2018 02 07

wangkuiyi

• CSP

  • Comments about the new test of the panic behavior of closed channels: https://github.com/PaddlePaddle/Paddle/pull/8197#pullrequestreview-94894463
  • Test the behavior of receiving from a closed channel: https://github.com/PaddlePaddle/Paddle/pull/8175
  • Rewrite the send/recv part of the CSP design doc: https://github.com/PaddlePaddle/Paddle/pull/8164
  • Correctly call CloseChannel in the destructor of a channel: https://github.com/PaddlePaddle/Paddle/issues/8248
  • Rewrite the send/recv part of CSP.md: https://github.com/PaddlePaddle/Paddle/issues/8163
  • Rewrite SimpleUnbufferedChannelTest in channel_test.cc: https://github.com/PaddlePaddle/Paddle/issues/8160
  • Make select.case a little more readable: https://github.com/PaddlePaddle/Paddle/issues/8165

• CI

  • Improve CI speed: https://github.com/PaddlePaddle/Paddle/issues/7992

• EDL

  • Update run_on_minikube.md: https://github.com/PaddlePaddle/cloud/pull/599
  • Add paddle cloud figure: https://github.com/PaddlePaddle/cloud/pull/597
  • Change to use the updated deepcopy generation program: https://github.com/PaddlePaddle/cloud/pull/584
  • Refactorize EDL controller and autoscaler: https://github.com/PaddlePaddle/cloud/pull/579
  • Remove buildconfig: https://github.com/PaddlePaddle/cloud/pull/574
  • Complete the minikube tutorial: https://github.com/PaddlePaddle/cloud/issues/602
  • Why Kubernetes version 1.6.4 #598: https://github.com/PaddlePaddle/cloud/issues/598
  • Check Python style: https://github.com/PaddlePaddle/cloud/issues/596
  • Restructure the doc directory: https://github.com/PaddlePaddle/cloud/issues/595
  • Translate Chinese documents into English and keep only the English version: https://github.com/PaddlePaddle/cloud/issues/594
  • Add a figure about component-interactions: https://github.com/PaddlePaddle/cloud/issues/593
  • Add document on how to build Go program into go/README.md: https://github.com/PaddlePaddle/cloud/issues/592
  • Rewrite root README.md file: https://github.com/PaddlePaddle/cloud/issues/591
  • Create python directory and move paddlecloud into there: https://github.com/PaddlePaddle/cloud/issues/590
  • .travis.yml refers to nonexisting bash script: https://github.com/PaddlePaddle/cloud/issues/586
  • Make it clear that what a training job is composed of: https://github.com/PaddlePaddle/cloud/issues/575

kavyasrinet

• Update copyright year of files: https://github.com/PaddlePaddle/Paddle/pull/8018
• Update gradient check writeup: https://github.com/PaddlePaddle/Paddle/pull/8057
• CSP in Fluid :
  • Proposing Python syntax for send/recv in CSP: https://github.com/PaddlePaddle/Paddle/pull/8093
  • Fix unbuffered test to a generic test: https://github.com/PaddlePaddle/Paddle/pull/8162
  • Add unit test with less receiver more senders for unbuffered channels: https://github.com/PaddlePaddle/Paddle/pull/8060
  • Add more receivers less senders. Receive should block: https://github.com/PaddlePaddle/Paddle/pull/8061
  • Add block when channel is empty, successfully transfer when there is a send: https://github.com/PaddlePaddle/Paddle/pull/8063
  • Check if receiver blocks on empty channel, resume on close : https://github.com/PaddlePaddle/Paddle/pull/8166
  • Return zero on closed channel: https://github.com/PaddlePaddle/Paddle/pull/8167
  • [WIP] Implement Select op https://github.com/PaddlePaddle/Paddle/issues/8247
  • [WIP] Make select.case a little more readable https://github.com/PaddlePaddle/Paddle/issues/8165
• PR review :
  • https://github.com/PaddlePaddle/Paddle/pull/8067
  • https://github.com/PaddlePaddle/Paddle/pull/8197
  • https://github.com/PaddlePaddle/Paddle/pull/8175
  • https://github.com/PaddlePaddle/Paddle/pull/8164
  • https://github.com/PaddlePaddle/Paddle/pull/8132
  • https://github.com/PaddlePaddle/Paddle/pull/8109
  • https://github.com/PaddlePaddle/Paddle/pull/8094
  • https://github.com/PaddlePaddle/Paddle/pull/8067
  • https://github.com/PaddlePaddle/Paddle/pull/7984

abhinavarora

• CSP in Fluid

  • Add proposed fluid syntax for select statement in Fluid's implementation of CSP https://github.com/PaddlePaddle/Paddle/pull/7908
  • Adding unit tests for checking that closing unbuffered channel closes all blocked senders and receivers https://github.com/PaddlePaddle/Paddle/pull/8067
  • Buffered channel should notify both condition variables on close https://github.com/PaddlePaddle/Paddle/pull/8094
  • Close buffered channel should unblock the blocked senders and receivers https://github.com/PaddlePaddle/Paddle/pull/8109
  • Channel should not accept senders once it is closed https://github.com/PaddlePaddle/Paddle/pull/8132
  • Fix for program crash when destructor is called before channel close with blocked readers/writers https://github.com/PaddlePaddle/Paddle/pull/8197
  • Correctly call CloseChannel in the destructor of a channel https://github.com/PaddlePaddle/Paddle/issues/8248
  • [WIP] Implement Select op https://github.com/PaddlePaddle/Paddle/issues/8247
  • [WIP] Make select.case a little more readable https://github.com/PaddlePaddle/Paddle/issues/8165
  • Add Unit Tests for Buffered and Unbuffered Channels https://github.com/PaddlePaddle/Paddle/issues/8066

• PR reviews

  • https://github.com/PaddlePaddle/Paddle/pull/8243#pullrequestreview-94817829
  • https://github.com/PaddlePaddle/Paddle/pull/8222#pullrequestreview-94852717
  • https://github.com/PaddlePaddle/Paddle/pull/8179#pullrequestreview-94226601
  • https://github.com/PaddlePaddle/Paddle/pull/8175#pullrequestreview-94206523
  • https://github.com/PaddlePaddle/Paddle/pull/8171#pullrequestreview-94474499
  • https://github.com/PaddlePaddle/Paddle/pull/8169#pullrequestreview-94816497
  • https://github.com/PaddlePaddle/Paddle/pull/8098#pullrequestreview-93819458
  • https://github.com/PaddlePaddle/Paddle/pull/8065#pullrequestreview-93501530
  • https://github.com/PaddlePaddle/Paddle/pull/8060#pullrequestreview-93483199

• Attend meeting with Yi at Intel for potential integration of Fluid with Intel NGraph

helinwang

• EDL code improvement with Yi.
  • Refactorize EDL controller and autoscaler: https://github.com/PaddlePaddle/cloud/pull/579#commits-pushed-29a0a8e
  • update EDL build command line: https://github.com/PaddlePaddle/cloud/pull/589
• CI improvements:
  • Terminate and fail long running training tests when got NaN loss: https://github.com/PaddlePaddle/Paddle/pull/8169
  • Install GPU and replace CPU for the big black CI machine.
  • Now we have 4 CI machines, each running 2 GPU for our CI test.
• Distributed PaddlePaddle reviews and discussions:
  • https://github.com/PaddlePaddle/Paddle/pull/8241#pullrequestreview-94931822
  • https://github.com/PaddlePaddle/Paddle/pull/8023#pullrequestreview-93451118
  • https://github.com/PaddlePaddle/Paddle/pull/8023#pullrequestreview-94163408
  • https://github.com/PaddlePaddle/Paddle/issues/8119#issuecomment-363226055
• CSP code reviews:
  • https://github.com/PaddlePaddle/Paddle/pull/8164#pullrequestreview-94114620

Yu Yang

• Debugging of Parallel.Do

  • https://github.com/PaddlePaddle/Paddle/pull/8186
  • https://github.com/PaddlePaddle/Paddle/pull/8224
  • https://github.com/PaddlePaddle/Paddle/pull/8181
  • https://github.com/PaddlePaddle/Paddle/pull/8210

• Add NMT Demo for Parallel.Do

  • https://github.com/PaddlePaddle/Paddle/pull/8005

• Reviewed PR

  • https://github.com/PaddlePaddle/Paddle/pull/8233
  • https://github.com/PaddlePaddle/Paddle/pull/8184
  • https://github.com/PaddlePaddle/Paddle/pull/8135
  • https://github.com/PaddlePaddle/Paddle/pull/8087
  • https://github.com/PaddlePaddle/Paddle/pull/8080
  • https://github.com/PaddlePaddle/Paddle/pull/8076

YanXu

• PR:
  • Move learning rate and releated op to pserver, https://github.com/PaddlePaddle/Paddle/pull/8209/files
  • Move python folder, https://github.com/PaddlePaddle/cloud/pull/601
  • Fix doc, https://github.com/PaddlePaddle/cloud/pull/600
  • Fix compile error with cmake flag WITH_DISTRIBUTED=ON, https://github.com/PaddlePaddle/Paddle/pull/8030
• PR Review:
  • https://github.com/PaddlePaddle/cloud/pull/587
  • Benchmark report, https://github.com/PaddlePaddle/Paddle/pull/7539#pullrequestreview-93236498

qijun

Memory optimization on Fluid:

machine translation demo saves 80.5% memory.

• memory optimization for dynamic RNN
  • https://github.com/PaddlePaddle/Paddle/pull/8041
• Refine get_cfgs method of memory optimization transpiler
  • https://github.com/PaddlePaddle/Paddle/pull/8080
• fix bug in memory optimization transpiler
  • https://github.com/PaddlePaddle/Paddle/pull/8233

Save/Load API:

• add design doc for save/load api
  • https://github.com/PaddlePaddle/Paddle/pull/8207

nickyfantasy

• Translated VisualDL backend design doc to english
  • https://github.com/PaddlePaddle/VisualDL/pull/248
• VisualDL current backend / SDK structure and development doc
  • https://github.com/PaddlePaddle/VisualDL/pull/262

Chenxi

• follow up with CI machine procurement

• faster CI build script update

  • https://github.com/PaddlePaddle/Paddle/pull/8017

• CI upgrade and database migration

• Paddle Cloud doc improvements

  • https://github.com/PaddlePaddle/cloud/issues/602

gongweibao

• PaddleCloud
  • Fix golint error https://github.com/PaddlePaddle/cloud/pull/606
  • Cleanup useless gen_config.sh and fix golint bug https://github.com/PaddlePaddle/cloud/pull/603
  • Check Python style https://github.com/PaddlePaddle/cloud/pull/607
• Review & discussion
  • Update run_on_minikube.md https://github.com/PaddlePaddle/cloud/pull/599
  • Add paddle cloud figure https://github.com/PaddlePaddle/cloud/pull/597
• Fluid distribution
  • kubeflow https://github.com/kubeflow/kubeflow

Yan Chunwei

• fluid
  • WIP CONV seq2seq
  • debuger with graphviz
  • debuger with better-formatted codes

https://github.com/PaddlePaddle/Paddle/pull/8140#pullrequestreview-93938166 https://github.com/PaddlePaddle/Paddle/pull/8031#pullrequestreview-93217701 https://github.com/PaddlePaddle/Paddle/pull/8031/files/05de2c7303639ec90d1f562bbba6c09085d61877#r165547659

• visualdl https://github.com/PaddlePaddle/VisualDL/pull/238#discussion_r165281264 https://github.com/PaddlePaddle/VisualDL/pull/238#pullrequestreview-93564206 https://github.com/PaddlePaddle/VisualDL/pull/251#pullrequestreview-93799454 https://github.com/PaddlePaddle/VisualDL/pull/260#pullrequestreview-94275005

fengjiayi

• Fundamental Data Reading in C++:
  • https://github.com/PaddlePaddle/Paddle/pull/8009 (C++ Readers and related ops)
  • https://github.com/PaddlePaddle/Paddle/pull/8135 (Reader supports in VarDesc)
  • https://github.com/PaddlePaddle/Paddle/pull/8237 (Design doc)
• Code clean:
  • https://github.com/PaddlePaddle/Paddle/pull/8087 (Simplify shape inference code) -Reviews:
  • https://github.com/PaddlePaddle/Paddle/pull/8091

wanghaoshuang

• OCR
  • Make CTC align op support for empty output
    • https://github.com/PaddlePaddle/Paddle/pull/8114
  • Fix normalization in edit_distance op
    • https://github.com/PaddlePaddle/Paddle/pull/8245
  • Add sequence error into EditDistance evaluator[WIP]
  • Refine and verify CTC train model
    • https://github.com/PaddlePaddle/models/pull/596
• Fix
  • https://github.com/PaddlePaddle/Paddle/pull/8243
• Review
  • https://github.com/PaddlePaddle/models/pull/618

luotao

• inference:
  • remove libwarpctc.so in core.so and libpaddle_fluid.so: https://github.com/PaddlePaddle/Paddle/pull/7762
  • Add make clean in docker/build.sh: https://github.com/PaddlePaddle/Paddle/pull/8076
  • make inference_lib_dist for fluid inference shared library: https://github.com/PaddlePaddle/Paddle/pull/7977
• document:
  • Reclassify and change of V2 API documentation structure: https://github.com/PaddlePaddle/Paddle/pull/8085
  • remove duplicated mobile index: https://github.com/PaddlePaddle/Paddle/pull/8123
  • add independent sphinx tree for api: https://github.com/PaddlePaddle/Paddle/pull/8145
  • adjust the structure of documentation: https://github.com/PaddlePaddle/Paddle/pull/8226
  • PaddlePaddle.org Website Enhancement: https://shimo.im/sheet/uoMu6uOR6sE3Mj5L/Q4ADL
• code review:
  • Fix default build_type env: https://github.com/PaddlePaddle/Paddle/pull/8019
  • Fix document errors: https://github.com/PaddlePaddle/Paddle/pull/8040
  • Refine the inference unittest recognize_digits: https://github.com/PaddlePaddle/Paddle/pull/8147

Dang Qingqing

• SSD on Fluid:
  • [Merged] multi-class non-maximum suppression operator.
    • https://github.com/PaddlePaddle/Paddle/pull/7953
  • [Merged] Add target assigner operator for SSD detection.
    • https://github.com/PaddlePaddle/Paddle/pull/8193
  • Fix the input dimension and warrnings in multiclass_nms_op.
    • https://github.com/PaddlePaddle/Paddle/pull/8232
    • https://github.com/PaddlePaddle/Paddle/pull/8158
  • Coder Review:
    • mAP Evalutor: https://github.com/PaddlePaddle/Paddle/pull/6588
    • https://github.com/PaddlePaddle/Paddle/pull/7921
• Model Conversion (caffe2fluid)
  • Add softmax into Python API.
    • https://github.com/PaddlePaddle/Paddle/pull/8065
  • Allow uers to specify the name of moving mean and variance in batch_norm interface.
    • https://github.com/PaddlePaddle/Paddle/pull/8069
  • coder review for caffe2fluid:
    • https://github.com/PaddlePaddle/models/pull/617#issuecomment-362254554
  • The AR model has been converted successfluly.
• Coder Review for OCR model:
  • https://github.com/PaddlePaddle/models/pull/596#pullrequestreview-93543154
  • https://github.com/PaddlePaddle/Paddle/pull/8114

zhangchao

• Change the 'mul_op' to 'matmul_op' in fc.
  • https://github.com/PaddlePaddle/Paddle/pull/8101.
• Profiling for book08 machine translation.

guosheng

• PR:
  • Add the Transformer implementation for NMT
    • https://github.com/PaddlePaddle/models/pull/659

wanghaox

• SSD Detection mAP operator :https://github.com/PaddlePaddle/Paddle/pull/6588
• SSD Target_assign operator discuss and review: https://github.com/PaddlePaddle/Paddle/pull/8193#pullrequestreview-94572241

wangmeng

• Refine SE-ResNeXt
  • https://github.com/PaddlePaddle/models/pull/601
  • https://github.com/PaddlePaddle/models/pull/611
• SE-ResNeXt forward verification of fluid and pytorch

wuyi

• GPU performance issue:https://github.com/PaddlePaddle/Paddle/issues/8119#issuecomment-363231595
• Transpiler enhancement
  • https://github.com/PaddlePaddle/Paddle/pull/8241
  • https://github.com/PaddlePaddle/Paddle/pull/8049
• Document updates for installation
• PaddleCloud CRD reviews
  • https://github.com/PaddlePaddle/cloud/pull/563
  • https://github.com/PaddlePaddle/cloud/pull/587
  • https://github.com/PaddlePaddle/cloud/pull/565
• PaddleCloud code cleanup reviews
  • https://github.com/PaddlePaddle/cloud/pull/600
  • https://github.com/PaddlePaddle/cloud/pull/607
  • https://github.com/PaddlePaddle/cloud/pull/606
  • https://github.com/PaddlePaddle/cloud/pull/604
  • https://github.com/PaddlePaddle/cloud/pull/603
  • https://github.com/PaddlePaddle/cloud/pull/601
  • https://github.com/PaddlePaddle/cloud/pull/599

Yibing Liu

DeepASR:

• [merged] Refactor model config: profiling, multi-device running, model saving etc.
  • https://github.com/PaddlePaddle/models/pull/636
• [under review] Add demo script for inference
  • https://github.com/PaddlePaddle/models/pull/653
• Profile on single GPU
  • https://github.com/PaddlePaddle/models/issues/637
• [merged] Fix the bug in profiler
  • https://github.com/PaddlePaddle/Paddle/pull/8116

Code review:

• DeepASR
  • https://github.com/PaddlePaddle/models/pull/635
  • https://github.com/PaddlePaddle/models/pull/641
  • https://github.com/PaddlePaddle/models/pull/656
  • https://github.com/PaddlePaddle/models/pull/642
  • https://github.com/PaddlePaddle/models/pull/628
  • https://github.com/PaddlePaddle/models/pull/625
• Fluid:
  • https://github.com/PaddlePaddle/Paddle/pull/8245
  • https://github.com/PaddlePaddle/Paddle/pull/8114

qiaolongfei

fluid

• learning rate decay (https://github.com/PaddlePaddle/Paddle/issues/7769)

  • Add polynomial_decay and piecewise_decay https://github.com/PaddlePaddle/Paddle/pull/8013
    • [WIP]Use force cpu in fill constant op https://github.com/PaddlePaddle/Paddle/pull/8254

• Switch design and implementation

  • switch-op design https://github.com/PaddlePaddle/Paddle/pull/8031
  • Impl scalar switch case op with condition op https://github.com/PaddlePaddle/Paddle/pull/8184

• Review 2. disable nccl and parallel_do test https://github.com/PaddlePaddle/Paddle/pull/8200 3. simplify shape inference code https://github.com/PaddlePaddle/Paddle/pull/8087 4. debug/format protobuf to human-readable codes https://github.com/PaddlePaddle/Paddle/pull/8086 5. Allow uers to specify the name of moving mean and variance in batch_norm interface. https://github.com/PaddlePaddle/Paddle/pull/8069 6. init graph debug https://github.com/PaddlePaddle/Paddle/pull/8026

yangyaming

• Parallel data reader
  https://github.com/PaddlePaddle/models/pull/635
  https://github.com/PaddlePaddle/models/pull/641
  https://github.com/PaddlePaddle/models/pull/656
• Fix CI for models repo
  https://github.com/PaddlePaddle/models/pull/625
• Directory structure
  https://github.com/PaddlePaddle/models/pull/628
• Review
  https://github.com/PaddlePaddle/models/pull/636
  https://github.com/PaddlePaddle/models/pull/640

dongzhihong

• Fluid
  • add c++ version recordio
    • https://github.com/PaddlePaddle/recordio/pull/12
  • accelerate elementwise_add_grad
    • https://github.com/PaddlePaddle/Paddle/pull/7961
  • fix nccl multigpu hang up
    • https://github.com/PaddlePaddle/Paddle/pull/8244
  • fix nccl disable dso
    • https://github.com/PaddlePaddle/Paddle/pull/8240
  • mixed_vector fix bug
    • https://github.com/PaddlePaddle/Paddle/pull/8045
  • add mixed vector copy to peer
    • https://github.com/PaddlePaddle/Paddle/pull/8149
• Benchmark
  • build nginx docker image

zhaochengduo

PR

• GPU implementation for layer normalization operator
  • https://github.com/PaddlePaddle/Paddle/pull/8006
• Refine elementwise_op
  • https://github.com/PaddlePaddle/Paddle/pull/8091
• CSP in Fluid
  • https://github.com/PaddlePaddle/Paddle/pull/8222
  • https://github.com/PaddlePaddle/Paddle/pull/8098
  • https://github.com/PaddlePaddle/Paddle/pull/8179
  • https://github.com/PaddlePaddle/Paddle/pull/8182
• Refine prior_box_op
  • https://github.com/PaddlePaddle/Paddle/pull/8151
• Refine for_range
  • https://github.com/PaddlePaddle/Paddle/pull/8152

Reviewed PR

• CSP in Fluid
  • https://github.com/PaddlePaddle/Paddle/pull/8197
  • https://github.com/PaddlePaddle/Paddle/pull/8175
  • https://github.com/PaddlePaddle/Paddle/pull/8063
  • https://github.com/PaddlePaddle/Paddle/pull/8109
  • https://github.com/PaddlePaddle/Paddle/pull/8067
  • https://github.com/PaddlePaddle/Paddle/pull/7984
• Add target assigner operator for SSD detection.
  • https://github.com/PaddlePaddle/Paddle/pull/8193
• Add depthwise conv op gpu
  • https://github.com/PaddlePaddle/Paddle/pull/7885
• Allow uers to specify the name of moving mean and variance in batch_norm interface.
  • https://github.com/PaddlePaddle/Paddle/pull/8069
• Feature/memory profiler
  • https://github.com/PaddlePaddle/Paddle/pull/7983

tonyyang-svail

NCCL:

• NCCL compilation and linking version doesn't match:
  • https://github.com/PaddlePaddle/Paddle/issues/8195
  • https://github.com/PaddlePaddle/Paddle/pull/8196

CI: driver, failed test

• https://github.com/PaddlePaddle/Paddle/pull/8200

Liu Yiqun

• Fluid
  • [Merged] Refine the inference unittest recognize_digits
    • https://github.com/PaddlePaddle/Paddle/pull/8147
  • [Reviewing] Simplify the inference unittests' cmake and codes
    • https://github.com/PaddlePaddle/Paddle/pull/8216
  • [Reviewing] Add C++ inference unittest of recommender system
    • https://github.com/PaddlePaddle/Paddle/pull/8227
  • Review
    • make inference_lib_dist for fluid inference shared library: https://github.com/PaddlePaddle/Paddle/pull/7977
    • modify C++ load() for inference: https://github.com/PaddlePaddle/Paddle/pull/8024
    • inference example and unittest for label_semantic_roles: https://github.com/PaddlePaddle/Paddle/pull/8058
    • inference example for image_classification and unittest for inference: https://github.com/PaddlePaddle/Paddle/pull/8020

cs2be (thuan)

VisualDL

• VDL slow when user sets incorrect log directory (https://github.com/PaddlePaddle/VisualDL/pull/249)
• Binary Record Fix (https://github.com/PaddlePaddle/VisualDL/pull/260)

PaddlePaddle.org

• Add Paddle API section (https://github.com/PaddlePaddle/PaddlePaddle.org/pull/391)

jetfuel (Jeff)

• tweak the script so the user doesn't need to specify abs path to the logdir: https://github.com/PaddlePaddle/VisualDL/pull/240
• Update the documentation description in the storage.py: https://github.com/PaddlePaddle/VisualDL/pull/251
• Fix python path typo in the start dev server script: https://github.com/PaddlePaddle/VisualDL/pull/253
• Provide instructions on virtualenv: https://github.com/PaddlePaddle/VisualDL/pull/258
• rename vdl_scratch.py to vdl_create_scratch_log: https://github.com/PaddlePaddle/VisualDL/pull/259
• Add how to develop frontend documentation: https://github.com/PaddlePaddle/VisualDL/pull/234
• I have been focusing on the VisualDL documentation. To make sure we have proper information on the website.
• Working the migrating the VisualDL component framework from San To Vue.

varunarora

• Researching and implementing 2D low-dimensional representations (PCA) of tensors in VDL (PR coming soon).
• Feedback on client side and server side documentation, leading to multiple PRs: https://github.com/PaddlePaddle/VisualDL/pull/253, https://github.com/PaddlePaddle/VisualDL/pull/259.
• Visual debugger user study summarizing / synthesis
• Prototyping developer experience using visual debugger.
• Managing VDL issues.

kexinzhao

Fluid Inference:

• Add inference example for label semantic roles: https://github.com/PaddlePaddle/Paddle/pull/8058
• Add inference example for rnn_encoder_decoder: https://github.com/PaddlePaddle/Paddle/pull/8176
• Add inference example for understand sentiment: https://github.com/PaddlePaddle/Paddle/pull/8251
• Add inference example for fit a line book chapter: https://github.com/PaddlePaddle/Paddle/pull/8208
• Enhance prune method for RNN nested blocks: https://github.com/PaddlePaddle/Paddle/issues/8062
• Enhance program desc constructor for RNN nested blocks: https://github.com/PaddlePaddle/Paddle/issues/8161
• Review:
  • https://github.com/PaddlePaddle/Paddle/pull/8147#pullrequestreview-94204518
  • https://github.com/PaddlePaddle/Paddle/pull/8024#pullrequestreview-94563877

sidgoyal78

Inference: PR/issue:

• Refine load() based on new combining ops: https://github.com/PaddlePaddle/Paddle/pull/8024 (under review)
• Inference example and unit test for inference api, chapter: image classification: https://github.com/PaddlePaddle/Paddle/pull/8020
• Add support for loading from buffer for inference framework: https://github.com/PaddlePaddle/Paddle/pull/8246
• Add inference unit-test for word2vec book chapter: https://github.com/PaddlePaddle/Paddle/pull/8198

Review:

• RNN PR: https://github.com/PaddlePaddle/Paddle/pull/8176
• Inference example and unit test for label_semantic_roles: https://github.com/PaddlePaddle/Paddle/pull/8058
• Refine the inference unittest recognize_digits: https://github.com/PaddlePaddle/Paddle/pull/8147