* Implement the operator with sprase matrix multiply
* Update the URL of mklml library.
test=develop
* Disable MKLML implematation when using no-linux.
test=develop
* optimize bp with mkl sparse matrix
test=develop
* add pybind interface to get all inplace ops, test=develop
* enhance OpTest to check whether the consistency of operator when using and not using inplace, test=develop
* handle corner cases in op_test, test=develop
* support outputs without tensor holder_, like XShape in reshape_op, test=develop
* fix bug, some op has GradOpMaker, but actually no grad_op in OpInfoMap, test=develop
* use reshape_grad instead of reshape in FlattenGradOp, test=develop
* fix error debug dims info for variables like XShape, test=develop
* change computational order in sum_op to relieve computation difference using inplace, test=develop
* add inplace_atol to check group_norm, and skip inplace_grad for mkldnn, test=develop
* follow sneaxiy's comments, test=develop
* remove unused DefaultGradOpDescMaker in mkldnn op, test=develop
* increase test_parallel_executor_seresnext time limit
test=develop
* split test_parallel_executor_seresnext
test=develop
* temporally disable reduce_and_allreduce test because of the random failure.
test=develop
* split gpu and cpu
test=develop
add fl_listen_and_serv op for Federated_learning and fl_distribute_transpiler add this op to pserver program . This op just listen the endpoint and sum&scale.
* change the default value of summarize from -1 to 20 in Print op to improve ease of use, test=develop
* change the doc of API Print to make the document easier to understand, test=develop
* instag lod tensor impl
* First PR for instag
* First PR for instag
* Before adding Selection Rows.
* Change name from instag to filter_instag, add upgrade the impl of filter_instag
* Change name from instag to filter_instag, add upgrade the impl of filter_instag
* Fix yapf error in gradient_checker.py to pass Travis-CI
* Fix Filter Instag Grad test=develop
* Fix Filter Instag Grad test=develop
* 1) Fix API.spec, add filter_instag Op. 2) Add Vector Support for CUDA. test=develop
* Impl Loss_weight and empty output handler
* change Loss Weight datatype to Float32, and add Loss Weight as 2nd output
* 1) Support Tensor Input(without LOD) 2) Add Unit test
* Filter By Instag Final test=develop
* Update API.spec for filter_by_instag test=develop
* Update API.spec for filter_by_instag 2 test=develop
* Add Filter By Instag Coverage
* code format of test_layers.py
* code format test_layers.py test=develop
* Make API args more readable test=develop
* Make API args more readable and pass code format test=develop
* Filter By Instag Op, Rename Map to Index Map test=develop
* Filter By Instag Op, code format err in filter_by_instag_op.cc test=develop
* Filter by instag op: code format of cpp files test=develop
* Filter by instag Op: Api spec modification test=develop
* Filter by instag Op: Api spec doc id modification test=develop
* Filter by instag Op: Api spec and doc preview test=develop test=document_preview
* Filter By Instag Op, fix doc erro test=document_preview test=develop
* Filter By Instag Op, fix doc err and Api spec test=document_preview test=develop
* Filter By Instag Op, fix Api spec test=document_preview test=develop
* Filter By Instag Op, fix Paddle Encoforce deprecated warning test=document_preview test=develop
* Filter By Instag Op, fix Paddle Encoforce deprecated and code format warning test=document_preview test=develop
* add hard_swish activation op (new op)
test=develop
* remove redundancy files
* modify document content of HardSwish OP
* add API test in test_layers.py
* add dynamic_graph for test_hard_swish
* add a place field in DataFeed to denote which place it will feed data to.
* abstract the copy process in CopyToFeedTensor function
* add UT for float32 type and for CUDAPlace
* Add call stack info during runtime and compile time
test=develop
* Rename operator_call_stack
test=develop
* Add unit test
test=develop
* follow comment
test=develop
* add train demo for imdb text classification task
* make inference library release data_feed dataset dataset_factory data_feed_factory
* add String Data Generator
* new feature of train demo: save model params
* New feature of train demo: set training config using gflags
* change code style for CI
* add readme and dataset for imdb demo trainer
* fix warpctc.dll not found issue, test=develop
* revert the linux platform change, test=develop
* delete warpctc_lib_path.h.in, test=develop
* add SetPySitePackagePath function
* fix warpctc.dylib not found issue on Mac, test=develop
* improve the paddle lib path setting logic, test=develop
* fix mac ci issue caused by test_warpctc_op unittest, test=develop
* tweak code, test=develop
* open gc by default, test=develop
* fix test_train_recognize_digits and disable gc when ngraph is enabled, test=develop
* fix conditional_block op eager deletion bug, test=develop
* add some comments to reviewers, test=develop
* support filelist size < trainer num
* pull dense when stop, to make sure local dense params are same as pserver, so save paddle model will save dense model same as pserver
* enable QueueDataset train same filelist for serveral times
* test=develop
Add the op of unique_with_counts, the op is calc the unqiue input of data, and output the corresponding indices and count of data.
* test=develop
Check the input and dtype in the op of unique_with_counts
* test=develop
test=document_preview
update the API.spec for `unique_with_counts`, at the same time, optimize the python api in the op of `unique_with_count`
* test=develop
test=document_preview
Fix some python api problem in the op of `unique_with_counts`, and change the error messsage in this op.
* Fix some API problem in the op of `unique_with_counts`
test=develop
test=document_preview
* test=develop
test=document_preview
Fix the api sample of op `unique_with_counts`, and update api.spec
(1) set fleet_send_batch_num a default value according to trainer num， the previous 80000 is fixed，if trainer num is much less or larger than 100，global shuffle may have timeout error.
(2) fix load one table bug, add barrier
* support center loss
* change tensor copy api to high level api tensorcopy
* test=develop rewrite the center_loss cuda_kernel to make it faster
and add document of the center loss api,also update test function
* test=document_preview test=develop
update document of center loss
* test=document_preview test=develop
modify API.spec modify test code remove nouse const_cast
* extend matmul op to support multiple head multiplication
With the support of multiple head, the multiplication of two big matrixes is
split into multiplication of several (head_number) small matrixes. e.g. if
Mat A is [3, 24] and Mat B is [24, 4], when multiple A and B with head_number
as 4, Mat A will be split as 4 matrix of [3, 6] and Mat B will be 4 matrix of
[6, 4]. The result of final matrix will be 4 matrix of [3, 4], i.e. [3, 16].
The change includes 2 things:
1. save delta model and shrink table are control by the same parameter before, now add delete_after_unseen_days to control shrink table.
2. value in sparse table has no slot before, now add slot in sparse table, and add DownpureCtrAccessor to support the new meta.
test=develop
（1）support patch data （merge slots of instances of same line id, modify dense layer which
changes its size）
（2）add fleet load_one_table interface, support load from paddle model and load from pslib model
（3）fix push sparse bug which cause push sparse cost more time（about 10% in my testcase）
（4）when some slots are not in one of your network (join/update, etc.)，data feed、collect label info、push/pull sparse will skip these slots， instead of throw error.
（5）add more debug info in TrainFilesWithProfiler
The change includes 3 things:
1. Set CPU_NUM to 1 in the tests because the ParallelExecutor will print warning that CPU_NUM is not set and use default 1.
2. Old tests compare two RNNs, hand written simple RNN and same RNN built by Paddle, but initialized RNN weights in numpy random and Paddle random separately. Fixed it by setting weights and bias values.
3. Also set numpy random seed in the tests. Now the two RNNs diff can be smaller (rtol from 0.1, 0.2 to. 0.01) in the tests.
test=develop
Test PaddingRNN on V100 GPU device.
Test configuration: large model, padding mode (which is the mode using recurrentOp), one GPU.
GPU memory (MiB): 6414 (this PR) vs 6837 (without this PR)
Speed (steps/s): 10.28 (this PR) vs 9.89 (without this PR)
* feature/auto_growth_allocator, test=develop
* add unittest of AlignedAllocator, test=develop
* try to turn on auto_growth to test on CI, test=develop
* fix segmentation fault in mixed_vector.h, test=develop
* add unittests, test=develop
1. Since allreduce op has 4 reduce types, We split these four reduce types into four ops
2. We also refined the collective op code, e.g. we separated the collective op kernel into CPUKernel and CUDAKernel, and remove the device specified DeviceContext parameter in template as we already knew the target DeviceContext
3. We remove the newly added Collective op role to reduce the complexity of program and graph analysis
* fix prepare context redundant code problem, optimize executor by caching create_varaiables
test=develop
* supports collective training in executor
* make fetch_list runable with variables, add more unittest for use_program_cache
test=develop
* fix comment
test=develop
* use unique name for nccl_id
* supports output to stream in program_to_code
* insert sync_comm_stream before regularization; add skip_op_callstack capability in program_to_code
* set op role in collective training
* add collective op role
* remove orig file
* add build optimizer by strategy
* add collective strategy
* refine collective strategy
* add multi-process role maker
* refine strategy building factory so that we can easily plugin more strategy
* scale loss grad in collective sgd transpiler
* add support for distributed fc
* code format
* revert some features for dist fc
* add support for distributed fc training
* fix prepare context redundant code problem, optimize executor by caching create_varaiables
test=develop
* supports collective training in executor
* make fetch_list runable with variables, add more unittest for use_program_cache
test=develop
* use unique name for nccl_id
* supports output to stream in program_to_code
* insert sync_comm_stream before regularization; add skip_op_callstack capability in program_to_code
* set op role in collective training
* add collective op role
* fix comment
test=develop
* remove orig file
* add build optimizer by strategy
* add collective strategy
* refine collective strategy
* add multi-process role maker
* refine strategy building factory so that we can easily plugin more strategy
* scale loss grad in collective sgd transpiler
* add support for distributed fc
* code format
* revert some features for dist fc
* add support for distributed fc training
* test=develop
add collective op unittest standard
* test=develop
remove the test_collective directory
* test=develop
remove the test_collective directory
* remove slicegather test
* code format for reducescatter
* update attr of shard_index_op
* Modify macro nccl_helper
* remove test without distribute
* macro collective_helper
* marcro update
* test=develop
update support python3.5
* test=develop change gpu memory use to 0.1 when test
* test=develop
update ut equal func
* test=develop
set flags to 1.5
* test=develop fix pickle dumple py35
* test=develop
fix divide in slice and add sync_comm_stream
update atol and rtol to 1e-05
rm shard_index op and test
modify read input from file to read from memory
remove origin_program in framework and add i/o in c_sync_calc_stream
* test=develop update unittest sync operator I/O
1. fix the bug that out_put_var in SaveSelectedRows would be empty string
2. use merge_sparse_lookup_table to replace sum op for load_persistables_for_inference
3. fix the bug in _clone_var_in_block_ when the var is SELECTED_ROWS.
(1) use channel instead of vector/BlockingQueue in Dataset，to keep same with existing implementation, and make code more readable and flexible (dataset single output channel or multi output channel). one previous memory out of limit problem is cause by not release memory after training.
(2) add Record because MultiSlotType costs too much memory (80B)，fix memory out of limit problem.
(3) add Channel, Archive in paddle/fluid/framework
(4) change dataset from shared_ptr to unique_ptr in pybind
(5) move create/destroy readers from trainer to dataset
(6) move shuffle from datafeed to dataset. dataset holds memory, datafeed is only for load data and feed data to network.
(7) fix thread num bug of Dataset when filelist size < thread num
(8) support set_queue_num in InMemoryDataset
Zhen Wang
Leo Chen
Yibing Liu
chengduo
翟飞跃
Aurelius84
Zeng Jinle
lidanqing
xiaoting
danleifeng
Zhaolong Xing
lvmengsi
zhang wenhui
kh2se2013
gongweibao
silingtong123
Tao Luo
zhaoyuchen2018
Hao Wang
Adam
wuzewu
liym27
LielinJiang
jiaqi
Jiawei Wang
wawltor
huangjun12
wopeizl
yaoxuefeng
hutuxian
ShenLiang
Kaipeng Deng
mapingshuo
wangguanzhong
Youwei Song
Chen Weihang
SunGaofeng
Jiabin Yang
liuwei1031
Liufang Sang
Dong Daxiang
xsrobin
hong
石晓伟
whs
HaoRen
Thunderbrook
chengduo
fuyinno4
Bob Zhu
JesseyXujin
Yi Liu
Huihuang Zheng
tangwei12
123malin
tianshuo78520a
lujun
bingyanghuang
Physher
tensor-tang
qingqing01
pkpk
zhoukunsheng
Yi Liu
Jie Fang
Wojciech Uss
HaoRen
Qiao Longfei
flame
Bai Yifan
songhao