2017 12 20

gongweibao

• cloud:
  • https://github.com/PaddlePaddle/cloud/pull/526
  • https://github.com/PaddlePaddle/cloud/pull/530
• cluster:
  • Deploy PaddleCloud on k8s-k40 cluster
• scheduler:
  • Survey: https://github.com/PaddlePaddle/cloud/issues/523#issuecomment-352240114

Yu Yang

WIP

• NTM with attention
  • https://github.com/PaddlePaddle/Paddle/pull/6789
• Rename XXDescBind --> XXDesc
  • https://github.com/PaddlePaddle/Paddle/pull/6797

Gradient Clipping

• https://github.com/PaddlePaddle/Paddle/pull/6729

Design Of switching kernel

• https://github.com/PaddlePaddle/Paddle/pull/6720

Enhancement

• https://github.com/PaddlePaddle/Paddle/pull/6771
• https://github.com/PaddlePaddle/Paddle/pull/6761
• https://github.com/PaddlePaddle/Paddle/pull/6673
• https://github.com/PaddlePaddle/Paddle/pull/6655
• https://github.com/PaddlePaddle/Paddle/pull/6653
• https://github.com/PaddlePaddle/Paddle/pull/6642

Yan Xu

• Add a simple threadpool: https://github.com/PaddlePaddle/Paddle/pull/6684
• Muilti CPU design: https://github.com/PaddlePaddle/Paddle/pull/6223
• Review:
  • Multi Node: https://github.com/PaddlePaddle/Paddle/pull/6297
  • Doc refine: https://github.com/PaddlePaddle/Paddle/pull/6674#pullrequestreview-84022202
  • https://github.com/PaddlePaddle/Paddle/pull/6537#pullrequestreview-83992105

tonyyang-svail

• [WIP] Parallel_do: https://github.com/PaddlePaddle/Paddle/pull/6730
• Issue:
  • https://github.com/PaddlePaddle/Paddle/issues/6798
  • https://github.com/PaddlePaddle/Paddle/issues/6728
• Review
  • https://github.com/PaddlePaddle/Paddle/pull/6732

qijun

Multi-device:

• add design doc on keys of operaror kernel type
  • https://github.com/PaddlePaddle/Paddle/pull/6782
• Discussion on integrating MKLDNN to fluid
  • https://github.com/PaddlePaddle/Paddle/issues/6815

Fix and Enhance

• add GetPlaces operator
  • https://github.com/PaddlePaddle/Paddle/pull/6732
• add more place test and rename Cudnn to CUDNN
  • https://github.com/PaddlePaddle/Paddle/pull/6621

wangmeng

• Add separable convolution:
  • https://github.com/PaddlePaddle/Paddle/pull/6631
• Add Xception model for image classification:
  • https://github.com/PaddlePaddle/models/pull/555

caoying

• Add documentation for C-API (have not passed all of reviewer's comments yet):
  • https://github.com/PaddlePaddle/Paddle/pull/6596
• Refine the documentation for fully connected layer's helper and mul_op:
  • https://github.com/PaddlePaddle/Paddle/pull/6792
• Some small fix:
  • https://github.com/PaddlePaddle/models/pull/547
  • https://github.com/PaddlePaddle/models/pull/550
  • https://github.com/PaddlePaddle/models/pull/552

kavyasrinet

• PR review:
  • https://github.com/PaddlePaddle/Paddle/pull/6640
  • https://github.com/PaddlePaddle/Paddle/pull/6688
  • https://github.com/PaddlePaddle/Paddle/pull/6806
  • https://github.com/PaddlePaddle/Paddle/pull/6810
  • https://github.com/PaddlePaddle/Paddle/pull/6808
  • https://github.com/PaddlePaddle/Paddle/pull/6854
• Set up and build PaddlePaddle for IOS in Xcode.
• Understand and write inference in Fluid: https://github.com/PaddlePaddle/Paddle/pull/6672
• Design doc for executor: https://github.com/PaddlePaddle/Paddle/pull/6721
• Collaborate with Yiqun and QingQing to understand current C-API and inference.
• Design Operator Markdown: https://github.com/PaddlePaddle/Paddle/pull/6805
• Polish and write documentations for operators in layers.py (issue: https://github.com/PaddlePaddle/Paddle/issues/6526)
  • https://github.com/PaddlePaddle/Paddle/pull/6817
  • https://github.com/PaddlePaddle/Paddle/pull/6807
  • https://github.com/PaddlePaddle/Paddle/pull/6820
  • https://github.com/PaddlePaddle/Paddle/pull/6853

abhinavarora

• Documentation of Layers [WIP]

  • Polishing the documentation of the less than layer https://github.com/PaddlePaddle/Paddle/pull/6816
  • Polish layer documentation for fill_constant ops https://github.com/PaddlePaddle/Paddle/pull/6808
  • Polishing the embedding layer and the fc layer documentation https://github.com/PaddlePaddle/Paddle/pull/6806
  • Polish the support new device doc https://github.com/PaddlePaddle/Paddle/pull/6594

• PR Review

  • https://github.com/PaddlePaddle/Paddle/pull/6799#pullrequestreview-84972870
  • https://github.com/PaddlePaddle/Paddle/pull/6817
  • https://github.com/PaddlePaddle/Paddle/pull/6810#pullrequestreview-84962168
  • https://github.com/PaddlePaddle/Paddle/pull/6807#pullrequestreview-84950293
  • https://github.com/PaddlePaddle/Paddle/pull/6805#pullrequestreview-84914464
  • https://github.com/PaddlePaddle/Paddle/pull/6792#pullrequestreview-84862892

fengjiayi

• refactoring of backward code [WIP]
  • https://github.com/PaddlePaddle/Paddle/pull/6614
  • https://github.com/PaddlePaddle/Paddle/pull/6623
  • [WIP] https://github.com/PaddlePaddle/Paddle/pull/6741
• reviews:
  • https://github.com/PaddlePaddle/Paddle/pull/6655
  • https://github.com/PaddlePaddle/Paddle/pull/6771

Dang Qingqing

• Profiling:
  • Add design documentation for profiling tool: https://github.com/PaddlePaddle/Paddle/pull/6700
  • Add the event to record the timeline: https://github.com/PaddlePaddle/Paddle/pull/6701
  • Refine CUDA profiler and delete the test file: https://github.com/PaddlePaddle/Paddle/pull/6715
  • VGG Net and RNN profiling (TODO): https://github.com/PaddlePaddle/Paddle/issues/6736
• Memory optimization for OCR model
  • Reduce memory usage in conv layer and RoI layer for mobile inference.
    • https://github.com/PaddlePaddle/Paddle/pull/6659
    • https://github.com/PaddlePaddle/Paddle/pull/6666
    • The result: http://agroup.baidu.com/paddlepaddle/md/article/586596
  • Memory optimization in convolution layer by the grouped im2col and gemm.
    • https://github.com/PaddlePaddle/Paddle/pull/6802
• Code Review:
  • Correct the dropout_op's computation in test #6804
  • feature/print op #6799

wanghaoshuang

• Doc
  • https://github.com/PaddlePaddle/Paddle/pull/6708
  • https://github.com/PaddlePaddle/Paddle/pull/6810
• Models regtest
  • https://github.com/PaddlePaddle/regtest/pull/8
  • https://github.com/PaddlePaddle/regtest/pull/9
• Review
  • https://github.com/PaddlePaddle/Paddle/pull/6670
  • https://github.com/PaddlePaddle/Paddle/pull/6279
  • https://github.com/PaddlePaddle/regtest/pull/10
  • https://github.com/PaddlePaddle/regtest/pull/4

hedaoyuan

• Review
  • [Memory optimization] https://github.com/PaddlePaddle/Paddle/pull/6659
  • https://github.com/PaddlePaddle/Mobile/pull/66
  • [Multi-Thread] https://github.com/PaddlePaddle/Paddle/pull/6751
  • [C-API] https://github.com/PaddlePaddle/Paddle/pull/6596

luotao

• MKL
  • [discuess] how to add MKL Packed python interface: https://github.com/PaddlePaddle/Paddle/issues/6612
  • refine mkl&openblas script for benchmark: https://github.com/PaddlePaddle/Paddle/pull/6706
  • review: MKL Packed design doc and code for RNN optimazation
    • https://github.com/PaddlePaddle/Paddle/pull/6636, https://github.com/PaddlePaddle/Paddle/pull/6680 ,https://github.com/PaddlePaddle/Paddle/pull/6719
    • use Intel OpenMP to speedup seq2batch when WITH_MKL: https://github.com/PaddlePaddle/Paddle/pull/6622
• docs:
  • update the link of doc.paddlepaddle.org in README.md: https://github.com/PaddlePaddle/Paddle/pull/6618, https://github.com/PaddlePaddle/Paddle/pull/6734
  • add fluid python api in chinese website: https://github.com/PaddlePaddle/Paddle/pull/6663
  • reopen linkchecker: https://github.com/PaddlePaddle/Paddle/pull/6748
  • [WIP] add python doc for sequence_pool: https://github.com/PaddlePaddle/Paddle/pull/6787
  • review: k8s related, mobile, c-api docs
    • [USER] Add refer code to get started: https://github.com/PaddlePaddle/Paddle/pull/6646
    • https://github.com/PaddlePaddle/Paddle/pull/6674, https://github.com/PaddlePaddle/Paddle/pull/6742, https://github.com/PaddlePaddle/Paddle/pull/6711, https://github.com/PaddlePaddle/Mobile/pull/63, https://github.com/PaddlePaddle/Mobile/pull/62, https://github.com/PaddlePaddle/Paddle/pull/6596

Chenxi

• distributed training discussion
• add ios cross compiling doc in en
  • https://github.com/PaddlePaddle/Paddle/pull/6640

Yan Chunwei

• fluid
  • WIP chapter 8 machine translation decode
  • print op
• model CI with @haoshuang
  • design and plan
  • old code clean
• visualDL with @longfei @daming
  • move repo to PaddlePaddle/
  • https://github.com/VisualDL/VisualDL/pull/52
  • https://github.com/PaddlePaddle/VisualDL/pull/3
• PR
  • https://github.com/PaddlePaddle/Paddle/pull/6744/files/927f5d567c737fc9e754f911a1ced3def77b0651
  • https://github.com/PaddlePaddle/models/pull/534#pullrequestreview-82481975
  • https://github.com/PaddlePaddle/regtest/pull/8#pullrequestreview-84398256
  • https://github.com/PaddlePaddle/regtest/pull/9#pullrequestreview-84693303
  • https://github.com/PaddlePaddle/regtest/pull/8#pullrequestreview-84694853

ranqiu

• Update doc of v2 api

  https://github.com/PaddlePaddle/Paddle/pull/6654

• performance validation of understand_sentiment in fluid

Liu Yiqun

• Framework
  • Read some documentation of fluid
  • Review
    • Auto pruning: https://github.com/PaddlePaddle/Paddle/pull/2603
    • add doc for how to use C-API: https://github.com/PaddlePaddle/Paddle/pull/6596
• Mobile
  • Refine the inference demo
    • https://github.com/PaddlePaddle/Mobile/pull/64
  • Refine the content of many README and add some Chinese README:
    • https://github.com/PaddlePaddle/Mobile/pull/65
  • Android AICamera Demo
    • https://github.com/PaddlePaddle/Mobile/pull/61

wanghaox

• Port MultiBox Loss (Doing)
  • https://github.com/PaddlePaddle/Paddle/issues/6818

Daming Lu

• Setup dev servers (Flask and NodeJS) for VisualDL
• Research on how TensorBoard API works on retrieving data for Image and Scalar
  • https://github.com/VisualDL/VisualDL/issues/53
• Write Demo / Tutorial for VisualDL
  • https://github.com/VisualDL/VisualDL/issues/54
• Table in Markdown is not rendered, fixing
  • https://github.com/PaddlePaddle/PaddlePaddle.org/issues/360

guosheng

• PR:
  • Add Python wrapper for reduceSum, reduceMean
    • https://github.com/PaddlePaddle/Paddle/pull/6716
    • https://github.com/PaddlePaddle/Paddle/pull/6779
  • Refine and merge the Inception-ResNet-v2 model
    • https://github.com/PaddlePaddle/models/pull/510
• Review:
  • https://github.com/PaddlePaddle/models/pull/539

qiaolongfei

Fluid

• Muiti Device

  • Expose const value
  • Kernel hint design
  • Add kernel hint
  • how to add kernel hint

• Code optimize

  • fix simple_gru2 doc

• Review

  • feature/print op
  • Support Clip in param_attr
  • Add design of switching kernel
  • Add python wrapper for lstm unit op.
  • Enable reshape_op to support dimension inference
  • Add ChunkEvaluator for Multi-batches

zhaochengduo

PR

• Refine accuracy_op.cu
  • https://github.com/PaddlePaddle/Paddle/pull/6774
• Refine elementwiseAdd and im2col
  • https://github.com/PaddlePaddle/Paddle/pull/6743
• Add get parameters directly
  • https://github.com/PaddlePaddle/Paddle/pull/6687
• Add dilation for conv_trans_op
  • https://github.com/PaddlePaddle/Paddle/pull/6279

Benchmark

• Summary of performance optimization：02.Recognize_digits (WIP)
  • https://github.com/dzhwinter/benchmark/issues/27
• Fix mnist
  • https://github.com/dzhwinter/benchmark/pull/24
• fix nvprof gpu consistence
  • https://github.com/dzhwinter/benchmark/pull/15
• 07.Semantic Role Labeling

Review

• Refine CUDA profiler and delete the test file
  • https://github.com/PaddlePaddle/Paddle/pull/6715

yangyaming

• Expose sequence_softmax_op
  https://github.com/PaddlePaddle/Paddle/pull/6726
• Expose sequence_expand_op
  https://github.com/PaddlePaddle/Paddle/pull/6670
• Expose lstm unit
  https://github.com/PaddlePaddle/Paddle/pull/6669
• seq2seq [WIP]
  https://github.com/dzhwinter/benchmark/pull/25
• Model update
  https://github.com/PaddlePaddle/DeepSpeech/pull/88
  https://github.com/PaddlePaddle/DeepSpeech/pull/90
• Code Review
  https://github.com/PaddlePaddle/Paddle/pull/6789#pullrequestreview-84958477
  https://github.com/PaddlePaddle/DeepSpeech/pull/92#pullrequestreview-84956016
  https://github.com/PaddlePaddle/Paddle/pull/6716#pullrequestreview-84671460
  https://github.com/PaddlePaddle/Paddle/pull/6708#pullrequestreview-84062907
  https://github.com/PaddlePaddle/Paddle/pull/6663#pullrequestreview-83794521
  https://github.com/PaddlePaddle/DeepSpeech/pull/68#pullrequestreview-83739165
  https://github.com/PaddlePaddle/Paddle/pull/6538#pullrequestreview-84057565

sweetsky0901(tiantian)

• spp(Spatial pyramid pooling) op done
  • https://github.com/PaddlePaddle/Paddle/pull/6204
• detection_output op（for SSD, doing, code review 
  • https://github.com/PaddlePaddle/Paddle/pull/6488
• norm op doing unittest
  • https://github.com/PaddlePaddle/Paddle/issues/6561

dongzhihong

• Fluid
  • remove GPU Sync Interface
    • https://github.com/PaddlePaddle/Paddle/pull/6793
  • change OperatorBase Run with Place
    • https://github.com/PaddlePaddle/Paddle/pull/6783
  • fix Place inherit
    • https://github.com/PaddlePaddle/Paddle/pull/6766
  • cudnn operators change to cudnn kernel
    • https://github.com/PaddlePaddle/Paddle/pull/6660
  • add global device pool
    • https://github.com/PaddlePaddle/Paddle/pull/6537
  • polish executor design doc
    • https://github.com/PaddlePaddle/Paddle/pull/6688
• Reviews
  • add kernel key
    • https://github.com/PaddlePaddle/Paddle/pull/6720

Yibing Liu

Deep Speech 2:

• Finish the BaiduEN8K model's training with lr decay + clip by global norm
  • Update benchmark result for BaiduEN8K model
  • Current benchmark has caught up with DS2 paper on some public test sets
• review:
  • https://github.com/PaddlePaddle/DeepSpeech/pull/90
  • https://github.com/PaddlePaddle/DeepSpeech/pull/88

Fluid:

• VGG16 convergence comparison on GPU between fluid and tensorflow
  • https://github.com/PaddlePaddle/Paddle/issues/6813
• VGG16 performance comparison between fluid and tensorflow
  • https://github.com/PaddlePaddle/Paddle/issues/6814
• [merged] Correct the dropout_op's computation in test
  • https://github.com/PaddlePaddle/Paddle/pull/6804
• [merged] Enable reshape_op to support dimension inference
  • https://github.com/PaddlePaddle/Paddle/pull/6609

Zhangchao

• Fix crf_layer, crf_decoding, chunk_evaluator when using GPU.

  https://github.com/PaddlePaddle/Paddle/pull/6522

• Refine the DOC of DSSM.

  https://github.com/PaddlePaddle/models/pull/554

• Book.04 word2vec performance comparison with V2.

  https://github.com/PaddlePaddle/Paddle/issues/6710

helinwang

• Concurrency discussions
• Multiple Machine Review: https://github.com/PaddlePaddle/Paddle/pull/6297#pullrequestreview-84871025
• Multiple Thread Review: https://github.com/PaddlePaddle/Paddle/pull/6684#pullrequestreview-84945672
• Paddle Cloud Review: https://github.com/PaddlePaddle/cloud/pull/526#pullrequestreview-83673564

typhoonzero(wuyi)

• Cluster train transpiler updating:
  • https://github.com/PaddlePaddle/Paddle/pull/6297
• Train DeepFM on kubernetes cluster
• PR reviews of Paddlecloud, thread pool and some design doc
• benchmarking 05 recommender

sidgoyal78

• WIP: Updating documentation of layers: https://github.com/PaddlePaddle/Paddle/issues/6526
• Related PR reviews:
  • https://github.com/PaddlePaddle/Paddle/pull/6807
  • https://github.com/PaddlePaddle/Paddle/pull/6806