This repository has been archived by the owner on Jul 1, 2024. It is now read-only.

Keras-mxnet 2.2.4 release#193

Merged

roywei merged 172 commits intomasterfrom keras-mxnet-2.2.4

Oct 9, 2018

+6,699-3,486

Commits on Aug 14, 2018

Commits on Aug 15, 2018

Style fixes for enabling PEP8 501 (keras-team#10916 )

taehoonlee
authored and
fchollet
committed
Update TensorFlow as 1.9 on Travis (keras-team#10674 )

taehoonlee
authored and
fchollet
committed

Commits on Aug 16, 2018

Commits on Aug 18, 2018

EarlyStopping: restore model weights corresponding to best value of monitored quantity (keras-team#10765 )

MarcoAndreaBuchmann
authored and
taehoonlee
committed

Commits on Aug 21, 2018

Fix CI on TensorFlow (keras-team#10929 )

taehoonlee
authored and
fchollet
committed
Fix multi-output Siamese networks in multi_gpu_model (keras-team#10856 )

datumbox
authored and
fchollet
committed
add installing pydot (keras-team#10937 )

kouml
authored and
fchollet
committed
Update cifar10_cnn_capsule.py (keras-team#10927 )

jankrepl
authored and
taehoonlee
committed
Fix Theano shape inference bug.
fchollet
committed
Speed up backend tests (keras-team#10930 )

taehoonlee
authored and
fchollet
committed
Style fixes for enabling PEP8 501 (keras-team#10931 )

taehoonlee
authored and
fchollet
committed
Bypass CNTK shape inference issue.
fchollet
committed
Merge branch 'master' of github.com:fchollet/keras
fchollet
committed
Support recurrent kernel using identity initializer (keras-team#10934 )

joelthchao
authored and
fchollet
committed
Fix val_step in fit_generator with Sequence (keras-team#10946 )

Frédéric Branchaud-Charron
authored and
fchollet
committed
Refactoring: Made an abstract class for all the cropping layers. (keras-team#10888 )

gabrieldemarmiesse
authored and
fchollet
committed

Commits on Aug 23, 2018

[P] add data_format support for Pooling1D layers (keras-team#10966 )

hsgkim
authored and
fchollet
committed
Removed duplicated backend test. (keras-team#10959 )

gabrieldemarmiesse
authored and
fchollet
committed
Speed up backend tests (keras-team#10956 )

taehoonlee
authored and
fchollet
committed
Fix dilated convolution for CNTK backend. (keras-team#10967 )

yanboliang
authored and
fchollet
committed
Generators now use same logic as Sequence (keras-team#10925 )

Frédéric Branchaud-Charron
authored and
fchollet
committed

Commits on Aug 25, 2018

Commits on Aug 27, 2018

Fix broken links in FAQ (keras-team#11010 )

JamesHinshelwood
authored and
fchollet
committed
ignore PEP8 W503 (keras-team#10976 )

joelthchao
authored and
fchollet
committed

Commits on Aug 28, 2018

Used "pip install cntk" to simplify the installation of CNTK. (keras-team#11011 )

gabrieldemarmiesse
authored and
fchollet
committed
Speed up backend tests (keras-team#10992 )

taehoonlee
authored and
fchollet
committed
Trying to find a clue about the flaky test of (keras-team#10995 )

gabrieldemarmiesse
authored and
fchollet
committed
Speed up random_normal backend test.
fchollet
committed
Convolutional layer supports float64 dtype after tensorflow 1.8.0 (keras-team#10977 )

joelthchao
authored and
fchollet
committed
Update causal Conv1D doc : uses zero padding keras-team#8751 (keras-team#10999 )

BertrandDechoux
authored and
fchollet
committed
Remove unnecessary pytest timeout
fchollet
committed
Made a base class for ZeroPadding. (keras-team#10984 )

gabrieldemarmiesse
authored and
fchollet
committed
Reduce test flakyness.
fchollet
committed
Merge branch 'master' of github.com:fchollet/keras
fchollet
committed
Conv1D and Conv3D supporting dilated conv for CNTK backend (keras-team#10997 )

yanboliang
authored and
fchollet
committed
Allow TB callback to display float values.
fchollet
committed
Fix issue with non-canonical TF version name format.
fchollet
committed

Commits on Aug 29, 2018

Separate pooling test from convolutional test and parameterize test case (keras-team#10975 )

joelthchao
authored and
fchollet
committed

Commits on Aug 30, 2018

[RELNOTES] Added the mode "bilinear" in the upscaling2D layer. (keras-team#10994 )

gabrieldemarmiesse
authored and
fchollet
committed
[P] Expose monitor value getter for easier subclass (keras-team#11002 )

joelthchao
authored and
fchollet
committed
Speeding up the tests by reducing the number of K.eval(). (keras-team#11036 )

gabrieldemarmiesse
authored and
Frédéric Branchaud-Charron
committed
fix a bug, load_weights doesn't return anything (keras-team#11031 )

linjinjin123
authored and
fchollet
committed
Update lstm text generation example (keras-team#11038 )

hadifar
authored and
fchollet
committed
Better UX (keras-team#11039 )

Frédéric Branchaud-Charron
authored and
fchollet
committed
Enable using last incomplete minibatch (keras-team#8344 )

ozabluda
authored and
fchollet
committed
Fix line too long in mnist_acgan (keras-team#11040 )

yanboliang
authored and
fchollet
committed

Commits on Aug 31, 2018

Doc Change: Change in shape for CIFAR Datasets (keras-team#11043 )

Sri Yogesh K
authored and
fchollet
committed
[P, RELNOTES] Conv2DTranspose supports dilation (keras-team#11029 )

yanboliang
authored and
fchollet
committed
Removing duplicated backend tests. (keras-team#11037 )

gabrieldemarmiesse
authored and
fchollet
committed
Cached the theano compilation directory. (keras-team#11048 )

gabrieldemarmiesse
authored and
fchollet
committed

Commits on Sep 1, 2018

Commits on Sep 2, 2018

Make sure the data_format argument defaults to ‘chanels_last’ for all 1D sequence layers.
fchollet
committed

Commits on Sep 3, 2018

Added in_train_phase and in_test_phase in the numpy backend. (keras-team#11061 )

gabrieldemarmiesse
authored and
fchollet
committed

Commits on Sep 4, 2018

Update Dockerfile to fix error introduced by keras-team#11011 (keras-team#11073 )

cshubhamrao
authored and
fchollet
committed

Commits on Sep 5, 2018

Made the backend tests faster by using the numpy backend in the (keras-team#11062 )

gabrieldemarmiesse
authored and
taehoonlee
committed
Update the RNN cell API to be explicit about output_size. (keras-team#11021 )

qlzh727
authored and
fchollet
committed
Revise the performance table for applications (keras-team#11085 )

taehoonlee
authored and
fchollet
committed
Update variational_autoencoder_deconv.py (keras-team#11081 )

linjinjin123
authored and
fchollet
committed
Skipped the training tests when the layer has the same behavior at training and testing time. (keras-team#11089 )

gabrieldemarmiesse
authored and
fchollet
committed
Refactor Enqueuers (keras-team#11079 )

Frédéric Branchaud-Charron
authored and
fchollet
committed
Section for Training history visualization (keras-team#11076 )

alfasst
authored and
fchollet
committed
Made an abstract class for upsampling. (keras-team#11065 )

gabrieldemarmiesse
authored and
fchollet
committed

Commits on Sep 6, 2018

Add missing raise for ValueError at Embedding compute_output_shape (keras-team#11091 )

yanboliang
authored and
fchollet
committed

Commits on Sep 10, 2018

Fix TF Function accepts session_kwargs: options and run_metadata (keras-team#11107 )

yanboliang
authored and
fchollet
committed

Commits on Sep 11, 2018

Commits on Sep 12, 2018

Simplified the cntk backend tests and made them more readable. (keras-team#11116 )

gabrieldemarmiesse
authored and
fchollet
committed
Fix bug when channel=1 (keras-team#11123 )

Frédéric Branchaud-Charron
authored and
fchollet
committed

Commits on Sep 13, 2018

Commits on Sep 14, 2018

Added the exponential activation. (keras-team#11136 )

gabrieldemarmiesse
authored and
farizrahman4u
committed
Breaking down the attention API PR: part 2 (keras-team#11140 )

gabrieldemarmiesse
authored and
Frédéric Branchaud-Charron
committed
Fixed the windows line endings in the CSVCallback. (keras-team#11124 )

gabrieldemarmiesse
authored and
fchollet
committed

Commits on Sep 15, 2018

Used a with statement instead of a close statement for robustness. (keras-team#11147 )

gabrieldemarmiesse
authored and
fchollet
committed

Commits on Sep 17, 2018

Fix bug in dilated conv for CNTK (keras-team#11125 )

yanboliang
authored and
gabrieldemarmiesse
committed
Style fixes for enabling PEP8 501 (keras-team#11160 )

taehoonlee
authored and
gabrieldemarmiesse
committed
Grouped common function calls together in the test_callbacks.py. (keras-team#11167 )

gabrieldemarmiesse
authored and
Frédéric Branchaud-Charron
committed
Grouped conda installations together to speed up the travis build. (keras-team#11166 )

gabrieldemarmiesse
authored and
Frédéric Branchaud-Charron
committed
Style fixes for enabling PEP8 501 (keras-team#11161 )

taehoonlee
authored and
fchollet
committed
Reduce BN test flakiness with seeding.
fchollet
committed
Splitted the convolutional recurrent tests. (keras-team#11154 )

gabrieldemarmiesse
authored and
fchollet
committed

Commits on Sep 20, 2018

Fix incorrect minimum input size for applications (keras-team#11181 )

taehoonlee
authored and
gabrieldemarmiesse
committed
Removed completely the keras_test decorator. (keras-team#11182 )

gabrieldemarmiesse
authored and
Frédéric Branchaud-Charron
committed
Refactoring: Added a data_generator to the test_utils.py. (keras-team#11153 )

gabrieldemarmiesse
authored and
fchollet
committed

Commits on Sep 21, 2018

Update “why use Keras?” page in the docs.
fchollet
committed

Commits on Sep 23, 2018

Style fixes for enabling PEP8 501 (keras-team#11199 )

taehoonlee
authored and
gabrieldemarmiesse
committed

Commits on Sep 24, 2018

Fix bug in sparse_top_k_categorical_accuracy (keras-team#11196 )

yanboliang
authored and
gabrieldemarmiesse
committed

Commits on Sep 25, 2018

refactor lagacy/interfaces (keras-team#11219 )

knightXun
authored and
gabrieldemarmiesse
committed
Added np.tile to reference_operations.py. (keras-team#11205 )

gabrieldemarmiesse
authored and
fchollet
committed

Commits on Sep 28, 2018

Commits on Sep 29, 2018

Add dilated_conv into reference operations (keras-team#11244 )

taehoonlee
authored and
gabrieldemarmiesse
committed
add trig funcs for numpy backend (keras-team#11254 )

danFromTelAviv
authored and
gabrieldemarmiesse
committed

Commits on Sep 30, 2018

add ndim (keras-team#11257 )

danFromTelAviv
authored and
gabrieldemarmiesse
committed
add more variable initializers to numpy backend (keras-team#11261 )

danFromTelAviv
authored and
gabrieldemarmiesse
committed

Commits on Oct 1, 2018

Commits on Oct 2, 2018

DOC: Add Conv3DTranspose class to autogen.py. (keras-team#11271 )

heathhenley
authored and
fchollet
committed
Gracefully handle legacy Sequential configs (keras-team#11280 )

GabrielBianconi
authored and
fchollet
committed

Commits on Oct 3, 2018

Add conv_transpose into reference operations (keras-team#11258 )

taehoonlee
authored and
gabrieldemarmiesse
committed
Fix fit_generator for workers=0 (keras-team#11285 )

arquolo
authored and
Frédéric Branchaud-Charron
committed
Bug fix: model save when file already exists (keras-team#11289 )

farizrahman4u
authored and
fchollet
committed
Prepare 2.2.4 release.
fchollet
committed

Commits on Oct 4, 2018

Commits on Oct 7, 2018

fix the broken format of links and references (keras-team#11319 )

p16i
authored and
gabrieldemarmiesse
committed
DOC: Closes keras-team#11306 multi_gpu_model examples do not show properly (keras-team#11310 )

Cheukting
authored and
gabrieldemarmiesse
committed

Commits on Oct 8, 2018

Commits on Oct 9, 2018