim Martens
2019-04-04 17:28:27 +02:00
parent 17537b944b
commit 05fa489905
1 changed files with 2 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

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

@ -59,8 +59,8 @@ def train_simple(dataset: tf.data.Dataset,
epoch as well as after finishing training (or stopping early). When starting
this function with the same ``iteration`` then the training will try to
continue where it ended last time by restoring a saved checkpoint.
The loss values are provided as scalar summaries. Reconstruction and sample
images are provided as summary images.
The loss values are provided as scalar summaries. Reconstruction images are
provided as summary images.
Args:
dataset: train dataset
@ -73,9 +73,6 @@ def train_simple(dataset: tf.data.Dataset,
verbose: if True prints train progress info to console (default: True)
"""
# non-preserved tensors
sample = K.expand_dims(K.expand_dims(K.random_normal((64, zsize)), axis=1), axis=1)
# checkpointed tensors and variables
checkpointables = {
'learning_rate_var': K.variable(lr),
@ -126,16 +123,6 @@ def train_simple(dataset: tf.data.Dataset,
f"Encoder + Decoder loss: {outputs['enc_dec_loss']:.3f}"
))
# save sample image summary
def _save_sample(decoder: model.Decoder, global_step: tf.Variable, **kwargs) -> None:
resultsample = decoder(sample).cpu()
grid = util.prepare_image(resultsample)
summary_ops_v2.image(name='sample', tensor=K.expand_dims(grid, axis=0),
max_images=1, step=global_step)
with summary_ops_v2.always_record_summaries():
_save_sample(**checkpointables)
# save weights at end of epoch
checkpoint.save(checkpoint_prefix)