2017 11 01

wangkuiyi

• Save/load models in Python: https://github.com/PaddlePaddle/Paddle/pull/5136#pullrequestreview-72352311
• Translate Build for Android document: https://github.com/PaddlePaddle/Paddle/pull/5029
• Translate Build for Raspberry Pi: https://github.com/PaddlePaddle/Paddle/pull/5177
• Unify Android build scripts: https://github.com/PaddlePaddle/Paddle/issues/5169
• Restructure mobile and cross-compiling documents: https://github.com/PaddlePaddle/Paddle/issues/5168
• Rename "C-API" into "inference library": https://github.com/PaddlePaddle/Paddle/issues/5175
• Update CONTRIBUTE.md and VLOG convention: https://github.com/PaddlePaddle/Paddle/pull/5181
• gflag's Python bind: https://github.com/PaddlePaddle/Paddle/pull/5193#pullrequestreview-72707181
• program API change: https://github.com/PaddlePaddle/Paddle/pull/5260#pullrequestreview-73317048

luotao

• update benchmark data on VGG19: https://github.com/PaddlePaddle/Paddle/pull/5148
• add Max strategy for sequence_pool op: https://github.com/PaddlePaddle/Paddle/pull/4864
• add test_Expand and simply the gserver/tests/CMakeLists: https://github.com/PaddlePaddle/Paddle/pull/5044
• correct the index of cluster_train_cn/en.md: https://github.com/PaddlePaddle/Paddle/pull/5249
• prepare for "Create with AI" meeting.
• code review:
  • MKLDNN:
    • refine check macro: https://github.com/PaddlePaddle/Paddle/pull/5154
    • refine mkldnn unit test: https://github.com/PaddlePaddle/Paddle/pull/5273
  • other:
    • Contribute and logging: https://github.com/PaddlePaddle/Paddle/pull/5181#pullrequestreview-72660558
    • nlp-dqa program on code.

srinetkavya

• Design documentation
  • Model Averaging design doc (https://github.com/PaddlePaddle/Paddle/pull/5137)
  • Background for multi-device training and inference(https://github.com/PaddlePaddle/Paddle/pull/5284). This is the first design doc for FPGA inference.
• Pull request review of raspberry pi documentation.

kexinzhao

• Design documentation
  • float16 design doc (https://github.com/PaddlePaddle/Paddle/pull/5313)
• Op Porting
  • python wrapper for adagrad (https://github.com/PaddlePaddle/Paddle/pull/4977)
• Code review
  • https://github.com/PaddlePaddle/Paddle/pull/5272#pullrequestreview-73654380
  • https://github.com/PaddlePaddle/Paddle/pull/5021#pullrequestreview-71589116

ChengduoZhao

• Op Porting
  • Add conv3d_gemm_op.(can be merged)
    • https://github.com/PaddlePaddle/Paddle/pull/4709
  • Add 3D convolution transposed operator.(can be merged)
    • https://github.com/PaddlePaddle/Paddle/pull/5118
• fix bug and code format refine
  • fix pool op (If globalPooling = true, ksize and paddings should be ignored.)
    • https://github.com/PaddlePaddle/Paddle/pull/5187
  • fix_sequence_conv_op(code format refine)
    • https://github.com/PaddlePaddle/Paddle/pull/5130
• Code review
  • deconv cudnn forward passed
    • https://github.com/PaddlePaddle/Paddle/pull/5235
  • Fix/sequence conv
    • https://github.com/PaddlePaddle/Paddle/pull/5266

nickyfantasy

• Start a web app with Django for PaddleBoard Demo https://github.com/PaddlePaddle/board/pull/11
• Integrate with D3.js and able to show line and bar chart in a web app with hardcode data https://github.com/PaddlePaddle/board/pull/12
• Update and finalize how TensorFlow writes and read data https://github.com/PaddlePaddle/board/wiki/How-TensorFlow-writes-data-and-read-by-TensorBoard

varunarora

• PEP-8 partial consistency and function level commenting of codebase: https://github.com/PaddlePaddle/PaddlePaddle.org/pull/220
• Cross-browser and cross-platform testing of site, font and color updates: https://github.com/PaddlePaddle/PaddlePaddle.org/issues/227
• Restyling blog and consolidating styles with rest of content: https://github.com/PaddlePaddle/PaddlePaddle.org/pull/230
• Mobile repo transform and sitemap generation support: https://github.com/PaddlePaddle/PaddlePaddle.org/pull/240
• Assisted with blog repo build/deploy

cs2be (thuan)

• Auto-generation of Paddle docs navigations. Dynamic generation of top level navigations: https://github.com/PaddlePaddle/PaddlePaddle.org/pull/216
• Merge search partials, merge sitemap template files, fix issue with paddle doc 0.10.0 generator: https://github.com/PaddlePaddle/PaddlePaddle.org/pull/231
• Add PaddlePaddle.org documentation mode: https://github.com/PaddlePaddle/PaddlePaddle.org/pull/246

daming-lu

• Added stripping and deploying script for blog:

  https://github.com/PaddlePaddle/PaddlePaddle.org/pull/224/

• Merged one-line deploy script to original Paddle content repos:

  Blog: https://github.com/PaddlePaddle/blog/pull/8

  Book: https://github.com/PaddlePaddle/book/pull/439

  Models: https://github.com/PaddlePaddle/models/pull/408

  Paddle: https://github.com/PaddlePaddle/Paddle/pull/5287

• (WIP) add one-line deploy script for Mobile: https://github.com/PaddlePaddle/PaddlePaddle.org/issues/243

abhinavarora

Regularization Design

• Add Python API for backward regularization ops https://github.com/PaddlePaddle/Paddle/pull/5135
• Adding the Sign Op for L1 Weight Decay Regularization https://github.com/PaddlePaddle/Paddle/pull/5138
• Implementing L1 regularization Python wrapper https://github.com/PaddlePaddle/Paddle/pull/5173
• Adding L2 Regularization to Recognize digits MLP example https://github.com/PaddlePaddle/Paddle/pull/5186

Variable Initialization

• Adding a python framework for variable initializers https://github.com/PaddlePaddle/Paddle/pull/5232
• Adding Normal distribution initializer and unit tests for python initializers https://github.com/PaddlePaddle/Paddle/pull/5256
• Adding Xavier Initialization https://github.com/PaddlePaddle/Paddle/pull/5270

Code Review

• https://github.com/PaddlePaddle/Paddle/pull/5084/#pullrequestreview-72372202
• https://github.com/PaddlePaddle/Paddle/pull/5275#pullrequestreview-73636722
• https://github.com/PaddlePaddle/Paddle/pull/5268#pullrequestreview-73362923
• https://github.com/PaddlePaddle/Paddle/pull/5216#pullrequestreview-73002965
• https://github.com/PaddlePaddle/Paddle/pull/5098#pullrequestreview-72029934

jetfuel (Jeff Wang)

• Translate the quick start to Chinese. https://github.com/PaddlePaddle/PaddlePaddle.org/pull/210
• Generate Book sitemap by parsing index.cn.html.json and index.en.html.json https://github.com/PaddlePaddle/PaddlePaddle.org/pull/212
• Version Selection Dropdown menu sorting and UI improvement. https://github.com/PaddlePaddle/PaddlePaddle.org/pull/238, https://github.com/PaddlePaddle/PaddlePaddle.org/pull/235
• End to end deployment testing https://github.com/PaddlePaddle/PaddlePaddle.org/pull/239, https://github.com/PaddlePaddle/PaddlePaddle.org/pull/232
• Update the algorithm on language switching. Properly associate the related files https://github.com/PaddlePaddle/PaddlePaddle.org/pull/219, https://github.com/PaddlePaddle/PaddlePaddle.org/pull/225, https://github.com/PaddlePaddle/PaddlePaddle.org/pull/221, https://github.com/PaddlePaddle/PaddlePaddle.org/pull/229

gongweibao

• Log received parameter config: https://github.com/PaddlePaddle/Paddle/pull/5162
• Run PaddleCloud on minikube: https://github.com/PaddlePaddle/cloud/pull/463
• Fix some bugs: https://github.com/PaddlePaddle/cloud/pull/459 https://github.com/PaddlePaddle/Paddle/pull/5074 https://github.com/PaddlePaddle/cloud/pull/452

CaoYing

PR review:

• BeamSearch operator:
  • https://github.com/PaddlePaddle/Paddle/pull/5052#pullrequestreview-72662075
• The tensor product operator:
  • https://github.com/PaddlePaddle/Paddle/pull/5014#pullrequestreview-72496819
• The scene text recognition example to model repo:
  • https://github.com/PaddlePaddle/models/pull/329#pullrequestreview-72497921
• Bug fix for LambdaRank:
  • https://github.com/PaddlePaddle/models/pull/413#pullrequestreview-73675103

PR:

• Merge CRF operator: https://github.com/PaddlePaddle/Paddle/pull/5084

tonyyang-svail

• static rnn with executor:
  • https://github.com/PaddlePaddle/Paddle/pull/5224
• operators:
  • https://github.com/PaddlePaddle/Paddle/pull/5222
  • https://github.com/PaddlePaddle/Paddle/pull/5286
• VLOG:
  • https://github.com/PaddlePaddle/Paddle/pull/5121

wanghaoshuang

• OP:
  • Merge sequence expand op
    • https://github.com/PaddlePaddle/Paddle/pull/4740
• Review:
  • https://github.com/PaddlePaddle/Paddle/pull/5016
  • https://github.com/PaddlePaddle/Paddle/pull/5076

Dang Qingqing

• LSTM Bakcward Implenmentaion [merged]
  • https://github.com/PaddlePaddle/Paddle/pull/5115
• Allocate aligned memory by posix_memalign [merged]
  • https://github.com/PaddlePaddle/Paddle/pull/5247
• Refine activation function pointer for LSTM operator.
  • https://github.com/PaddlePaddle/Paddle/pull/5251
• Code Review:
  • Hack auc for dense vector labels: https://github.com/PaddlePaddle/Paddle/pull/5274
  • GRU Op: https://github.com/PaddlePaddle/Paddle/pull/5255
  • Fix Pooling Op: https://github.com/PaddlePaddle/Paddle/pull/5187
  • Seqeunce Conv Op: https://github.com/PaddlePaddle/Paddle/pull/4814
  • fix_sequence_conv_op: https://github.com/PaddlePaddle/Paddle/pull/5130
  • WarpCTC Op: https://github.com/PaddlePaddle/Paddle/pull/5107
  • Chunk Evalutor Op: https://github.com/PaddlePaddle/Paddle/pull/5016
  • CRF Op: https://github.com/PaddlePaddle/Paddle/pull/5084

typhoonzero(wuyi)

• autoscaling experiment: https://github.com/PaddlePaddle/cloud/issues/399
• Some paddle/fault tolernet job bug fixes:
  • https://github.com/PaddlePaddle/Paddle/pull/5274
  • https://github.com/PaddlePaddle/Paddle/pull/5160
  • https://github.com/PaddlePaddle/Paddle/pull/5132
  • https://github.com/PaddlePaddle/Paddle/pull/5083
  • https://github.com/PaddlePaddle/Paddle/pull/4972
  • https://github.com/PaddlePaddle/Paddle/pull/4828
  • https://github.com/PaddlePaddle/Paddle/pull/5111
• paddle native HDFS reader: https://github.com/PaddlePaddle/Paddle/pull/5282
• op porting refining and reviews:
  • https://github.com/PaddlePaddle/Paddle/pull/5111
  • https://github.com/PaddlePaddle/Paddle/pull/5208

fengjiayi

• compelete Python API of model save/load:
  • https://github.com/PaddlePaddle/Paddle/pull/5020
• conv net of Book Chap6
  • https://github.com/PaddlePaddle/Paddle/pull/5258
• fixes:
  • https://github.com/PaddlePaddle/Paddle/pull/5201
  • https://github.com/PaddlePaddle/Paddle/pull/5221
  • https://github.com/PaddlePaddle/Paddle/pull/5257

ranqiu

• Fix bugs of and Refine DSSM

  https://github.com/PaddlePaddle/models/pull/407 https://github.com/PaddlePaddle/models/pull/409

• Update ConvS2S

  https://github.com/ranqiu92/models/commit/4b428e561b20c98acaeb1414838fad8c55e528d6

Peng Li

• PR
  • https://github.com/PaddlePaddle/models/pull/403 : add pre-trained models -> merged
  • https://github.com/PaddlePaddle/Paddle/pull/5276 : enhance test cases for gaussian_random_op and uniform_random_op
• code review:
  • https://github.com/PaddlePaddle/Paddle/pull/5084 : CRF op
  • https://github.com/PaddlePaddle/Paddle/pull/5137 : design doc for model average
  • https://github.com/PaddlePaddle/Paddle/pull/5270 : Xavier Initializer

Chen xi

• Auto-scaling Experiment https://github.com/PaddlePaddle/cloud/issues/399 https://github.com/PaddlePaddle/cloud/pull/458

• Experiment data visualization https://github.com/PaddlePaddle/cloud/pull/453

• Auto scaling PR article follow up

• Contact "cheap learning" paper author

guosheng

• OP Porting:
  • Add GRU Operator: https://github.com/PaddlePaddle/Paddle/pull/5255
• Review:
  • https://github.com/PaddlePaddle/Paddle/pull/5251

qijun

Paddle Book:

• refine square_error_cost layer
  • https://github.com/PaddlePaddle/Paddle/pull/5216
• Refine word2vec demo
  • https://github.com/PaddlePaddle/Paddle/pull/5268

SelectedRows:

• support sparse output for lookup table grad op
  • https://github.com/PaddlePaddle/Paddle/pull/5145
• support adagrad sparse update
  • https://github.com/PaddlePaddle/Paddle/pull/5272
• fix bug in lookup table grad operator
  • https://github.com/PaddlePaddle/Paddle/pull/5228

Fix and Enhance:

• remove unused code
  • https://github.com/PaddlePaddle/Paddle/pull/5219
• add init_gflags interface
  • https://github.com/PaddlePaddle/Paddle/pull/5193

Liu Yiqun

• Arm deployment
  • [Reviewing] Enable the build for iOS simulator
    • https://github.com/PaddlePaddle/Paddle/pull/5211
  • [Merged] Add documenttation of cross-compiling for iOS
    • https://github.com/PaddlePaddle/Paddle/pull/5239
  • [Updated] Remove duplicated linking of libraries
    • https://github.com/PaddlePaddle/Paddle/pull/4890
  • Learn the ssd model and try to do the inference using capi on linux
  • Review
    • Add merge model scripts: https://github.com/PaddlePaddle/Paddle/pull/5151

yangyaming

• DS2
  • Remove padding before computing loss and gradients (WIP)
    https://github.com/PaddlePaddle/models/issues/414
• OP porting
  • Merge huber loss op
    https://github.com/PaddlePaddle/Paddle/pull/3989
    https://github.com/PaddlePaddle/Paddle/issues/5281
  • Merge precision recall op
    https://github.com/PaddlePaddle/Paddle/pull/5111

dongzhihong

• Refactorization:
  • book chapter5 recommender system
    • https://github.com/PaddlePaddle/Paddle/pull/5143
  • add sequence conv layer
    • https://github.com/PaddlePaddle/Paddle/pull/5117
  • polish sequence pool
    • https://github.com/PaddlePaddle/Paddle/pull/5229
  • sequence conv operator fix
    • https://github.com/PaddlePaddle/Paddle/pull/5264
  • network visualizer
    • https://github.com/PaddlePaddle/Paddle/pull/5292
  • polish embedding table
    • https://github.com/PaddlePaddle/Paddle/pull/5140
  • design evaluator for network
• Reviews
  • regularizer design doc
    • https://github.com/PaddlePaddle/Paddle/pull/5105
  • some fixs and enhance
    • https://github.com/PaddlePaddle/Paddle/pull/5228
    • https://github.com/PaddlePaddle/Paddle/pull/5298
    • https://github.com/PaddlePaddle/Paddle/pull/5142
    • https://github.com/PaddlePaddle/Paddle/pull/5290

wangmeng

• Finish factorization machine layer
  • https://github.com/PaddlePaddle/Paddle/pull/4859
• Add plot to file
  • https://github.com/PaddlePaddle/Paddle/pull/5283

Yibing Liu

Deep Speech 2:

• Enable the clipping of gradient by local & global norm
  • https://github.com/PaddlePaddle/Paddle/pull/4970
• Follow the sequence padding bug
• Tiny fix
  • https://github.com/PaddlePaddle/models/pull/383

Operator Porting:

• Add ctc edit distance operator
  • https://github.com/PaddlePaddle/Paddle/pull/5300

helinwang

• Autoscaling experiment:
  • https://github.com/PaddlePaddle/cloud/issues/399
  • https://github.com/PaddlePaddle/cloud/pull/453
• TeamCity CI will not test if only *.md and *.rst changed.

Zhang Chao

• Fix lambda_rank bug for inference:

  https://github.com/PaddlePaddle/models/pull/413

• Merge scene text recognition:

  https://github.com/PaddlePaddle/models/pull/329

• [WIP] Update the bilinear tensor product for batch input:

  https://github.com/PaddlePaddle/Paddle/pull/5014

qiaolongfei

refactoring

• Image classification in new C++ core:
  • update create_op_func and support generate dropout layer
  • 03 image classification: VGG16
  • 03 image classification: add resnet
  • improve unique_name, uniq id is related to prefix
• Code Improve && code clean
  • memory log level change from 3 to 10
  • add shareLod
  • Optimizer use init program

code review

• Optimizer Design and Implementation
  • Handling global step increment in optimizer python wrapper
• Image classification in new C++ core:
  • Batch norm latest
• Other
  • Remove test_mnist, since we replace it with compile time concepts
  • Fix clang compile
  • Extract InferShape to many cc files
  • Extract apply_backward_pass to backward.py
  • deconv cudnn forward passed

hedaoyuan

• Mobile
  • https://github.com/PaddlePaddle/Paddle/pull/5252
  • https://github.com/PaddlePaddle/Mobile/pull/21
• Review
  • https://github.com/PaddlePaddle/Mobile/pull/20

HuWeiwei

• porting clip_by_norm op
• https://github.com/PaddlePaddle/Paddle/pull/5296

Zhuoyuan Chen

Op implementation:

• Batch norm test and some fixes with Long fei: https://github.com/PaddlePaddle/Paddle/pull/5103
• Deconv Op on Cudnn: https://github.com/PaddlePaddle/Paddle/pull/5235

Op review:

• LSTM fix for Acknowledgements: https://github.com/PaddlePaddle/Paddle/pull/5293
• Conv 3d Transpose (Chengduo): https://github.com/PaddlePaddle/Paddle/pull/5118

Working and next step:

• Work with Longfei on why demo3 (image classification does not produce correct result)

xingzhaolong

• pr
  • https://github.com/PaddlePaddle/Paddle/pull/5165 add dilation for exconv
  • https://github.com/PaddlePaddle/Paddle/pull/5151 add merge model scripts
• Enet
  • upsample layer (done, wait pr)
  • dilation (done)
  • reproduce on paddle(done)
• review
  • https://github.com/PaddlePaddle/Mobile/pull/21

zhouxiao

• Operators
  • PR(under review): finished pnpair evaluator, add several optional features https://github.com/PaddlePaddle/Paddle/pull/5108
  • Doing:
    • layer normalization based on new BN operator. https://github.com/PaddlePaddle/Paddle/issues/4701

Yu Yang

• Cast Operator
  • https://github.com/PaddlePaddle/Paddle/pull/5149
• Save/Load tensor in python
  • https://github.com/PaddlePaddle/Paddle/pull/5136
• Static RNN with Executor
  • with tony
  • https://github.com/PaddlePaddle/Paddle/pull/5224
• Code clean up
  • Extract InferShape from operator.h
    • https://github.com/PaddlePaddle/Paddle/pull/5174
• Fix several bugs
  • backward https://github.com/PaddlePaddle/Paddle/pull/5262/files
  • Clang Compile https://github.com/PaddlePaddle/Paddle/pull/5217
  • AddBiasOp should take care of flatten_dims https://github.com/PaddlePaddle/Paddle/pull/5200

Yan Chunwei

• PR
  • closed, dynamic rnn test with RNNOp
  • open, rank sort lod op

xuwei

• https://github.com/PaddlePaddle/Paddle/issues/5289