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

Added TODOs to train module

Browse Source 
Signed-off-by: Jim Martens <github@2martens.de>
This commit is contained in:
Jim Martens
2019-02-08 21:14:45 +01:00
parent 90fa27d9ef
commit 9bf7229b1e
1 changed files with 5 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
src/twomartens/masterthesis/aae/train.py
View File

@ -30,6 +30,10 @@ Functions:
prepare_training_data(...): prepares the mnist training data prepare_training_data(...): prepares the mnist training data
train(...): trains the AAE models train(...): trains the AAE models
Todos:
- fix early stopping
- fix losses reaching exactly zero
""" """
import functools import functools
@ -621,7 +625,7 @@ def _normalize(feature: tf.Tensor, label: tf.Tensor) -> Tuple[tf.Tensor, tf.Tens
if __name__ == "__main__": if __name__ == "__main__":
tf.enable_eager_execution() tf.enable_eager_execution()
inlier_classes = [0] inlier_classes = [3]
iteration = 1 iteration = 1
train_dataset, _ = prepare_training_data(test_fold_id=0, inlier_classes=inlier_classes, train_dataset, _ = prepare_training_data(test_fold_id=0, inlier_classes=inlier_classes,
total_classes=10) total_classes=10)