38,760
edits
TensorFlow/fr: Difference between revisions
Updating to match new version of source page
No edit summary |
(Updating to match new version of source page) |
||
| Line 645: | Line 645: | ||
Il ne semble pas possible de définir un nombre de CPU depuis la version 2.1. | Il ne semble pas possible de définir un nombre de CPU depuis la version 2.1. | ||
==Known issues== | ==Known issues== | ||
A bug sneaked into the Keras implementation of Tensorflow after version 2.8.3. It affects the performance of the layers used for data augmentation with prefix <i>tf.keras.layers.Random</i> (like <i>tf.keras.layers.RandomRotation</i>, <i>tf.keras.layers.RandomTranslation</i>, etc). It significantly slows down the training process by more than 100 times. Please switch back to version 2.8.0 if you want to use those data augmentation layers until the bug is fixed in future release of Tensorflow. | A bug sneaked into the Keras implementation of Tensorflow after version 2.8.3. It affects the performance of the layers used for data augmentation with prefix <i>tf.keras.layers.Random</i> (like <i>tf.keras.layers.RandomRotation</i>, <i>tf.keras.layers.RandomTranslation</i>, etc). It significantly slows down the training process by more than 100 times. Please switch back to version 2.8.0 if you want to use those data augmentation layers until the bug is fixed in future release of Tensorflow. | ||