- 07 3月, 2018 2 次提交
-
-
由 Francois Chollet 提交于
-
由 Sebastián Ramírez 提交于
* Add optional dependencies' links to download pages in README * Update Graphviz link in README as they just changed it * Update cuDNN link in README
-
- 06 3月, 2018 5 次提交
-
-
由 Taehoon Lee 提交于
-
由 Pasha Khosravi 提交于
There was an extra period in the middle of few sentences in the recurrent.py documentation.
-
由 Taehoon Lee 提交于
-
由 Taehoon Lee 提交于
-
由 Taehoon Lee 提交于
* Make rnn backend tests efficient * Fix PEP8
-
- 05 3月, 2018 4 次提交
-
-
由 Francois Chollet 提交于
-
由 Francois Chollet 提交于
-
由 Sachin Joglekar 提交于
* Rough draft for TimeSeriesSequence * Added basic tests * Remove delay, rename lookback, PEP8 changes * Fixed float error * Add docstring example * Added tests for shuffle & reverse * Consider shape params instead of output_dim * Fixed unpacking error * Rectified doc example * Added multi-dimensional test * Fix naming error
-
由 Taehoon Lee 提交于
-
- 04 3月, 2018 1 次提交
-
-
* Made Sequence iterable * Made it python2 compliant * clean fit_generator * keep mp specs * Fix tests and expl
-
- 03 3月, 2018 2 次提交
-
-
由 Oleg Zabluda 提交于
* Sync docstring of Sequential.fit_generator() with that of improved Model.fit_generator()
-
由 Kevin Wang 提交于
"Invert" is misleading, since it's not clear what "inverting" an input means. "Reverse" is clearly defined for a sequence. The terminology from the second cited paper (cited by the first), is "reverse".
-
- 02 3月, 2018 1 次提交
-
-
由 saralajew 提交于
Just add K.sum() around the margin_loss
-
- 01 3月, 2018 6 次提交
-
-
由 Francois Chollet 提交于
-
由 Francois Chollet 提交于
-
由 Shanqing Cai 提交于
in its function doc string.
-
由 Taehoon Lee 提交于
-
* Made Sequence iterable * Made it python2 compliant
-
由 lenjoy 提交于
* Update Dockerfile by fixing conda dep * Fix comments
-
- 27 2月, 2018 2 次提交
-
-
由 paulpister 提交于
-
由 Nikita Pokidyshev 提交于
WARNING:tensorflow:From .../lib/python3.6/site-packages/keras/backend/tensorflow_backend.py:3148: calling l2_normalize (from tensorflow.python.ops.nn_impl) with dim is deprecated and will be removed in a future version. Instructions for updating: dim is deprecated, use axis instead
-
- 26 2月, 2018 2 次提交
-
-
由 Taehoon Lee 提交于
* Fix `pool_2d` of Theano for backend compatibility * Add a test
-
由 Taehoon Lee 提交于
-
- 25 2月, 2018 2 次提交
-
-
由 gabrieldemarmiesse 提交于
* Started refactoring. * Added ConvRecurrent2D to legacy/layers. * Updated docstrings. * Fixed a small bug due to K.int_shape when using Theano. * Fixed docstrings. * Made ConvRNN2D sub-class of RNN. * Used K.is_keras_tensor, made shorter lines, among various other fixes. * Fixed Theano dropout issue. * Fixed theano issue with keras shape. * Fixed losses attribute of RNN class.
-
由 Taehoon Lee 提交于
-
- 24 2月, 2018 1 次提交
-
-
由 Taehoon Lee 提交于
* Add depthwise conv2d for Theano and CNTK * Resolve a conflict
-
- 23 2月, 2018 3 次提交
-
-
由 davinnovation 提交于
* Fix ImageDataGenerator preprocessing_function
-
由 Taehoon Lee 提交于
-
由 treszkai 提交于
`cifar10.load_data()` and `mnist.load_data()` do not shuffle the datasets. Fix comments in examples accordingly.
-
- 22 2月, 2018 7 次提交
-
-
由 abrad1212 提交于
* Add Random image brightness
-
由 Francois Chollet 提交于
-
由 Kamil Sindi 提交于
* Add train test split to DirectoryIterator * Fix image.py errors * Add tests to directory_iterator_with_validation_split * Update image.py documentation * Update image.py documentation * Fix pep8 W293 error * Remove do_validatoin_split from _list_valid_filenames_in_directory * Update comment * Add validation_split to generator flow * Add test_image_data_generator_with_split_value_error * Remove num_subdir_files * Refactor to using split * Simplify split logic * Remove MAX_NUM_IMAGES_PER_CLASS * Update validation docs * Make validation_split strictly between 0 and 1 * Remove validation_split is None statements
-
由 DomHudson 提交于
* Slight optimisation * PEP8 fix
-
* Force update of Sequences * Now test both Unix and Win * pep8 on datautils
-
由 Taehoon Lee 提交于
-
由 Taehoon Lee 提交于
-
- 21 2月, 2018 2 次提交
-
-
由 Taehoon Lee 提交于
-
由 Pavithra Vijay 提交于
-