33fde6b3013694aa9721bb24996bb6ff28a04d70,returnn/tf/layers/basic.py,BinaryCrossEntropyLoss,get_value,#BinaryCrossEntropyLoss#,6653
Before Change
:rtype: tf.Tensor
with tf.name_scope("loss_bin_ce"):
out = tf.nn.sigmoid_cross_entropy_with_logits(
logits=self.output_flat, labels=tf.cast(self.target_flat, self.output_flat.dtype))
return self.reduce_func(out) * (1.0 / (self.output.dim or 1))
def get_error(self):
After Change
with tf.name_scope("loss_bin_ce"):
target_flat = tf.cast(self.target_flat, self.output_flat.dtype)
if self._pos_weight is None:
out = tf.nn.sigmoid_cross_entropy_with_logits(logits=self.output_flat, labels=target_flat)
else:
out = tf.nn.weighted_cross_entropy_with_logits(
logits=self.output_flat, labels=target_flat, pos_weight=self._pos_weight)
return self.reduce_func(out) * (1.0 / (self.output.dim or 1))
In pattern: SUPERPATTERN
Frequency: 5
Non-data size: 3
Instances Project Name: rwth-i6/returnn
Commit Name: 33fde6b3013694aa9721bb24996bb6ff28a04d70
Time: 2020-08-07
Author: pwilken@apptek.com
File Name: returnn/tf/layers/basic.py
Class Name: BinaryCrossEntropyLoss
Method Name: get_value
Project Name: dhlab-epfl/dhSegment
Commit Name: c9accdd97bb8071058fecf87aab5c166687ecc9b
Time: 2018-01-11
Author: sofia.oliveiraares@epfl.ch
File Name: doc_seg/model.py
Class Name:
Method Name: model_fn
Project Name: tensorflow/ranking
Commit Name: 6bf3f51cd0a312da842157665663c2dad9983248
Time: 2021-01-29
Author: xuanhui@google.com
File Name: tensorflow_ranking/python/losses_impl.py
Class Name: ClickEMLoss
Method Name: compute_unreduced_loss
Project Name: tensorflow/models
Commit Name: 0f7616bdade868fc1dd8c90adb6f7789176305a6
Time: 2020-08-10
Author: 30733558+plakal@users.noreply.github.com
File Name: research/audioset/vggish/vggish_train_demo.py
Class Name:
Method Name: main
Project Name: p2irc/deepplantphenomics
Commit Name: 2e7c32174cdc305765e72a439333634611bc74e5
Time: 2019-10-31
Author: dbl599@mail.usask.ca
File Name: deepplantphenomics/semantic_segmentation_model.py
Class Name: SemanticSegmentationModel
Method Name: _graph_problem_loss