Code

update#5

Code

Merged

AnnaTrainingG merged 117 commits intoAnnaTrainingG:developAnnaTrainingG/Paddle:developfrom PaddlePaddle:developPaddlePaddle/Paddle:developCopy head branch name to clipboardMay 25, 2021

+13,742-4,773

Commits

Commits on May 7, 2021

• add timeout for queue.get (#32747)

  chenwhql

  authored
  db5eac2

  Copy full SHA for db5eac2

• model_benchmark (#32600)

  tianshuo78520a

  authored
  7468253

  Copy full SHA for 7468253

• add other 15 activation ops (#32622)

  ZzSean

  authored
  b2160e7

  Copy full SHA for b2160e7

• fix distro (#32771)

  tianshuo78520a

  authored
  816afb9

  Copy full SHA for 816afb9

• Fix compile error on jetson platform (#32748)

  Show description for 8ce6b39
  LielinJiang

  authored
  8ce6b39

  Copy full SHA for 8ce6b39

• fix distro in manylinux (#32784)

  tianshuo78520a

  authored
  3753416

  Copy full SHA for 3753416

• remove packages in __all__ (#32759)

  Show description for a77ade0
  

  zhiboniu

  and

  ZHUI

  authored
  a77ade0

  Copy full SHA for a77ade0

Commits on May 8, 2021

• Add raw program meta optimizer (#32597)

  Show description for c1c18b0
  

  lilong12

  authored
  c1c18b0

  Copy full SHA for c1c18b0

• [Paddle-TRT]fix trt-converter-fc_op (#32671)

  Show description for 62d848d
  Wangzheee

  authored
  62d848d

  Copy full SHA for 62d848d

• bugfix: parallel_executor for xpu should use BindThreadedSSAGraphExecutor (#32792)

  houj04

  authored
  e8e4a9c

  Copy full SHA for e8e4a9c

• Remove np Deprecation Warning since np.bool is alias of bool (#32798)

  Show description for 8a42b1f
  zhhsplendid

  authored
  8a42b1f

  Copy full SHA for 8a42b1f

• [NPU] refine update_loss_scaling npu kernel (#32580)

  Show description for 4628b6f
  pangyoki

  authored
  4628b6f

  Copy full SHA for 4628b6f

• add c_identity op npu (#32787)

  Show description for c8affff
  Baibaifan

  authored
  c8affff

  Copy full SHA for c8affff

• 【heterps】support cuda11 for heterps; add profiler in oneps (#32640)

  Show description for beab956
  danleifeng

  authored
  beab956

  Copy full SHA for beab956

Commits on May 10, 2021

• Dynamic amp support sync_batch_norm op (#32770)

  sljlp

  authored
  23ab01e

  Copy full SHA for 23ab01e

• make check_op_desc.py support python3 (#32807)

  tianshuo78520a

  authored
  92adece

  Copy full SHA for 92adece

• fix paddle_build bug (#32813)

  pangyoki

  authored
  fd9a236

  Copy full SHA for fd9a236

• Revert "make check_op_desc.py support python3 (#32807)" (#32818)

  Show description for 5fc734c
  tianshuo78520a

  authored
  5fc734c

  Copy full SHA for 5fc734c

• update unittest for uint8 problem (#32790)

  lyuwenyu

  authored
  e357cfd

  Copy full SHA for e357cfd

• fix npu compile error (#32820)

  zhiqiu

  authored
  5aa8faa

  Copy full SHA for 5aa8faa

• [pslib] pslib with cmake (#32800)

  Show description for fbbc339
  Thunderbrook

  authored
  fbbc339

  Copy full SHA for fbbc339

• Support different data type between input and output (#32823)

  ZzSean

  authored
  3419de5

  Copy full SHA for 3419de5

• modify en_doco of spectral norm test=document_fix (#32812)

  wangna11BD

  authored
  1eb59ef

  Copy full SHA for 1eb59ef

Commits on May 11, 2021

• fix ce bug in label value, test=develop

  

  huangjun12

  authored and

  chajchaj

  committed
  400eb9d

  Copy full SHA for 400eb9d

• fix ci bug

  

  huangjun12

  authored and

  chajchaj

  committed
  e2c293f

  Copy full SHA for e2c293f

• fix ci coverage

  

  huangjun12

  authored and

  chajchaj

  committed
  6cd96c1

  Copy full SHA for 6cd96c1

• imporve efficiency

  

  huangjun12

  authored and

  chajchaj

  committed
  ef7e5fc

  Copy full SHA for ef7e5fc

• fix ci coverage bug

  

  huangjun12

  authored and

  chajchaj

  committed
  e1ea895

  Copy full SHA for e1ea895

• add weigth data to unit test

  

  huangjun12

  authored and

  chajchaj

  committed
  9495211

  Copy full SHA for 9495211

• add ignore_index for test case

  

  huangjun12

  authored and

  chajchaj

  committed
  9cdf6bd

  Copy full SHA for 9cdf6bd

• fix ci bug

  

  huangjun12

  authored and

  chajchaj

  committed
  a9e5305

  Copy full SHA for a9e5305

• Support control flow in DataParallel (#32826)

  Show description for 298f210
  ForFishes

  authored
  298f210

  Copy full SHA for 298f210

• fix rccl bug (#32808)

  xymyeah

  authored
  93fce18

  Copy full SHA for 93fce18

• fix cmake expressions error, test=develop (#32815)

  Avin0323

  authored
  84eca16

  Copy full SHA for 84eca16

• Fix retry error in download when exception occurs (#32816)

  Show description for 1ef2327
  lyuwenyu

  authored
  1ef2327

  Copy full SHA for 1ef2327

Commits on May 12, 2021

• Polish unittest test_decoupled_py_reader (#32832)

  Show description for 2611ed2
  chenwhql

  authored
  2611ed2

  Copy full SHA for 2611ed2

• add varbasecopy func to fix the ParamBase type bug in layers.to API (#32789)

  Show description for 067f558
  MingMingShangTian

  authored
  067f558

  Copy full SHA for 067f558

• change check_op_desc.py to py3 (#32825)

  tianshuo78520a

  authored
  cefc063

  Copy full SHA for cefc063

• Simple authors change (#32763)

  andreazanetti

  authored
  e1a4c83

  Copy full SHA for e1a4c83

• Optimize/fleet save (#32817)

  Show description for 890f626
  seiriosPlus

  authored
  890f626

  Copy full SHA for 890f626

• [NPU] Support npu pinned allocator and manage Tensor on NPUPinnedPlace (#32840)

  liym27

  authored
  6b3bb79

  Copy full SHA for 6b3bb79

• fix dataloader exit hang when join re-enter (#32827)

  Show description for f1d6302
  heavengate

  authored
  f1d6302

  Copy full SHA for f1d6302

• [NPU] Support async copy for TensorFromVector with event (#32563)

  liym27

  authored
  85512d6

  Copy full SHA for 85512d6

• Fix comments in framework (#32861)

  Show description for e676360
  lyuwenyu

  authored
  e676360

  Copy full SHA for e676360

• Fix error in device __all__ (#32860)

  lyuwenyu

  authored
  eff84a5

  Copy full SHA for eff84a5

• fix the error of fake_quant_dequant op name (#32866)

  juncaipeng

  authored
  24ffcd0

  Copy full SHA for 24ffcd0

• Polish Windows CI and open the normal GPU unittest on CI (#32794)

  Show description for eeca963
  zhwesky2010

  authored
  eeca963

  Copy full SHA for eeca963

Commits on May 13, 2021

• change unique op VisitaDataType from small to tiny (#32872)

  MingMingShangTian

  authored
  b60ab6b

  Copy full SHA for b60ab6b

• fix stack grad gpu (#32781) (#32877)

  bjjwwang

  authored
  3e47eee

  Copy full SHA for 3e47eee

• solved some npu bugs (#32793)

  Baibaifan

  authored
  c3ae0d4

  Copy full SHA for c3ae0d4

• add varbase_copy support CUDAPinnedPlace (#32883)

  MingMingShangTian

  authored
  48fc16f

  Copy full SHA for 48fc16f

• [NPU] support global accumulator for adam (#32780)

  Show description for dace3fd
  zhiqiu

  authored
  dace3fd

  Copy full SHA for dace3fd

• fix wait server ready (#32889)

  wangxicoding

  authored
  a8625aa

  Copy full SHA for a8625aa

Commits on May 14, 2021

• Polish code for _getitem_impl_ (#32868)

  liym27

  authored
  096b2f5

  Copy full SHA for 096b2f5

• Optimization the broadcast performance of elementwise_add (#32512)

  JamesLim-sy

  authored
  b035c8b

  Copy full SHA for b035c8b

• [Dy2Static]Add param_guard in ParameterList to support @to_static

  Aurelius84

  authored
  e48091d

  Copy full SHA for e48091d

• solove_matmulv2_npu_bugs (#32896)

  Baibaifan

  authored
  2d9d8f5

  Copy full SHA for 2d9d8f5

• fix launch port already in use (#32892)

  wangxicoding

  authored
  6473e27

  Copy full SHA for 6473e27

• test=document_fix (#32906)

  tianshuo78520a

  authored
  ed9e772

  Copy full SHA for ed9e772

• use the required instruction to determine if the environment fits the sample code's required. (#32766)

  Show description for 42aad30
  wadefelix

  authored
  42aad30

  Copy full SHA for 42aad30

• [oneDNN] Refactoring of softmax grad onednn kernel to match common API (#32851)

  jczaja

  authored
  479689f

  Copy full SHA for 479689f

• Fix four error messages (#32899)

  Show description for c4787d7
  Kqnonrime

  authored
  c4787d7

  Copy full SHA for c4787d7

• [Custom Op]Remove PADDLE_WITH_MKLDNN in custom_op (#32903)

  Aurelius84

  authored
  15b05c7

  Copy full SHA for 15b05c7

• Doc of paddle.save/load (#32900)

  Show description for 62f6550
  hbwx24

  authored
  62f6550

  Copy full SHA for 62f6550

• Set the default value of protocol to 4. (#32904)

  hbwx24

  authored
  e89fb25

  Copy full SHA for e89fb25

• change ut file map (#32841)

  lelelelelez

  authored
  a3debea

  Copy full SHA for a3debea

Commits on May 17, 2021

• solove_device_guard_bugs (#32915)

  Baibaifan

  authored
  585564d

  Copy full SHA for 585564d

• BugFix with ParseInputDataType from LodTensorArray (#32918)

  Show description for 5f1c07d
  Aurelius84

  authored
  5f1c07d

  Copy full SHA for 5f1c07d

• remove redundant graph files (#32924)

  Show description for 906db71
  seemingwang

  authored
  906db71

  Copy full SHA for 906db71

• [HybridParallel]Fix precision problem of model parallel (#32897)

  Show description for c809530
  ForFishes

  authored
  c809530

  Copy full SHA for c809530

Commits on May 18, 2021

• fix the paddle compare op for the broadcast when the element equal (#32941)

  Show description for c72ed82
  wawltor

  authored
  c72ed82

  Copy full SHA for c72ed82

• notest;test=zcltest (#32821)

  lelelelelez

  authored
  59997d5

  Copy full SHA for 59997d5

• unit double (#32902)

  Show description for 29bbeb0
  Thunderbrook

  authored
  29bbeb0

  Copy full SHA for 29bbeb0

• update kunlun bkcl to support multi-machine (#32577)

  QingshuChen

  authored
  59b74ee

  Copy full SHA for 59b74ee

• [Dy2Static] Refactor param_guard logic of @to_static (#32867)

  Show description for b8d493d
  Aurelius84

  authored
  b8d493d

  Copy full SHA for b8d493d

• [UnitTest]Enhance grep syntax to avoid random failed of test_dist_mnist_dgc_nccl (#32946)

  Show description for b5882c6
  Aurelius84

  authored
  b5882c6

  Copy full SHA for b5882c6

• relu supports bfloat16 data type (#32542)

  Avin0323

  authored
  bcd40f2

  Copy full SHA for bcd40f2

• add unit8 for concat (#32850)

  vslyu

  authored
  53580bb

  Copy full SHA for 53580bb

• Update paths to Quant models (#32870)

  Show description for 5d62748
  wozna

  authored
  5d62748

  Copy full SHA for 5d62748

• [NPU] fix accuracy npu op bug and change top_k's output to int64 (#32935)

  Show description for c66586b
  pangyoki

  authored
  c66586b

  Copy full SHA for c66586b

Commits on May 19, 2021

• Fix Link unittest exe random fail (#32891)

  zhwesky2010

  authored
  d7d7fae

  Copy full SHA for d7d7fae

• add enforce check for set_value (#32972)

  chenwhql

  authored
  af89a94

  Copy full SHA for af89a94

• [oneDNN] Pool softmax and LRN access to cache optimized (#32922)

  jczaja

  authored
  56008aa

  Copy full SHA for 56008aa

• [Dy2Stat]BugFix StaticAanlysis with gast.Subscript (#32969)

  Show description for c285261
  Aurelius84

  authored
  c285261

  Copy full SHA for c285261

• remove ut from parallel_ut list (#32788)

  Show description for f0b2f59
  XieYunshen

  authored
  f0b2f59

  Copy full SHA for f0b2f59

• Optimize 102Flowers dataset reading speed (#31408)

  Show description for 67c2700
  GT-ZhangAcer

  authored
  67c2700

  Copy full SHA for 67c2700

• [Rocm] fix test of random_crop_op & logsumexp (#32824)

  Show description for aa4a56f
  Ray2020BD

  authored
  aa4a56f

  Copy full SHA for aa4a56f

• fix the jetson allocator strategy, test=develop (#32932)

  Shixiaowei02

  authored
  1e1600e

  Copy full SHA for 1e1600e

• CI skip inference test if only python files modified (#32962)

  Show description for 7896b51
  Avin0323

  authored
  7896b51

  Copy full SHA for 7896b51

• fix test_paddle_save_load and test_paddle_save_load_binary (#32949)

  Show description for 6f8de31
  hbwx24

  authored
  6f8de31

  Copy full SHA for 6f8de31

Commits on May 20, 2021

• Binary functor envoking of elementwise broadcast (#32928)

  JamesLim-sy

  authored
  1494952

  Copy full SHA for 1494952

• handle remove files in pr (#32940)

  lelelelelez

  authored
  7e27b5a

  Copy full SHA for 7e27b5a

• remove unused shell (#32954)

  seemingwang

  authored
  8854786

  Copy full SHA for 8854786

• Add complex template type (#32857)

  Show description for 738bf20
  MingMingShangTian

  authored
  738bf20

  Copy full SHA for 738bf20

• Polish code for setitem and getitem (#32911)

  liym27

  authored
  848cabf

  Copy full SHA for 848cabf

• [Dy2Stat]Support convert sublayers in Sequential Container (#32978)

  Show description for e409c7c
  Aurelius84

  authored
  e409c7c

  Copy full SHA for e409c7c

• revert_matmulv2_npu (#33014)

  Baibaifan

  authored
  be8e94a

  Copy full SHA for be8e94a

• fix gather op and add logsumexp op on kunlun (#32931)

  Show description for a96e8bc
  tangzhiyi11

  authored
  a96e8bc

  Copy full SHA for a96e8bc

Commits on May 21, 2021

• [NPU] cast indices and label if their type is not consistent in accuracy npu op (#33016)

  Show description for 70dc5f4
  pangyoki

  authored
  70dc5f4

  Copy full SHA for 70dc5f4

• update conda build script for cuda11 (#29594)

  Show description for 44668a7
  XieYunshen

  authored
  44668a7

  Copy full SHA for 44668a7

• paddle.to_tensor supports LoDTensor (#33027)

  zhiqiu

  authored
  a85eddd

  Copy full SHA for a85eddd

• fix model_benchmark ci (#33035)

  Show description for 0e5d832
  tianshuo78520a

  authored
  0e5d832

  Copy full SHA for 0e5d832

• optimize softmax with cross entropy hard label (#32290)

  Show description for 7be6191
  xingfeng01

  authored
  7be6191

  Copy full SHA for 7be6191

• add method for enhance pass,test=develop (#33004)

  winter-wang

  authored
  79ed717

  Copy full SHA for 79ed717

• replace complex64/128 with complex template in cast Op (#33019)

  Show description for 79d918d
  MingMingShangTian

  authored
  79d918d

  Copy full SHA for 79d918d

Commits on May 22, 2021

• Added oneDNN matmul grad BF16/FP32 kernel (#32968)

  Show description for e2a3a6f
  jakpiase

  authored
  e2a3a6f

  Copy full SHA for e2a3a6f

• refine conv2d doc (#33045)

  jerrywgz

  authored
  a6dc68b

  Copy full SHA for a6dc68b

Commits on May 24, 2021

• Support OutType tmeplate argument in elementwise_broadcast branch (#33060)

  JamesLim-sy

  authored
  d6aea4a

  Copy full SHA for d6aea4a

• fix potential overflow problem & node add & node remove & node clear (#33055)

  Show description for 60ac160
  

  6 peopleauthored
  60ac160

  Copy full SHA for 60ac160

• open launch ps test=develop (#33044)

  gongweibao

  authored
  d0d5586

  Copy full SHA for d0d5586

• [oneDNN] bump up oneDNN to 2.2.2 (#32685)

  Show description for b8e4ec7
  jczaja

  authored
  b8e4ec7

  Copy full SHA for b8e4ec7

• enhance unittest for yolo_box (#33070)

  jerrywgz

  authored
  99a11e3

  Copy full SHA for 99a11e3

• Revert "fix model_benchmark ci (#33035)" (#33080)

  Show description for 6ad5ece
  tianshuo78520a

  authored
  6ad5ece

  Copy full SHA for 6ad5ece

• [HybridParallel]Fix pipeline in dygraph (#33007)

  Show description for 4920c47
  ForFishes

  authored
  4920c47

  Copy full SHA for 4920c47

Commits on May 25, 2021

• Add a new high performance framework for reduce ops (#32697)

  AnnaTrainingG

  authored
  88b43b5

  Copy full SHA for 88b43b5

• Added scale op FP32/BF16 FWD/BWD kernels (#32975)

  jakpiase

  authored
  86ea8dc

  Copy full SHA for 86ea8dc

• modify Ops to complex template (#33041)

  Show description for 5fa44c3
  MingMingShangTian

  authored
  5fa44c3

  Copy full SHA for 5fa44c3