2017 08 23

wangkuiyi

Key concepts structuring

• https://github.com/PaddlePaddle/Paddle/issues/3570#issuecomment-323665697

Code polishment

• https://github.com/PaddlePaddle/Paddle/issues/3622
• https://github.com/PaddlePaddle/Paddle/pull/3581

Add new op

• https://github.com/PaddlePaddle/Paddle/pull/3620#pullrequestreview-57853960

Yu Yang

Draft design doc of Python Package

• https://github.com/PaddlePaddle/Paddle/pull/3569

New Operator

• scale, identity https://github.com/PaddlePaddle/Paddle/pull/3592
• minus https://github.com/PaddlePaddle/Paddle/pull/3634

Enhancement of current codebase:

• Resolve a cycle depenedence https://github.com/PaddlePaddle/Paddle/pull/3587
• Rename Net::AddOp --> Net::AppendOp https://github.com/PaddlePaddle/Paddle/pull/3585

Tiny fixes:

• https://github.com/PaddlePaddle/Paddle/pull/3556
• https://github.com/PaddlePaddle/Paddle/pull/3553

caoying

• GNR model:
  • add the last layer for GNR model: cross entropy over beams with early stop:
    • https://github.com/PaddlePaddle/Paddle/issues/3638
    • https://github.com/PaddlePaddle/Paddle/pull/3639
  • follow comments to update sequence_slice_layer:
    • https://github.com/PaddlePaddle/Paddle/pull/3367
  • update GNR models:
    • https://github.com/lcy-seso/models/tree/GNR/globally_normalized_reader
  • all of the implementations have been finished, now I begin to focus on tuning the model.

qiaolongfei

refactoring

• complete mnist_demo:
  • https://github.com/PaddlePaddle/Paddle/pull/3564
• fix sgd_op:
  • https://github.com/PaddlePaddle/Paddle/commit/8b3d33a055b2a1556adedeb41a16b794249a3848
• fix rowwise_add_grad_op:
  • https://github.com/PaddlePaddle/Paddle/commit/82b820e97b90f21d7b46629bba72436a69e888e1
• reduce some compile warning:
  • https://github.com/PaddlePaddle/Paddle/pull/3645
• Fix rowwise add:
  • https://github.com/PaddlePaddle/Paddle/pull/3604
• many discussion about python api and the work for next Phase:
  • PaddlePaddle Refactoring: Phase 3
  • Port Operators

Yan Chunwei

• paddle refactor

  • discussion about the design of python syntax wrapper with others
    • https://github.com/PaddlePaddle/Paddle/pull/3566
    • https://github.com/PaddlePaddle/Paddle/pull/3626
    • https://github.com/PaddlePaddle/Paddle/pull/3646
  • discussion about refactor tasks

• paddle models

  • fix DSSM bug about GRU

typhoonzero(wuyi)

• large model bug:
  • https://github.com/PaddlePaddle/Paddle/pull/3518 (merged)
  • Test on dnn model (Baidu Feed, WIP)
• paddle cloud:
  • bug fix reviews
    • https://github.com/PaddlePaddle/cloud/pull/333
    • https://github.com/PaddlePaddle/Paddle/pull/3624
    • https://github.com/PaddlePaddle/Paddle/pull/3649
    • https://github.com/PaddlePaddle/Paddle/pull/3568
  • Fault tolerant demos debug-[WIP]
• Clusters:
  • CPU Cluster: Changed to another 144 nodes, reinstalling OS.
  • DLNEL Cluster: GPU resource not releasing bug-[WIP], feature requests.

wanghaoshuang

• Image:

  • Make cudnn convolution layer and projection support for dilation:
    • https://github.com/PaddlePaddle/Paddle/pull/3629

• PaddleCloud:

  • Run recognize_digits and understand_sentiment demo with fault tolerant mode:
    • https://github.com/PaddlePaddle/cloud/pull/343

fengjiayi

• Discuss Python API design
• code review
  • https://github.com/PaddlePaddle/Paddle/pull/3540
  • https://github.com/PaddlePaddle/Paddle/pull/3587
  • https://github.com/PaddlePaddle/Paddle/pull/3645

qijun

• dynet api survey
  • https://github.com/PaddlePaddle/Paddle/pull/3562
• refine random related operators
  • https://github.com/PaddlePaddle/Paddle/pull/3596
• Add with refactoring cmake option
  • https://github.com/PaddlePaddle/Paddle/pull/3631

Code Reviews:

• https://github.com/PaddlePaddle/Paddle/pull/3620#discussion_r134664308
• https://github.com/PaddlePaddle/Paddle/pull/3592#discussion_r134376072

hedaoyuan

• Mobile
  • https://github.com/PaddlePaddle/Paddle/pull/3549
• Review
  • https://github.com/PaddlePaddle/Paddle/pull/3602
  • https://github.com/PaddlePaddle/Paddle/pull/3591

Dang Qingqing

• PaddlePaddle Refactoring

  • Add a lookup table op and a CUDA helper
    • https://github.com/PaddlePaddle/Paddle/pull/3620
  • Merge GPU kernel for cross entropy: https://github.com/PaddlePaddle/Paddle/pull/3501
  • Move pybind from package paddle/framework into paddle/pybind:
    • https://github.com/PaddlePaddle/Paddle/pull/3640
  • code review:
    • https://github.com/PaddlePaddle/Paddle/pull/3634

• Other:

  • Tune relative precision for unit test img_conv2 in test_NetworkCompare:
    • https://github.com/PaddlePaddle/Paddle/pull/3591
  • Code Review:
    • dilation conv: https://github.com/PaddlePaddle/Paddle/pull/3629
    • ROIPooling: https://github.com/PaddlePaddle/Paddle/pull/2982
    • Faster-RCNN Related: https://github.com/PaddlePaddle/Paddle/pull/3496
    • 3DConv/Deconv:
      • https://github.com/PaddlePaddle/Paddle/pull/3348#pullrequestreview-55203521
      • https://github.com/PaddlePaddle/Paddle/pull/3348#pullrequestreview-57995587
    • 3DPooling:
      • https://github.com/PaddlePaddle/Paddle/pull/3504#pullrequestreview-57753344

dongzhihong

• Refactorization: discuss about the Python API with guys.
  • Clarify the concept of Block, Network and run target interface.
  • discuss how initialized op and forward/backward/optimize op should be organized. survey the NCCL library
• PR & fixes unify operator Input/Output name: design doc. https://github.com/PaddlePaddle/Paddle/pull/3452 add unsigned value in attr https://github.com/PaddlePaddle/Paddle/pull/3623 fix ndcg metrics https://github.com/PaddlePaddle/models/pull/215
• Reviews init minst demo https://github.com/PaddlePaddle/Paddle/pull/3564
  fix rowwise add https://github.com/PaddlePaddle/Paddle/pull/3604 Gather operator https://github.com/PaddlePaddle/Paddle/pull/3540 refine random operator https://github.com/PaddlePaddle/Paddle/pull/3596 fix cudnn path https://github.com/PaddlePaddle/Paddle/pull/3581

Xinghai Sun

• DS2:
  • Update network structure (Add GRU support) https://github.com/PaddlePaddle/models/pull/214
  • Tuning bigger model: WER for Librispeech -1.7%
• RL:
  • Implement A3C with PyTorch (As a preparation for its paddle version)

Chen xi

• Made training plan and documented the training process
  • https://docs.google.com/document/d/1T5DTnNPv_f7VeARJj8PYMUsksZfJd0pPB8AsenYq35U/edit?usp=sharing
  • http://wiki.baidu.com/display/~chenxi44/Start+Coding+for+Paddle
• Received training from Helin
  • Big picture understanding about Paddlepaddle
  • Paddle cloud architect
• Design doc review and update

Yancey1989(yanxu)

• DLNEL production system maintaining
  • https://github.com/PaddlePaddle/cloud/issues/345
  • https://github.com/PaddlePaddle/cloud/issues/344
• PaddleCloud
  • Repo: Custom Kubernetes Scheduler Demo, https://github.com/Yancey1989/k8s-scheduler
  • PR: GPU limit, https://github.com/PaddlePaddle/cloud/pull/326
  • issue: improve logs storage/view, https://github.com/PaddlePaddle/cloud/issues/335
  • issue: Prevent data leaks, https://github.com/PaddlePaddle/cloud/issues/332

guosheng

• PaddlePaddle Image Enhancement
  • Faster R-CNN: refine the ROIPoolLayer.
• Cooperation with Video Analysis Team in IDL:
  • Add ScaleShiftLayer to get rid of differences with the Pytorch model. https://github.com/PaddlePaddle/Paddle/pull/3560
  • Debug the Attention Net and validate the forward precession. Additionally, another network used in the ensemble model has gotten the same evaluating result with the Pytorch implementation.
• Code Review
  • https://github.com/PaddlePaddle/Paddle/pull/3571
• Other
  • fix: https://github.com/PaddlePaddle/Paddle/pull/3602

luotao

• [WIP] refine Huber loss, add huber_regression_cost: https://github.com/PaddlePaddle/Paddle/pull/3571
• [WIP] V2 API save and load param header: https://github.com/PaddlePaddle/Paddle/pull/3619
• [bug fix] help @yanchunwei to fix dssm infer error with gru: https://github.com/PaddlePaddle/models/pull/212
• code review:
  • Add a sequence slice layer: https://github.com/PaddlePaddle/Paddle/pull/3367#pullrequestreview-57111394
  • Remove flag use_mkldnn_wgt (Intel MKLDNN): https://github.com/PaddlePaddle/Paddle/pull/3548

yangyaming

• DS2
  • Train unicom data on paddle cloud
    • Learn paddle cloud
    • Prepare environment
    • Submit job
  • Enhance decoder with @yibing
    • Support character based LM
    • Support vocabulary filtering
    • https://github.com/PaddlePaddle/models/pull/139

chengduoZhao

• Adaptedto the BatchNorm structure to support 3D data
  • https://github.com/PaddlePaddle/Paddle/pull/3594
  • https://github.com/PaddlePaddle/Paddle/issues/3597
• Add config parser for Conv3D and DeConv3D
  • https://github.com/PaddlePaddle/Paddle/pull/3594
• fix 3D-maxPooling function
  • https://github.com/PaddlePaddle/Paddle/pull/3504

Liu Yiqun

• Cooperate with gemone for the Album project
  • Profile Paddle, and look into why Paddle is slower than Caffe
  • build Paddle for armeabi
  • build Paddle with clang

Yibing Liu

• Optimize the beam search decoder for deployment

  • https://github.com/PaddlePaddle/models/pull/139

  • Simplify the build process

  • Add the support of multiple threads, reducing evaluation time to 10min

  • Choose better data structure (trie tree) and decoding logic (finite state transducer)

• Carry on some experiments on changing the text data preprocessing

helinwang

• Refactor discussion (Python interface)
• Paddle Cloud test demo converge, fixes:
  • https://github.com/PaddlePaddle/Paddle/pull/3649
  • https://github.com/PaddlePaddle/Paddle/pull/3624
• Bring Chen Xi up to date

zhuoyuan

• Refactor discussion (GAN and Python API)
• Paddle operators
  • gather_op, forward backward
  • scatter_op, forward backward
• fix gradient check for in-place operators