172 Commits (dfdcb7eaa8550595e93b0212d8f9d9504bd31641)

Author	SHA1	Message	Date
Luo Tao	16b70f3ef5	reduce time of test_TrainerOnePass	8 years ago
liaogang	d1e754336e	Merge branch 'clang-format' of https://github.com/gangliao/Paddle; branch 'develop' of https://github.com/PaddlePaddle/Paddle into clang-format	8 years ago
Helin Wang	f0cf8ac6f0	reduce compare sparse test time by reducing dimensions	8 years ago
liaogang	736bc95b09	Fix proto_data	8 years ago
liaogang	c5714728eb	add blank	8 years ago
liaogang	b626a7ccd2	change suffix proto to proto_data	8 years ago
tensor-tang	3c3a11a0dc	add use_mkldnn flag	8 years ago
liaogang	1d4fa2433c	ClangFormat for proto and cuda	8 years ago
武毅	39af255959	Fix new optimizer lr (#3074) ... * default learning rate, temperary fix * update	8 years ago
武毅	23b8346072	Fault tolerant distributed training, just work version, with etcd (#2849) ... * using etcd as fault tolerant training * update * workable version, ft not tested * small fix * update * remove TODO	8 years ago
hedaoyuan	010adb99b4	Remove useless empty pointer check.	8 years ago
wuyi05	26d95a6bbf	fix new remote updater for go pserver	8 years ago
Yu Yang	5eb8bf0324	Correct GLOG CHECK in Paddle ... Use CHECK instead of PCHECK, because PCHECK is used for errno.	8 years ago
helinwang	fae606fc96	Merge pull request #2659 from helinwang/cmake ... Improve Go cmake, add go/master/c into cmake include path..	8 years ago
Helin Wang	59cf5e7796	Fix Go cmake	8 years ago
Yu Yang	64b78b1656	Fix TravisCI	8 years ago
Yu Yang	b1a311c44d	Add pb_cc_library in generic.cmake ... Fix #2567	8 years ago
Cao Ying	6fa84d40f4	Merge pull request #2521 from emailweixu/fix_rnn_gen ... Fix bugs for rnn generation.	8 years ago
qiaolongfei	9e13b68f01	refine code	8 years ago
qiaolongfei	d9aac1e13d	add WITH_Go to disable compile go to paddle	8 years ago
xuwei06	3438d650ed	Fix bugs for rnn generation ... 1. v2.layer.parse_network does not correctly handle the generation output. 2. GatherAgentLayer does not correctly handle generation output when batch_size > 1 3. Fix CustomStackTrace for rnn group	8 years ago
xuwei06	17994e38aa	RecurrentGroup with mixed input sequence types ... No longer need to use SubsequenceInput. The framework will detect.	8 years ago
qiaolongfei	c093a24336	clang format check	8 years ago
qiaolongfei	ebba2b139b	update code with new cclient	8 years ago
qiaolongfei	37594eae17	add paramConfig for each parameter	8 years ago
qiaolongfei	da3e84a6d2	change trainer_id	8 years ago
qiaolongfei	4f366be4e2	clang format	8 years ago
qiaolongfei	39d0b3de99	add test file mnist_test.py, free resource of newRemoteParameterUpdater properly	8 years ago
qiaolongfei	c44f5dd883	add simple updater, this version can train a model	8 years ago
qiaolongfei	966bf9ae1f	fix the problem in cclient when malloc paddle_parameter	8 years ago
qiaolongfei	28476f5f6e	fix the problem of paddle_send_grad	8 years ago
qiaolongfei	6f1c91da99	refine code	8 years ago
qiaolongfei	99dc60642d	new parameterupdater use paddle pserver cclient of go	8 years ago
Yu Yang	273e3f444f	Remove not necessary functionalities in Parameter	8 years ago
liaogang	6237f6f57a	revert clang-format	8 years ago
liaogang	f27fd9dc28	follow comments	8 years ago
Yu Yang	966e89cd68	Disable some unit tests when PYTHON or SWIG_PY are not enabled. ... * When SWIG_PY is not enabled, paddle.v2 package cannot be used. * When PYTHON is not enabled, PyDataProvider cannot be used. Fix issue #1920	8 years ago
Luo Tao	7a81327c28	remove compile option "with_metric_learning"	8 years ago
Helin Wang	3219c831a1	Rename Argument::sumCost to Argument::cost since Argument should not know about cost. ... cost is Argument, but argument does not have to be cost.	8 years ago
wangkuiyi	90268e1258	Merge pull request #1222 from reyoung/feature/remove_unused_function_in_trainer ... Remove unused init method in paddle::Trainer	8 years ago
wangkuiyi	dac4a59793	Merge pull request #1223 from reyoung/feature/remove_unused_clean_method ... Removed unused GradientMachine::create	8 years ago
Yu Yang	5b1f66fe97	Removed unused GradientMachine::create	8 years ago
Yu Yang	1f97236177	Remove unused init method in paddle::Trainer	8 years ago
backyes	b6dfd2891b	fix bug: ... * update function can not be inherited since updateImpl exsits * remove update implementation in one inherited updater	8 years ago
qiaolongfei	aa9f516260	code refine, add comment and some naming problem	8 years ago
qiaolongfei	b1eeb2e1d7	Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into add-pserver-util	8 years ago
Yu Yang	d41ac09090	Merge pull request #1017 from gangliao/external ... Download, build, install External dependencies via cmake	8 years ago
qiaolongfei	5aaaef4468	rename PServerController to ParameterServerController	8 years ago
liaogang	be8b126845	Move Execepts into arch/osx dir	8 years ago
liaogang	b090ce329a	Fix conflict with develop	8 years ago
liaogang	8acd1ac31a	Add extern python interp	8 years ago
qiaolongfei	93e74f89c6	rename PServerUtil to PServerController	8 years ago
qiaolongfei	cfbb4c481e	use unique_ptr	8 years ago
qiaolongfei	95f20b9472	add comment and refine code	8 years ago
qiaolongfei	f3c61cbc4c	add pserver util and parameter server config	8 years ago
qiaolongfei	8ff42cd37d	merge develop	8 years ago
qiaolongfei	32b28c6429	add remote updater in api and swig	8 years ago
qiaolongfei	027aaf9ef2	add cluster train for quick_start	8 years ago
Yu Yang	adc58397d2	Merge pull request #969 from reyoung/feature/clean_gradient_machine_start ... Remove not used params in GradientMachine::start	8 years ago
Yu Yang	4e34220852	Merge pull request #970 from reyoung/feature/clean_parameter_updater_finish_pass ... Remove unused cost parameter in ParameterUpdater	8 years ago
Yu Yang	677c79b6a1	Merge branch 'feature/clean_parameter_updater_finish_pass' into feature/mnist_train_api	8 years ago
Yu Yang	71a316ea1f	Remove unused cost parameter in ParameterUpdater	8 years ago
Yu Yang	ad6cb60d74	Merge branch 'feature/clean_gradient_machine_start' into feature/mnist_train_api	8 years ago
Yu Yang	56f29658ba	Remove not used params in GradientMachine::start	8 years ago
Yu Yang	0d1703d91f	Add const in ParameterUpdater init	8 years ago
Yu Yang	7aad9f539e	Merge branch 'develop' of github.com:baidu/Paddle into feature/add_unittest	8 years ago
Yu Yang	96eab5046a	Complete unittest setting in CMake	8 years ago
Yu Yang	04fb1fc13e	Merge pull request #891 from backyes/sparse_bug ... [Sparse Bug] Test and sparse_remote_update can not co-exsit, crash trainer if necessary	8 years ago
Yu Yang	22b9b6662b	Add unittest to coverage SgdThreadUpdater's enableBufType	8 years ago
wangyanfei01	7c7430470d	follow comments: more readable LOG	8 years ago
wangyanfei01	45bd0a5e16	more accurate to early stop train sparse model	8 years ago
liaogang	3d0e73bd32	Remove custom glog-like and gflags-like macros	8 years ago
wangyanfei01	2294b819cc	Merge remote branch 'origin/develop' into sparse_bug	8 years ago
wangyanfei01	f53e8d7a09	fix bug: if test and sparse_remote_update can not co-exsit, crash trainer if necessary	8 years ago
Yu Yang	068bfbb817	All file pass pre-commit hook	8 years ago
Yi Wang	e9549cbb78	Change "Baidu, Inc" into "PaddlePaddle Authors"	8 years ago
Yu Yang	ba68704edb	Fix api test, fix testing script for macos	8 years ago
Yu Yang	3100406cc4	Simplify the testOnePeriod method.	8 years ago
wangyanfei01	1f2423a919	follow comments: more docs	8 years ago
wangyanfei01	290ee32f2d	fix confliction	8 years ago
wangyanfei01	63de3ec2ba	merge origin/develop and fix confliction	8 years ago
Luo Tao	7830893a83	fix conflict	8 years ago
xuwei06	66520f8b41	Fix --job=test ... When the define_py_data_sources2 has both train_list and test_list, for job=test, the trainer will create both dataProvider_ and testDataProvider_. But dataProvider_ is not used. This causes SIGSEGV at finishAsync() because asyncLoader_ is not created. Change-Id: If579f715f80a70ebc795094792c3436bfa0f5746	8 years ago
wangyanfei01	1f743d381c	Redesign test_period meaning: ... * always do test on all test data * do test at the end of each pass if test_period=0, otherwise do test if test_period batches passed	8 years ago
Luo Tao	96615fe329	merge develop, fix conflict	8 years ago
Luo Tao	80c68d38ff	clang format .cc .h .cpp .c and .hpp file	8 years ago
Luo Tao	496d64ebdb	Support rectangle input for CNN	8 years ago
Yu Yang	836d61382f	Update pre-commit-config ... * Check all files by pre commit hooks	8 years ago
qijun	319742c641	format python code in demo, doc, doc_cn and paddle directories	8 years ago
wangyanfei01	b62c80f156	qfg	8 years ago
wangyanfei01	0feecbd13c	modify on docs	8 years ago
wangyanfei01	0a0c55d228	more friendly test options	8 years ago
wangyanfei01	c6a0298e2a	create PR to polish test_period meaning	8 years ago
Yu Yang	968464cc60	Fix a bug in testOnePeriod. (#322) ... * Forget to finishTestPeriod in testOnePeriod. * Fix #318	8 years ago
qingqing01	45c81a414f	Add job=time in trainer, refine cudnn_conv to reduce gpu memory and speed up training. (#218) ... * Add benchmark for PaddlePaddle, tensorflow and caffe * ConvProjection to reduce memory for goolenet * Add unit test for ConvProjection. 1. unit test in test_LayerGrad. 2. compare the ConvPorjection and CudnnConvLayer, also compare the concat_layer+img_conv_layer and concat_layer_conv_projection. * Reduce cudnn_conv memory and add benchmark document. 1. Use TmpMatrix as the workspace in cudnn_conv to reduce gpu memory. It reduce lots of memory. 2. Add benchmark document. 3. fix smallnet_mnist_cifar.py in paddle. * Add job=time and refine cudnn_conv to reduce gpu memroy and speed up * Refine cudnn_conv and shared biases operation in concat_layer and mixed_layer. * follow comments * follow comments * Use unique_ptr to prevent memory leaks in CudnnConvLayer.	8 years ago
luotao1	ca5a5ec480	Make Paddle --save_dir support a directory name (#277) ... * Also fix #243	8 years ago
backyes	0e1a22d0fc	set test_period default value to 0 (#279)	8 years ago
emailweixu	cbe734b396	Python trainer api (#193) ... * Python trainer API and demo * Adding missing PaddleAPIPrivate.h * Adding api_train.sh * More comments * Bump up patch version to 0b3	8 years ago
luotao1	e20ff3b3e9	fix build bug in gcc46 (#236)	8 years ago
emailweixu	28bc05b126	Fix sparse training for trainer_count=1 (#204) ... * Fix sparse training for trainer_count=1 For trainer_count=1, the gradient machine is NeuralNetwork, which does not create parameter buf for PARAMETER_GRADIENT for sparse update in Parameter::enableType. But gradient parameter buf is still used in SgdThreadUpdater. * Minor update to comment	8 years ago