Code

Add matmul_v2 kernel in pten#36844

Code

Merged

zyfncg merged 155 commits intoPaddlePaddle:developPaddlePaddle/Paddle:developfrom zyfncg:matmul_v2_newzyfncg/Paddle:matmul_v2_newCopy head branch name to clipboardNov 2, 2021

+866-12

Commits

Commits on Jul 9, 2021

• initial tensor design & sign kernel demo

  chenwhql

  committed
  3f545c4

  Copy full SHA for 3f545c4

Commits on Jul 12, 2021

• add move constructor for meta & add lodtensor

  chenwhql

  committed
  1f4ea40

  Copy full SHA for 1f4ea40

• add dirs & sign xpu kernel

  chenwhql

  committed
  44bf926

  Copy full SHA for 44bf926

Commits on Jul 15, 2021

• add mean cpu&cuda kernel impl

  chenwhql

  committed
  b20689d

  Copy full SHA for b20689d

• move sign & mean xpu & npu kernel

  chenwhql

  committed
  79d2a1a

  Copy full SHA for 79d2a1a

Commits on Jul 16, 2021

• add selected_rows basic impl

  chenwhql

  committed
  434136f

  Copy full SHA for 434136f

Commits on Jul 27, 2021

• refactor design, BaseTensor to DenseTensor, etc.

  chenwhql

  committed
  6c6ee22

  Copy full SHA for 6c6ee22

• Merge branch 'develop' of /~https://github.com/PaddlePaddle/Paddle into op2func_refactor

  chenwhql

  committed
  013c3fb

  Copy full SHA for 013c3fb

Commits on Jul 28, 2021

• add scale mkldnn kernel

  chenwhql

  committed
  33bba06

  Copy full SHA for 33bba06

• polish xpu & npu impl details

  chenwhql

  committed
  d895a11

  Copy full SHA for d895a11

Commits on Jul 29, 2021

• fix mkldnn reuse compile failed

  chenwhql

  committed
  62ebf01

  Copy full SHA for 62ebf01

• change tensor operation lib name

  chenwhql

  committed
  7c09726

  Copy full SHA for 7c09726

• resolve conflit with develop

  chenwhql

  committed
  7ae7f2f

  Copy full SHA for 7ae7f2f

• rename util filename

  chenwhql

  committed
  288efc2

  Copy full SHA for 288efc2

• add more comments

  chenwhql

  committed
  be3ddd5

  Copy full SHA for be3ddd5

Commits on Jul 30, 2021

• change TensorImplInterface to TensorInterface

  chenwhql

  committed
  3386c49

  Copy full SHA for 3386c49

Commits on Aug 4, 2021

• add kernel key and factory

  chenwhql

  committed
  4ef6be5

  Copy full SHA for 4ef6be5

• remove MKLDNNTensorMeta, add MKLDNNDenseTensor

  chenwhql

  committed
  b69066e

  Copy full SHA for b69066e

• resolve conflict with develop

  chenwhql

  committed
  1d4f90e

  Copy full SHA for 1d4f90e

Commits on Aug 5, 2021

• change XXDeviceContext to XXContext

  chenwhql

  committed
  c732d57

  Copy full SHA for c732d57

Commits on Aug 16, 2021

• add base kernel registrar utils & test on sign

  chenwhql

  committed
  374345f

  Copy full SHA for 374345f

• resolve conflict with develop

  chenwhql

  committed
  bbb6473

  Copy full SHA for bbb6473

• replace boost::any by paddle::any

  chenwhql

  committed
  0e18ff4

  Copy full SHA for 0e18ff4

Commits on Aug 17, 2021

• fix several ci failed

  chenwhql

  committed
  805896b

  Copy full SHA for 805896b

• fix npu compile error

  chenwhql

  committed
  fc4442b

  Copy full SHA for fc4442b

• add ordered map util

  chenwhql

  committed
  cefe30a

  Copy full SHA for cefe30a

• Merge branch 'develop' of /~https://github.com/PaddlePaddle/Paddle into op2func_refactor

  chenwhql

  committed
  aa3e79b

  Copy full SHA for aa3e79b

• fix multiple ordered_map compile errors

  chenwhql

  committed
  a1753a0

  Copy full SHA for a1753a0

Commits on Aug 18, 2021

• move dev into include dir

  chenwhql

  committed
  05a82e7

  Copy full SHA for 05a82e7

Commits on Aug 19, 2021

• support sign op in static op run

  chenwhql

  committed
  90e9090

  Copy full SHA for 90e9090

Commits on Aug 23, 2021

• fix static op run error

  chenwhql

  committed
  a94eefd

  Copy full SHA for a94eefd

• resolve confilt with develop

  chenwhql

  committed
  19da152

  Copy full SHA for 19da152

• fix new executor compile failed

  chenwhql

  committed
  021a505

  Copy full SHA for 021a505

Commits on Aug 25, 2021

• add dygraph branch & remove sign_op.h

  chenwhql

  committed
  f24e45e

  Copy full SHA for f24e45e

Commits on Aug 26, 2021

• fix test_infer_no_need_buffer_slots

  chenwhql

  committed
  44acc84

  Copy full SHA for 44acc84

• fix rocm compile link error

  chenwhql

  committed
  2b66ab4

  Copy full SHA for 2b66ab4

• fix unitybuild error & clear glog

  chenwhql

  committed
  2a5ce9b

  Copy full SHA for 2a5ce9b

• fix npu compile failed

  chenwhql

  committed
  39b7d06

  Copy full SHA for 39b7d06

• skip quant trans test

  chenwhql

  committed
  d4dec61

  Copy full SHA for d4dec61

• fix part windows compile problem

  chenwhql

  committed
  461f146

  Copy full SHA for 461f146

• Merge branch 'develop' into op2func_refactor

  chenwhql

  authored
  35aee9a

  Copy full SHA for 35aee9a

• fix xpu enforce error

  chenwhql

  committed
  ddfbbdd

  Copy full SHA for ddfbbdd

• Merge branch 'op2func_refactor' of /~https://github.com/chenwhql/Paddle into op2func_refactor

  chenwhql

  committed
  57bcd67

  Copy full SHA for 57bcd67

Commits on Aug 27, 2021

• fix inference test failed

  chenwhql

  committed
  7d82352

  Copy full SHA for 7d82352

• Merge branch 'op2func_refactor' of /~https://github.com/chenwhql/Paddle into op2func_refactor

  chenwhql

  committed
  d55bb4b

  Copy full SHA for d55bb4b

• Merge branch 'develop' into op2func_refactor

  chenwhql

  authored
  d9476dd

  Copy full SHA for d9476dd

Commits on Aug 30, 2021

• remove ordered_map to solve quant failed

  chenwhql

  committed
  193ee9d

  Copy full SHA for 193ee9d

• Merge branch 'develop' of /~https://github.com/PaddlePaddle/Paddle into op2func_refactor

  chenwhql

  committed
  f2db581

  Copy full SHA for f2db581

• fix part of rcom compile faild

  chenwhql

  committed
  db6ff09

  Copy full SHA for db6ff09

• Merge branch 'op2func_refactor' of /~https://github.com/chenwhql/Paddle into op2func_refactor

  chenwhql

  committed
  80bf6b8

  Copy full SHA for 80bf6b8

Commits on Aug 31, 2021

• add more register kernels

  chenwhql

  committed
  9031ab3

  Copy full SHA for 9031ab3

Commits on Sep 3, 2021

• revert scale kernel temporarily

  chenwhql

  committed
  f7bbaca

  Copy full SHA for f7bbaca

• Merge branch 'develop' into op2func_refactor

  chenwhql

  authored
  cea19d0

  Copy full SHA for cea19d0

Commits on Sep 6, 2021

• fix code format error

  chenwhql

  committed
  568bebd

  Copy full SHA for 568bebd

Commits on Sep 7, 2021

• add new kernel registrar marco

  chenwhql

  committed
  0eedc92

  Copy full SHA for 0eedc92

• rename top to tcmpt

  chenwhql

  committed
  509d13e

  Copy full SHA for 509d13e

• revert xpu, npu, mkldnn impl & remove op def

  chenwhql

  committed
  7146f92

  Copy full SHA for 7146f92

Commits on Sep 8, 2021

• add kernel args parse functor to auto parse args

  chenwhql

  committed
  321b141

  Copy full SHA for 321b141

• resolve confilt with develop

  chenwhql

  committed
  57a14c6

  Copy full SHA for 57a14c6

Commits on Sep 9, 2021

• revert some change & add scale kernels

  chenwhql

  committed
  c3ebfea

  Copy full SHA for c3ebfea

• add op proto in dygraph kernelcontext building

  chenwhql

  committed
  b67de9c

  Copy full SHA for b67de9c

Commits on Sep 10, 2021

• polish kernel dispatch logic & nameing rule

  chenwhql

  committed
  13c02aa

  Copy full SHA for 13c02aa

• fix scale kernel match error

  chenwhql

  committed
  1987ce9

  Copy full SHA for 1987ce9

• fix scale test failed

  chenwhql

  committed
  33a4c41

  Copy full SHA for 33a4c41

Commits on Sep 13, 2021

• add mean API and unittest

  chenwhql

  committed
  c32fde9

  Copy full SHA for c32fde9

Commits on Sep 17, 2021

• test mean api success

  chenwhql

  committed
  a4e53ef

  Copy full SHA for a4e53ef

Commits on Sep 18, 2021

• add branch to solve compiled error

  chenwhql

  committed
  1d9f33f

  Copy full SHA for 1d9f33f

• skip clang format error

  chenwhql

  committed
  b0cf02c

  Copy full SHA for b0cf02c

• add mean skip rule in op_library

  chenwhql

  committed
  95a612e

  Copy full SHA for 95a612e

• add dot kernel, api and unittest (#6)

  MingMingShangTian

  authored
  83d6f77

  Copy full SHA for 83d6f77

• remove old kernel and add symbol link

  chenwhql

  committed
  dad5e61

  Copy full SHA for dad5e61

• resolve conflit with tianyu

  chenwhql

  committed
  027f0b2

  Copy full SHA for 027f0b2

• fix dot compiled failed

  chenwhql

  committed
  8add5e4

  Copy full SHA for 8add5e4

• Merge branch 'develop' of /~https://github.com/PaddlePaddle/Paddle into op2func_refactor

  chenwhql

  committed
  01b5ded

  Copy full SHA for 01b5ded

Commits on Sep 22, 2021

• add merco for module declare

  chenwhql

  committed
  71a3403

  Copy full SHA for 71a3403

• fix npu and xpu compile error

  chenwhql

  committed
  4663033

  Copy full SHA for 4663033

Commits on Sep 23, 2021

• revert sign, mean, scale, dot kernel removing

  chenwhql

  committed
  be15b02

  Copy full SHA for be15b02

• add comment for keeping old kernel impl

  chenwhql

  committed
  8371096

  Copy full SHA for 8371096

• fix mutable_data error

  chenwhql

  committed
  f1f6c8e

  Copy full SHA for f1f6c8e

Commits on Sep 24, 2021

• fix bfloat16 conflit

  chenwhql

  committed
  5547b44

  Copy full SHA for 5547b44

• fix inference undef error

  chenwhql

  committed
  dd3323d

  Copy full SHA for dd3323d

• Merge branch 'op2func_refactor' of /~https://github.com/chenwhql/Paddle into op2func_refactor

  chenwhql

  committed
  65e68c6

  Copy full SHA for 65e68c6

Commits on Sep 26, 2021

• adapt to msvc compile rules

  chenwhql

  committed
  caaed19

  Copy full SHA for caaed19

• Merge branch 'op2func_refactor' of /~https://github.com/chenwhql/Paddle into op2func_refactor

  chenwhql

  committed
  c9a3f38

  Copy full SHA for c9a3f38

• polish comment for template inst

  chenwhql

  committed
  46b7762

  Copy full SHA for 46b7762

Commits on Sep 27, 2021

• add cmake template instantiation for win

  chenwhql

  committed
  4253f49

  Copy full SHA for 4253f49

Commits on Sep 28, 2021

• Merge branch 'develop' of /~https://github.com/PaddlePaddle/Paddle into op2func_refactor

  chenwhql

  committed
  4e871ea

  Copy full SHA for 4e871ea

Commits on Sep 29, 2021

• Merge branch 'develop' into op2func_refactor

  chenwhql

  authored
  395a50f

  Copy full SHA for 395a50f

• fix backend to place device id bug

  chenwhql

  committed
  817f052

  Copy full SHA for 817f052

• Merge branch 'op2func_refactor' of /~https://github.com/chenwhql/Paddle into op2func_refactor

  chenwhql

  committed
  7f640a6

  Copy full SHA for 7f640a6

• fix ifdef error

  chenwhql

  committed
  bf0f99b

  Copy full SHA for bf0f99b

Commits on Sep 30, 2021

• Op2functor (#7)

  Show description for 73de891
  MingMingShangTian

  authored
  73de891

  Copy full SHA for 73de891

• fill_any_like kernel refactor (#10)

  Show description for e9b219d
  zyfncg

  authored
  e9b219d

  Copy full SHA for e9b219d

Commits on Oct 11, 2021

• skip dtype for fill_any_like

  chenwhql

  committed
  9789890

  Copy full SHA for 9789890

• add attrs for kernel key constrcut

  chenwhql

  committed
  9b33270

  Copy full SHA for 9b33270

Commits on Oct 12, 2021

• add use_pt_kernel Flags to control whether to use pt kernel (PaddlePaddle#13)

  Show description for aa6ed57
  MingMingShangTian

  authored
  aa6ed57

  Copy full SHA for aa6ed57

• fix mutable_data cuda place error

  chenwhql

  committed
  9db8e4a

  Copy full SHA for 9db8e4a

• Merge branch 'develop' into op2func_refactor

  chenwhql

  authored
  12c1178

  Copy full SHA for 12c1178

Commits on Oct 13, 2021

• move high level apis into hapi

  chenwhql

  committed
  c882b5c

  Copy full SHA for c882b5c

• Merge branch 'op2func_refactor' of /~https://github.com/chenwhql/Paddle into op2func_refactor

  chenwhql

  committed
  e30ca2a

  Copy full SHA for e30ca2a

Commits on Oct 14, 2021

• remove selectedrows adapting temporarily

  chenwhql

  committed
  46ba70c

  Copy full SHA for 46ba70c

• Support Scalar in Tensor Compute Library (PaddlePaddle#14)

  Show description for 073aef3
  zyfncg

  authored
  073aef3

  Copy full SHA for 073aef3

• resolve conflit with yunfei

  chenwhql

  committed
  06789ba

  Copy full SHA for 06789ba

• remove mkldnn tensor & polish details

  chenwhql

  committed
  3f5f789

  Copy full SHA for 3f5f789

Commits on Oct 15, 2021

• use flat_hash_map and small_vector in kernel factory

  chenwhql

  committed
  2309149

  Copy full SHA for 2309149

• Refactor flatten kernel (PaddlePaddle#12)

  Show description for 6ce92e5
  YuanRisheng

  authored
  6ce92e5

  Copy full SHA for 6ce92e5

• Revert "use flat_hash_map and small_vector in kernel factory"

  Show description for e0322d5
  chenwhql

  committed
  e0322d5

  Copy full SHA for e0322d5

• Move cpu, cuda and other device code into kernels (PaddlePaddle#15)

  Show description for d3ab655
  zyfncg

  authored
  d3ab655

  Copy full SHA for d3ab655

Commits on Oct 18, 2021

• Perfect unitests (PaddlePaddle#16)

  Show description for ddc7de8
  YuanRisheng

  authored
  ddc7de8

  Copy full SHA for ddc7de8

• replace with flat_hash_map, small_vector (PaddlePaddle#19)

  Show description for 37791f7
  MingMingShangTian

  authored
  37791f7

  Copy full SHA for 37791f7

• Perfect unitests (PaddlePaddle#20)

  Show description for 28a6374
  YuanRisheng

  authored
  28a6374

  Copy full SHA for 28a6374

• refactor execution adapting impl

  chenwhql

  committed
  e3e2b50

  Copy full SHA for e3e2b50

• resolve conflits

  chenwhql

  committed
  e0710fd

  Copy full SHA for e0710fd

Commits on Oct 19, 2021

• fix insert conflit

  chenwhql

  committed
  ff19bd0

  Copy full SHA for ff19bd0

• Fix CI bug of test_yolov3 (PaddlePaddle#21)

  Show description for 1dd0145
  zyfncg

  authored
  1dd0145

  Copy full SHA for 1dd0145

• add the tensor base class, test=develop (PaddlePaddle#17)

  Show description for b77d1ee
  

  Shixiaowei02

  and

  chenwhql

  authored
  b77d1ee

  Copy full SHA for b77d1ee

• [no-verify] commit backend and tensor signature changes

  chenwhql

  committed
  320b5f1

  Copy full SHA for 320b5f1

• resolve conflit with xiaowei

  chenwhql

  committed
  5b2999f

  Copy full SHA for 5b2999f

Commits on Oct 20, 2021

• Rename tcmpt to pten (PaddlePaddle#23)

  Show description for 466ce03
  zyfncg

  authored
  466ce03

  Copy full SHA for 466ce03

• remove k of all enum var

  chenwhql

  committed
  beec280

  Copy full SHA for beec280

• resolve conflit with yunfei

  chenwhql

  committed
  a49fd44

  Copy full SHA for a49fd44

• Merge branch 'develop' into op2func_refactor

  chenwhql

  authored
  373f9c1

  Copy full SHA for 373f9c1

• remove kernel_instantiate (PaddlePaddle#26)

  MingMingShangTian

  authored
  ce210b4

  Copy full SHA for ce210b4

• remove symbols and spatial_tensor

  chenwhql

  committed
  4e71d15

  Copy full SHA for 4e71d15

• change common to functions

  chenwhql

  committed
  04cf058

  Copy full SHA for 04cf058

• readd share tensor impl methods

  chenwhql

  committed
  ab8db2d

  Copy full SHA for ab8db2d

• add a candidate dense tensor class, test=develop (PaddlePaddle#28)

  Shixiaowei02

  authored
  f1c9661

  Copy full SHA for f1c9661

• change all Pt to Pten

  chenwhql

  committed
  d3674e9

  Copy full SHA for d3674e9

• Merge branch 'op2func_refactor' of /~https://github.com/chenwhql/Paddle into op2func_refactor

  chenwhql

  committed
  4e2c0dd

  Copy full SHA for 4e2c0dd

Commits on Oct 21, 2021

• resolve conflit with xiaowei

  chenwhql

  committed
  bbe59bc

  Copy full SHA for bbe59bc

• Op2functor opt1 (PaddlePaddle#27)

  Show description for 76a588e
  

  MingMingShangTian

  and

  chenwhql

  authored
  76a588e

  Copy full SHA for 76a588e

• polish kernel factory and kernel registry

  chenwhql

  committed
  fb224ab

  Copy full SHA for fb224ab

Commits on Oct 22, 2021

• fix operator test error msg mismatch

  chenwhql

  committed
  252fb79

  Copy full SHA for 252fb79

• remove tensor signature and backend set member

  chenwhql

  committed
  19b1095

  Copy full SHA for 19b1095

• move scalar and polish enforce

  chenwhql

  committed
  24ef6c5

  Copy full SHA for 24ef6c5

• revert dtype layout change to fix error

  chenwhql

  committed
  1685b67

  Copy full SHA for 1685b67

• fix enum operator override error

  chenwhql

  committed
  7b7e988

  Copy full SHA for 7b7e988

• add several base unittests

  chenwhql

  committed
  52fead0

  Copy full SHA for 52fead0

Commits on Oct 23, 2021

• add pten utils tests

  chenwhql

  committed
  2ff2721

  Copy full SHA for 2ff2721

• Merge branch 'develop' of /~https://github.com/PaddlePaddle/Paddle into op2func_refactor

  chenwhql

  committed
  e3ed2c6

  Copy full SHA for e3ed2c6

Commits on Oct 24, 2021

• polish some details

  chenwhql

  committed
  b5c77e5

  Copy full SHA for b5c77e5

Commits on Oct 26, 2021

• Dev/op2func refactor 3 (PaddlePaddle#30)

  Show description for 5240ac0
  

  chenwhql

  and

  Shixiaowei02

  authored
  5240ac0

  Copy full SHA for 5240ac0

• Merge branch 'op2func_refactor' of /~https://github.com/chenwhql/Paddle into op2func_refactor

  chenwhql

  committed
  5fb285c

  Copy full SHA for 5fb285c

• Merge branch 'develop' into op2func_refactor

  chenwhql

  authored
  3276897

  Copy full SHA for 3276897

• polish some details

  chenwhql

  committed
  558a848

  Copy full SHA for 558a848

• Merge branch 'op2func_refactor' of /~https://github.com/chenwhql/Paddle into op2func_refactor

  chenwhql

  committed
  72910fa

  Copy full SHA for 72910fa

• polish kernel signature details

  chenwhql

  committed
  8f100da

  Copy full SHA for 8f100da

Commits on Oct 27, 2021

• fix a bug about offsets of the tensor, test=develop (PaddlePaddle#31)

  Show description for be9df70
  

  chenwhql

  and

  Shixiaowei02

  authored
  be9df70

  Copy full SHA for be9df70

Commits on Oct 28, 2021

• add matmul kernel in pten

  zyfncg

  committed
  0fb60d0

  Copy full SHA for 0fb60d0

• add unittest for new matmul_v2 kernel

  zyfncg

  committed
  7758f14

  Copy full SHA for 7758f14

• fix bug of CI compile

  zyfncg

  committed
  e7bf067

  Copy full SHA for e7bf067

• fix bug of CI compile

  zyfncg

  committed
  2e76b24

  Copy full SHA for 2e76b24

Commits on Nov 1, 2021

• Merge branch 'develop' of /~https://github.com/PaddlePaddle/Paddle into matmul_v2_new

  zyfncg

  committed
  8ee3572

  Copy full SHA for 8ee3572

• merge conflict

  zyfncg

  committed
  2b6a565

  Copy full SHA for 2b6a565

• remove useless file

  zyfncg

  committed
  0de9aae

  Copy full SHA for 0de9aae