wellton/keras
1
0
Fork 0
Código Issues Pull Requests Ações Pacotes Projetos Versões Wiki Atividade
3.717 Commits 6 Branches 30 Tags
master
Gráfico de Commits

314 Commits

Autor SHA1 Mensagem Data
Taehoon Lee 3fe7837691 Update the old interfaces (#7252) 2017-07-06 04:26:47 -07:00
Taehoon Lee 59cd1c3994 Fix typos (#7213) 2017-07-02 19:23:33 -07:00
Hideaki Kanehara a85263fb3e unify coding style (#7129) 2017-06-26 11:16:03 -07:00
Taehoon Lee f430de10fb Style fixes (#7073) 2017-06-21 10:58:38 -07:00
Zafarali Ahmed 8d5b2ce60c Add example to compare RELU with SELU (#6990) 
* Add exampe to compare RELU with SELU fchollet/keras#6924

* Add header description

* Add axes labels

* Increase size of MLP #6990

* Reduce network size, reduce dropout rate, reduce dense units

* Reduce network size, add recommendations to reduce overfitting

* Encapsulate hyperparameters and create generic network builder

* Rename file to be more descriptive

* Add @tboquet's suggestion to export to png #6990

* Docstring clean-up

* Change optimizer to sgd, increase epochs

* Update docstrings

* Fix PEP8
 2017-06-16 15:00:16 -07:00
Simon Brugman 295e4f8064 Update lstm_benchmark.py (#6966) 
Removed typo
 2017-06-13 11:13:21 -07:00
Somshubra Majumdar 53303fdb10 Improvements to style transfer as discussed in https://github.com/fchollet/keras/pull/6872 (#6877) 2017-06-06 23:40:32 -07:00
webzjuyujun 1b67c59de8 Style fix. 2017-05-24 19:27:23 -07:00
rejunity e177397427 Small fixes for Neural_Doodle example (#6577) 
* Fixed type conversion in neural_doodle example. Shape returns number of channels as int32 however further calculations require it to be float

* Updated neural doodle example to follow Keras2 API. Renamed ‘border_mode’ argument to ‘padding’.

* Fixed apostrophe for consistency.
 2017-05-11 14:33:39 -07:00
Francois Chollet 7481b5d060 Update deep dream config. 2017-04-24 19:03:39 -07:00
Francois Chollet 9295efb216 Simplify the deep dream example 2017-04-24 18:23:09 -07:00
Mohanson 5bd3976e79 Spelling errors (#6232) 2017-04-12 22:10:15 -07:00
Francois Chollet 9eb7ecd3e5 Merge branch 'master' of github.com:fchollet/keras 2017-04-11 13:56:43 -07:00
Francois Chollet 4aa41625bf Switch variational examples to new API. 2017-04-11 13:43:04 -07:00
Francois Chollet c029fa2f62 Merge branch 'origin/vae_add_loss' of https://github.com/Spotlight0xff/keras into Spotlight0xff-origin/vae_add_loss 2017-04-11 12:57:28 -07:00
Nigel Ng 52b1377fe6 Update mnist_siamese_graph example (#6223) 
Take max of squared distance and K.epsilon() because some data points will throw `nan` for euclidean distance.
 2017-04-11 12:09:44 -07:00
TimHo 98ec9fc972 fix rmsprop learning rate for convergence (#6182) 
Rmsprop with default learning rate (0.001) cannot converge in this example. 
Initialize learning rate to (0.0001) and add weight decay fix the problem.
 2017-04-06 10:07:25 -07:00
Carl Thomé 7c6463da6f Spelling (#6149) 2017-04-04 11:28:16 -07:00
Mike Henry 655f5af76e Fixed URL for wordlist.tgz in image_ocr.py (#6136) 2017-04-03 23:55:18 -07:00
Wang Cheng fe48b41c22 remove unused variables in cifar10_cnn (#6112) 2017-04-02 08:02:35 -07:00
Fariz Rahman b260333eed Bug fix: ocr example; python 3 (#6060) 2017-03-30 13:24:12 +02:00
Junwei Pan fa4c747b7e Typo Fix (#6017) 2017-03-28 13:44:56 +02:00
scott-vsi 9f6fb452a2 Fix typo 
Looks like eec61d9 changed the stride from 1 to 2.
 2017-03-28 00:49:48 +02:00
Junwei Pan befbdaa076 Style fix for examples. (#5980) 2017-03-26 16:27:49 +02:00
Dr. Kashif Rasul 2b1c4779ff updated image_ocr to keras-2 API (#5843) 
* updated image_ocr to keras-2 API

* fixed doc

* pep8

* renamed filter_size to kernel_size
 2017-03-19 11:47:58 +01:00
Ben 5d512f82b5 Replace references to get_output_shape_for with compute_output_shape (#5848) 2017-03-18 20:43:17 -07:00
Francois Chollet 459d7fe3d7 Style fixes in example scripts 2017-03-15 21:13:31 -07:00
Fariz Rahman 1a0792ae13 Update babi_memnn.py 2017-03-16 09:13:58 +05:30
Fariz Rahman b5a02391e0 Upgrade memory network example to functional api 2017-03-16 09:12:07 +05:30
3h4 aff40d8008 Clarifying comment to stateful LSTM example (#5787) 
* Clarifying comment to stateful LSTM example

* Update stateful_lstm.py
 2017-03-15 11:04:50 -07:00
Spotlight0xff e848463347 using .add_loss in custom layer for VAE example 2017-03-15 13:21:45 +01:00
Francois Chollet ce3093a3b2 PEP8 fixes. 2017-03-15 00:16:44 -07:00
antonmbk c45f48eaea SWWAE Example: Conv kernel size in resid pathway to 1x1 and activation from BN+RELU to ELU (#5756) 
* Changed conv kernel size in resid pathway to 1x1, and changed activation from BN+RELU to ELU.

* Added a more informative docstring decsribing elu argument and its two behaviors.
 2017-03-14 22:54:25 -07:00
Francois Chollet aa826d684d Finish updating examples. 2017-03-11 19:44:29 -08:00
Francois Chollet d51fc7659e Update a couple examples. 2017-03-11 15:03:43 -08:00
Hiroya Chiba 65ce238f03 skip newsgroup header (#5585) 2017-03-03 17:45:54 -08:00
Francois Chollet 5d0cb10949 Make utils globally importable & update examples. 2017-02-28 14:41:30 -08:00
Spotlight0xff 4e50446279 Updated VAE examples (MLP and ConvNet) to the new API (#5552) 
* update VAE examples (MLP and ConvNet) to the new API

* renamed objectives to metrics for xent_loss
* std -> stddev for random_normal
* adjusted arguments to Conv2D/Deconv2D
* fixed typo (filterss -> filters)

* change to conv2dtranspose and tuple strides
 2017-02-28 10:21:05 -08:00
Francois Chollet fbe7873fc0 sum -> add 2017-02-28 09:04:00 -08:00
Hannah Vivian Shaw eec61d9d49 Update several examples to work with the new API (#5548) 
* Update mnist_transfer_cnn for new API

* Update mnist_siamese_graph.py for new API

* Refactor example a little bit for clarity

* Update mnist_irnn.py for new API

* Fix variable name

* Update mnist_heirarchial_rnn.py for new api

* Fix a few api calls i missed

* Update mnist_acgan.py for new API

* Fix variable name

* Update imdb_cnn for new API

* Update benchmark.py to work with new API

* PEP8 fix

* Change filter_length to kernel_size

* Update imdb_cnn_lstm.py for new API

* PEP8 indentation fix
 2017-02-27 18:53:41 -08:00
Francois Chollet 5ae158e8b0 Update fit_generator APIs. 2017-02-23 12:46:00 -08:00
Pokey Rule 7e7d2ed1f9 Fix deprecated scoring function (#5452) 
`'log_loss'` is deprecated; replaced by `'neg_log_loss'`
 2017-02-21 15:07:20 -08:00
fchollet 803e2869c7 Update a number of example scripts. 2017-02-19 19:24:32 -08:00
Francois Chollet 10bfb1c565 Integration tests passing. 2017-02-14 16:08:30 -08:00
fchollet ca23406974 Update 2 mnist examples 2017-02-10 18:16:11 -08:00
Francois Chollet 023331ec2a Progress on early draft of Keras 2 2017-02-09 17:16:25 -08:00
Francois Chollet 97f327317f fix merge conflicts 2017-02-08 11:04:30 -08:00
Shusen Liu 757b3ed1b0 fix mnist_sklearn_wrapper.py error when default image_dim_ordering=tf (#5130) 
* fix mnist_sklearn_wrapper.py error when default image_dim_ordering=tf

as title, mnist_sklearn_wrapper.py assumes image_dim_ordering to be "th"
                                                                                                                                                                                                                      Test Plan:                                                                                                                                                                                                        KERAS_BACKEND=theano condapython examples/mnist_sklearn_wrapper.py                                                                                                                                                KERAS_BACKEND=tensorflow condapython examples/mnist_sklearn_wrapper.py

* 2nd commit

Summary:

Test Plan:

Reviewers:

Subscribers:

Tasks:

Blame Revision:
 2017-02-06 14:04:53 -08:00
Minkoo Seo f1a95869eb Fix off by one error in WE example script 
Tokenizer returns sequence values in the range of [0, nb_words). In this
example, MAX_NB_WORDS is 20000 and the data's min value is 19999. There
is no need to use 'nb_words + 1'.
 2017-02-06 10:53:33 -08:00
Minkoo Seo 0f4be6d17b Style fixes in example script 'addition_rnn'. 
Adds more explanation. Make the code fit 80 column and intuitive.
 2017-01-31 11:57:28 -08:00