2017 11 08 - PaddlePaddle/Paddle GitHub Wiki

• Fix warnings https://github.com/PaddlePaddle/Paddle/pull/5359#pullrequestreview-74220912
• Too slow unit test https://github.com/paddlepaddle/paddle/issues/5342
• Compilation warnings https://github.com/paddlepaddle/paddle/issues/5341
• Compilation warnings https://github.com/paddlepaddle/paddle/issues/5340
• Compilation warnings https://github.com/paddlepaddle/paddle/issues/5339
• Compilation warnings https://github.com/paddlepaddle/paddle/issues/5338
• test_CustomStackTrace takes too much time https://github.com/paddlepaddle/paddle/issues/5330
• test_LayerGrad takes too much time to run https://github.com/paddlepaddle/paddle/issues/5329
• C/C++ compilation warnings about comparing signed and unsigned integers https://github.com/paddlepaddle/paddle/issues/5322
• {init_program,program} => {startup_program,main_program} https://github.com/paddlepaddle/paddle/issues/5319
• Pip install support https://github.com/paddlepaddle/paddle/issues/5318
• Unused parameters in framework.layers.py https://github.com/paddlepaddle/paddle/issues/5316
• Rename Python package layers into layer https://github.com/paddlepaddle/paddle/issues/5315
• Polish C++ operators documentation https://github.com/paddlepaddle/paddle/issues/5265
• Issues related with buildtools:
  • https://github.com/PaddlePaddle/buildtools/issues/3
  • https://github.com/PaddlePaddle/buildtools/issues/2
  • https://github.com/PaddlePaddle/buildtools/issues/1

• Edit documentation for all operators

  • https://github.com/PaddlePaddle/Paddle/pull/5374
  • https://github.com/PaddlePaddle/Paddle/pull/5373
  • https://github.com/PaddlePaddle/Paddle/pull/5372
  • https://github.com/PaddlePaddle/Paddle/pull/5317

• Write documentation for layers

  • Prototype: https://github.com/PaddlePaddle/Paddle/pull/5500

• Design documentation:

  • Background for multi-device: https://github.com/PaddlePaddle/Paddle/pull/5284
  • Explaining how multi-device works on Tensorflow: https://github.com/PaddlePaddle/Paddle/pull/5412

• Edit documentation for C++ operators

  • https://github.com/PaddlePaddle/Paddle/pull/5334
  • https://github.com/PaddlePaddle/Paddle/pull/5347
  • https://github.com/PaddlePaddle/Paddle/pull/5354
  • https://github.com/PaddlePaddle/Paddle/pull/5357
  • https://github.com/PaddlePaddle/Paddle/pull/5367
  • https://github.com/PaddlePaddle/Paddle/pull/5375
  • https://github.com/PaddlePaddle/Paddle/pull/5376
  • https://github.com/PaddlePaddle/Paddle/pull/5377

• Check that operators follow naming conventions

  • https://github.com/PaddlePaddle/Paddle/issues/5366

• Review

  • https://github.com/PaddlePaddle/Paddle/pull/5353#pullrequestreview-74180186
  • https://github.com/PaddlePaddle/Paddle/pull/5373#pullrequestreview-74261115
  • https://github.com/PaddlePaddle/Paddle/pull/5317#pullrequestreview-74261187

• Mobile
  • Build for minimum size https://github.com/PaddlePaddle/Mobile/pull/21
  • Inference Acceleration with Mobile GPU https://github.com/PaddlePaddle/Paddle/issues/5469
• Review
  • https://github.com/PaddlePaddle/Paddle/pull/5211
  • https://github.com/PaddlePaddle/Paddle/pull/5343
  • https://github.com/PaddlePaddle/Paddle/pull/5313
  • https://github.com/PaddlePaddle/Paddle/pull/5436
  • https://github.com/PaddlePaddle/Mobile/pull/24

• Define a customized simple protobuf format for PaddleBoard https://github.com/PaddlePaddle/board/pull/13
• Discussed with YongFeng and YiQun for a AICamera mobile demo (SSD Model) that will be shown in Baidu World https://github.com/PaddlePaddle/Mobile/issues/25
• Successfully compiled iOS and Android Paddle library https://github.com/PaddlePaddle/Paddle/blob/develop/doc/mobile/cross_compiling_for_ios_cn.md https://github.com/PaddlePaddle/Paddle/blob/develop/doc/mobile/cross_compiling_for_android_cn.md
• Initial check in iOS code with Paddle C++ code integration https://github.com/PaddlePaddle/Mobile/pull/26

• Contributed to building a development guide: https://github.com/PaddlePaddle/PaddlePaddle.org/pull/259
• Fixed quick start guide to make instructions clearer and for the code to reliably work with the current API: https://github.com/PaddlePaddle/PaddlePaddle.org/pull/260
• Built operators documentation and injected into documentation from JSON from print_operators_doc: https://github.com/PaddlePaddle/PaddlePaddle.org/pull/257
• Helped with fixing formula rendering on Chinese pages: https://github.com/PaddlePaddle/PaddlePaddle.org/pull/263
• Completed pending UI/UX proposal document: https://github.com/PaddlePaddle/PaddlePaddle.org/wiki/User-Research-(UI-UX)

• Autoscaling experiment and report: https://github.com/PaddlePaddle/cloud/blob/develop/doc/autoscale/experiment/README.md
• Reviews:
  • https://github.com/PaddlePaddle/cloud/pull/468#pullrequestreview-75298461
• Fixes:
  • https://github.com/PaddlePaddle/Paddle/pull/5497

• Add Mobile documentation to "DOC_MODE" https://github.com/PaddlePaddle/PaddlePaddle.org/pull/250
• Update sitemap generator for Mobile to pull in li>a and add them to side menu https://github.com/PaddlePaddle/PaddlePaddle.org/pull/254
• Add development guide https://github.com/PaddlePaddle/PaddlePaddle.org/pull/259
• Add manage.py method to update operators documentation. https://github.com/PaddlePaddle/PaddlePaddle.org/pull/261
• Fix line height issues on side navigation when text is too long https://github.com/PaddlePaddle/PaddlePaddle.org/pull/262
• Add document view only mode https://github.com/PaddlePaddle/PaddlePaddle.org/pull/269

• Fixed mal-format math formulas in book :
  • https://github.com/PaddlePaddle/PaddlePaddle.org/issues/256
• Added one-line deployment to documentation:
  • https://github.com/PaddlePaddle/Paddle/pull/5287
• Add cron job for PaddlePaddle.org's daily release (WIP)
  • https://github.com/PaddlePaddle/PaddlePaddle.org/issues/267

Working on Move core job management codes to controller in go

• https://github.com/PaddlePaddle/cloud/issues/462
• code branch is here: https://github.com/gongweibao/cloud/commits/movetogo

• Add convs2s to models

  https://github.com/PaddlePaddle/models/pull/431

• Update doc of layers.py

  https://github.com/PaddlePaddle/Paddle/pull/5437

  https://github.com/PaddlePaddle/Paddle/pull/5487

• discuss with Abhinav and longfei about beam search support
• update design of sequence decoder (beamsearch)
• beam search op

• PR
  • fix bugs of softmax with cross entropy:
    • https://github.com/PaddlePaddle/Paddle/pull/5394
  • CRF decoding:
    • https://github.com/PaddlePaddle/Paddle/pull/5352
  • fix NCE:
    • https://github.com/PaddlePaddle/Paddle/pull/5494
    • https://github.com/PaddlePaddle/models/pull/435
• PR review
  • The tensor operator:
    • https://github.com/PaddlePaddle/Paddle/pull/5014#pullrequestreview-74961291
    • https://github.com/PaddlePaddle/Paddle/pull/5014#pullrequestreview-75317130
  • The convolution seq2seq model:
    • https://github.com/PaddlePaddle/models/pull/431#pullrequestreview-75306024

• Refine bilinear tensor product op :

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

• Help to transform resnet from tf to paddle.

• [WIP] Refine doc for capi in models:

  https://github.com/PaddlePaddle/models/pull/385

• PR
  • https://github.com/PaddlePaddle/models/pull/415 : fix bug in document
• code review:
  • https://github.com/PaddlePaddle/Paddle/pull/4389
  • https://github.com/PaddlePaddle/Paddle/pull/4390
  • https://github.com/PaddlePaddle/Paddle/pull/5283

• Refine sequence max-pooling and add unit testing of gradient check.
  • https://github.com/PaddlePaddle/Paddle/pull/5310
• Check errors for the CUDA kernel calls.
  • https://github.com/PaddlePaddle/Paddle/pull/5436
• Enable hidden/cell state initialization and enhance unit testing in LSTM operator.
  • https://github.com/PaddlePaddle/Paddle/pull/5429
• Fix cmake error when building with WITH_AVX=OFF.
  • https://github.com/PaddlePaddle/Paddle/pull/5425
• Fix the doc for Momentum and Adam optimizer.
  • https://github.com/PaddlePaddle/Paddle/pull/5391
• [Speed up compiling]: reduce the NVCC compiling (some .cu operators can be compiled by G++)
  • https://github.com/PaddlePaddle/Paddle/issues/5491
• Code Review:
  • https://github.com/PaddlePaddle/Paddle/pull/5418
  • https://github.com/PaddlePaddle/Paddle/pull/5467
  • https://github.com/PaddlePaddle/Paddle/pull/5388

• Op porting

  • Add NCE op(WIP):
    • https://github.com/PaddlePaddle/Paddle/pull/5480

• code review

  • https://github.com/PaddlePaddle/Paddle/pull/5488
  • https://github.com/PaddlePaddle/Paddle/pull/5108

• reduce elasped time of test_LayerGrad: https://github.com/PaddlePaddle/Paddle/pull/5478
• add doc for image.py: https://github.com/PaddlePaddle/Paddle/pull/5485
• code review:
  • MKLDNN:
    • add MKLDNNAddtoLayer: https://github.com/PaddlePaddle/Paddle/pull/5309
    • enable benchmark resnet with MKLDNN：https://github.com/PaddlePaddle/Paddle/pull/5422
    • enable bias for mkldnn_addto: https://github.com/PaddlePaddle/Paddle/pull/5440
  • Op/compiler/doc/models/website:
    • Enable hidden/cell state initialization and enhance unit testing in LSTM operator. https://github.com/PaddlePaddle/Paddle/pull/5429#pullrequestreview-75032423
    • Put files to platlib：https://github.com/PaddlePaddle/Paddle/pull/5396
    • Fix ci not exit 1 when error：https://github.com/PaddlePaddle/Paddle/pull/5431
    • Fix the doc for Momentum and Adam optimizer：https://github.com/PaddlePaddle/Paddle/pull/5391
    • Rename doc/howto/cross_compiling into doc/mobile：https://github.com/PaddlePaddle/Paddle/pull/5399
    • rename the directory of neural sequence QA：https://github.com/PaddlePaddle/models/pull/425
    • update the README for the Models repo：https://github.com/PaddlePaddle/models/pull/426
    • uci_housing.py can download the trained model automatically: https://github.com/PaddlePaddle/Paddle/pull/5463
    • Add quick start code: https://github.com/PaddlePaddle/book/pull/443

Deep Speech 2:

• Add the document about docker running for DS2
  • https://github.com/PaddlePaddle/models/pull/418
• Fix a bug in CTC decoder
  • https://github.com/PaddlePaddle/models/pull/427
• Re-tune the decoder and update benchmark results for different models
  • https://github.com/PaddlePaddle/models/pull/436
  • https://github.com/PaddlePaddle/models/pull/423
  • https://github.com/PaddlePaddle/models/pull/430

Intel MKL_DNN:

• Carry out benchmark test on laptop and desktop computer
  • https://github.com/PaddlePaddle/Paddle/issues/5462

• DS2
  • Remove padding
    https://github.com/PaddlePaddle/models/issues/414
    https://github.com/PaddlePaddle/Paddle/pull/5388
    https://github.com/PaddlePaddle/Paddle/pull/5488
  • Refine PR
    https://github.com/PaddlePaddle/models/pull/396
  • Docs
    https://github.com/PaddlePaddle/models/pull/368
    https://github.com/PaddlePaddle/models/pull/364
• Paddle
  • More robust memory handler
    https://github.com/PaddlePaddle/Paddle/pull/5418
  • Refine expand_op
    https://github.com/PaddlePaddle/Paddle/pull/4061

• autoscaler scheduler design:
  • https://github.com/PaddlePaddle/cloud/pull/468
• manylinux compile env and CI:
  • https://github.com/PaddlePaddle/Paddle/pull/5430
  • https://github.com/PaddlePaddle/buildtools
• Some op refines
• Send Recv op WIP.

• Auto-scaling experiment
  • Bug fix:https://github.com/PaddlePaddle/cloud/pull/467
  • Bug fix:https://github.com/PaddlePaddle/cloud/pull/470
  • Review:https://github.com/PaddlePaddle/cloud/pull/468
• Op porting
  • Fix seq concat op with refactoring LoD:https://github.com/PaddlePaddle/Padd

• OP Porting
  • Refine GRU Operator(Merged):
    • https://github.com/PaddlePaddle/Paddle/pull/5255
  • Refine ChunkEval Operator:
    • https://github.com/PaddlePaddle/Paddle/pull/5016
• Others
  • Convert Model from tensorflow to PaddlePaddle

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

• Op Porting
  • Add conv3d_gemm_op. (merged)
    • conv2d kernel and conv3d kernel are wrote together
    • https://github.com/PaddlePaddle/Paddle/pull/4709
  • Add 3D convolution transposed operator. (merged)
    • deconv2d kernel and deconv3d kernel write together
    • https://github.com/PaddlePaddle/Paddle/pull/5118
• enhancement
  • Add dilations for conv2d and optimize conv2d code
    • https://github.com/PaddlePaddle/Paddle/pull/5472
  • Add 3D data type for cudnn (conv3d/deconv3d/pool3d)
  • fix_pool_op_attr_name
    • https://github.com/PaddlePaddle/Paddle/pull/5457

• Mobile deployment
  • [Merged] Rename doc/howto/cross_compiling into doc/mobile
    • https://github.com/PaddlePaddle/Paddle/pull/5399
    • https://github.com/PaddlePaddle/Mobile/pull/24
  • [Merged] Fix the lack of linking libraries to libpaddle_capi_engine
    • https://github.com/PaddlePaddle/Paddle/pull/5343
  • [Merged] Enable the build for iOS simulator
    • https://github.com/PaddlePaddle/Paddle/pull/5211
  • SSD Demo on Android and iOS
    • issue: https://github.com/PaddlePaddle/Mobile/issues/25
    • capi example on linux: https://github.com/Xreki/Mobile/tree/add_ssd_linux_demo/demo/linux/ssd
    • Review
      • Init commit for iOS AICamera demo: https://github.com/PaddlePaddle/Mobile/pull/26

PR:

• Add attribute check in operator constructor
• Forward: feature/while_op
• Refactor Static RNN unittest
• Update lod_tensor.md
• Static lstm sanity check
• Add fill_constant_batch_size_like_op to Static RNN's h_boot

review:

• Rewrite StaticRNN with Executor
• Remove fill_constant_batch_size_like_op.h
• Feature/shrink memory op
• Feature/rnn to array to lod tensor
• Add unittest, backward of array read/write op
• Add To Array operator for dynamic RNN
• ReadFromArray/WriteToArray op
• Compare Operator
• Feature/rank sort lod op
• Remove warning in executor
• Add net drawer for visualizing the graph

• Update the Chinese navigation bar text.
  • https://github.com/PaddlePaddle/PaddlePaddle.org/pull/265
• Update translated Chinese strings.
  • https://github.com/PaddlePaddle/PaddlePaddle.org/pull/251
• Consolidate the URL pattern for documentation, book, models, and mob…
  • https://github.com/PaddlePaddle/PaddlePaddle.org/pull/249
• uci_housing.py can download the trained model automatically.
  • https://github.com/PaddlePaddle/Paddle/pull/5463
• Add quickstart code
  • https://github.com/PaddlePaddle/book/pull/443
• SSD iOS Demo

• PR for fixing attribute naming: https://github.com/PaddlePaddle/Paddle/pull/5453