Skip to content

Releases: franneck94/TensorCross

2.0.0

20 May 18:52
Compare
Choose a tag to compare

Version 2.0.0: May 20, 2024

  • This will be the stable version for TF 2.13+ which is the first version that supports keras 3.0

Version 1.0.0

29 Oct 10:41
Compare
Choose a tag to compare

Version 1.0.0: October 29, 2023

  • Some rework, updates docs and versioning. This will be the stable version for TF (Keras) 2.8+ and Python 3.9+

Version 0.4.4

13 Mar 07:20
ba9cb84
Compare
Choose a tag to compare

Version 0.4.4: March 13, 2023

  • Bugfix for MacOS tensorflow metal

Release 0.4.3

12 Sep 16:50
Compare
Choose a tag to compare

Changelog

  • Bugfix for computing the mean score in CV

Release 0.4.2

12 Sep 07:00
Compare
Choose a tag to compare

Release 0.4.2

  • Bugfix for detecting if the mode is either minimize the loss or maximize the metric

Release 0.4.1

11 Sep 07:54
Compare
Choose a tag to compare

Version 0.4.1: September 11, 2021

  • Added detailed information to docstrings
  • The summary method now returns the printed string, such that it can be stored to a file
  • Now also exporting the dataset_join function

Release 0.4.0

10 Sep 08:29
Compare
Choose a tag to compare

Changes

  • Updated TF Version

Release 0.3.0

16 Dec 15:19
Compare
Choose a tag to compare

Features

Added support for TensorFlow 2.0+ (previously only 2.3)

Bug FIxes

Bug fix in CrossValidation search, by replacing the parameter name train_dataset with dataset.

Release 0.2.1

28 Nov 18:48
Compare
Choose a tag to compare

Changes in Release 0.2.1

Updated Version number, due to missing update in last release.

New Features from Release 0.2.0

Implemented the GridSearchCV and RandomSearchCV.
ALso implemented a modification of the TensorBoard callback to store the model
results of each parameter combination in a distinct sub-sir.

Release 0.2.0

28 Nov 18:40
b3b0e67
Compare
Choose a tag to compare

New Features

Implemented the GridSearchCV and RandomSearchCV.
ALso implemented a modification of the TensorBoard callback to store the model
results of each parameter combination in a distinct sub-sir.