提交 · 033ebe7e652aed9de07ea525abe3d15ba659a1b9 · PaddlePaddle / Paddle

09 12月, 2021 11 次提交
- S
  Refine CUDA atomicAdd for FP16 by CUDA primitive methods (#37895) · 033ebe7e
  由 sneaxiy 提交于 12月 09, 2021
```
* fix cuda atomicAdd for FP16

* try to fix ci
```
  033ebe7e
- Z
  Fixed eager compilation issues by temporarily turn off AutoCodeGen fo… (#37992) · 34a06cf5
  由 Zhanlue Yang 提交于 12月 09, 2021
```
* Fixed eager compilation issues by temporarily turn off AutoCodeGen for specific ops

* Removed op_types
```
  34a06cf5
- J
  
  add ipu device p2 (#37840) · cb636a48
  由 jianghaicheng 提交于 12月 09, 2021
  
  cb636a48
- R
  
  optimize flip op, removing duplicated computation when dim size is one (#37825) · 890638cf
  由 Roc 提交于 12月 09, 2021
  
  890638cf
- F
  
  format softmax forward (#37927) · 18aca3f5
  由 Feng Xing 提交于 12月 09, 2021
  
  18aca3f5
- B
  
  Add varbase init name (#37947) · fdf62e1e
  由 Baibaifan 提交于 12月 09, 2021
  
  fdf62e1e
- W
  default accessor and multi table config (#37714) · a9e0d28c
  由 wangguanqun 提交于 12月 09, 2021
```
* default accessor and multi table config

* add unittest

* add unittest

* delete print
```
  a9e0d28c
- W
  [Paddle-Inference] fix_ele_convert: IElementWiseLayer can broadcast (#37908) · f695dc97
  由 Wangzheee 提交于 12月 09, 2021
```
* fix_ele_convert: IElementWiseLayer can broadcast

* fix_ele_convert
```
  f695dc97
- C
  
  adjust main dir (#37916) · 1911b6f0
  由 Chen Weihang 提交于 12月 08, 2021
  
  1911b6f0
- C
  
  fix make error by alias name change (#37971) · e3f68f42
  由 Chen Weihang 提交于 12月 08, 2021
  
  e3f68f42
- Z
  Adjusted Eager AutoCodeGen to Support Operators with Multiple OpBases & Enable... · 9aed9ea0
  由 Zhanlue Yang 提交于 12月 09, 2021
```
Adjusted Eager AutoCodeGen to Support Operators with Multiple OpBases & Enable Passing Output Tensor as Input Argument (#37943)

* Rearranged Eager AutoCodeGen directory structure

* Removed USE_OP in Eager AutoCodeGen

* Enabled generation for Operators without Grad/Inputs/Outputs

* Resolved operators without input

* Fixed merge conflicts

* Enabled Eager AutoCodeGen for 10+ more operators

* Refactored Eager AutoCodeGen with more organized helper objects

* Enabled Eager AutoCodeGen for operators with multiple OpBases

* Adjusted Eager AutoCodeGen to Enable Passing Output Tensor as Input Argument
```
  9aed9ea0
08 12月, 2021 17 次提交

[Eager] coreops to 495 (#37926) · aff7397b

由 wanghuancoder 提交于 12月 08, 2021

* refine a test case, test=develop

* publish python c api for eager, test=develop

* revert modify about test_allclose_layer.py, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* delete numpy includes, use pybind11 numpy.h, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* suport eager error msg, and add grad test case, test=develop

* refine, test=develop

* refine, test=develop

* generate eager core ops, only 4 ops, test=develop

* Rearranged Eager AutoCodeGen directory structure

* Removed USE_OP in Eager AutoCodeGen

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* open 500 list

* refine, test=develop

* refine, test=develop

* refine, test=develop

* fix auto code gen, test=develop

* Enabled generation for Operators without Grad/Inputs/Outputs

* refine, test=develop

* refine, test=develop

* refine, test=develop

* add to pyobject, test=develop

* Resolved operators without input

* merge pr 37837

* refine

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine,test=develop
Co-authored-by: Njim19930609 <jim19930609@gmail.com>

aff7397b

C
Revert "pylayer support tuple/list type args (#37727)" (#37956) · d848ff04
由 chentianyu03 提交于 12月 08, 2021
```
This reverts commit a73064f2.
```
d848ff04

add a subdirectory named cinn in operators and move releated files into it (#37938) · 9cb637ed

由 CtfGo 提交于 12月 08, 2021

1. add a subdirectory named `cinn` in `paddle/fluid/operators` directory and move releated files into it
2. seperate CinnLaunchContext class from `cinn_launch_op.h` and put it in a  new independent file named `cinn_launch_context.h`, so that it can be included by others clearly.

9cb637ed

Fix host event recorder (#37944) · 20471de7

由 liutiexing 提交于 12月 08, 2021

* add align for WorkQueue

* add spinlock

* merge develop

* merge

* Add EventsWaiter

* Revert "Add EventsWaiter"

This reverts commit e206173aa9be7401b83a53581627bfaf557c8fb2.

* Fix RecordEvent
Co-authored-by: Nliutiexing <liutiexing@google.com>

20471de7

T

Delete git useless files (#37952) · 5cecfd23
由 tianshuo78520a 提交于 12月 08, 2021

5cecfd23
Y
[PTen]Add alias kernel name (#37881) · ff6507db
由 YuanRisheng 提交于 12月 08, 2021
```
* add alias kernel name

* modify code as suggestions
```
ff6507db

Add paddle.lerp API to do a linear interpolation (#37253) · 1716324c

由 wuhuanzhou 提交于 12月 08, 2021

* save temp

* add unittest, test=develop

* fix ci error, test=develop

* fix grad accuracy error, test=develop

* fix unused error, test=develop

* fix compilation error on Windows, test=develop

* add unittest, test=develop

* modify by review comment and add lerp_

* fix inplace api, test=develop

* fix inplace api, test=develop

* fix coverage error, test=develop

1716324c

W

[fleet_executor] Add interceptor gc (#37889) · 6b48dfe9
由 WangXi 提交于 12月 08, 2021

6b48dfe9
C
implementation of broadcast sub backward by reduce (#37754) · 567e6bbc
由 crystal 提交于 12月 08, 2021
```
* add boardcast_sub

* add boardcast_sub
```
567e6bbc

Fix CUDAGraphAllocator bug for StreamSafeCUDAAllocator (#37821) · b4a67491

由 From00 提交于 12月 08, 2021

* Fix CUDAGraph bug for StreamSafeCUDAAllocator

* Add CUDAGrapthAllocator check in multi-stream interface

* Set FLAGS_use_stream_safe_cuda_allocator defaulted to false

* Fix environment error for cmake

* Fix cmake error

* Add UT of GetAllocatorInterfaceTest

* Add UT of CUDAGraphExceptionTest

* Enhance CUDAGraphExceptionTest

b4a67491

C

add check whether tensor is inplace and leaf when calcute gradient (#37931) · 2c02a580
由 chentianyu03 提交于 12月 08, 2021

2c02a580
F
fix: when ceil_model==true && Padding_algo!=SAME, (x-size)/stride != int, this... · d1ab323f
由 feng_shuai 提交于 12月 08, 2021
```
fix: when ceil_model==true && Padding_algo!=SAME, (x-size)/stride != int, this convert is wrong (#37929)
```
d1ab323f

[Eager] generate eager core ops, only 4 ops (#37813) · 52f63cd2

由 wanghuancoder 提交于 12月 08, 2021

* refine a test case, test=develop

* publish python c api for eager, test=develop

* revert modify about test_allclose_layer.py, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* delete numpy includes, use pybind11 numpy.h, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* suport eager error msg, and add grad test case, test=develop

* refine, test=develop

* refine, test=develop

* generate eager core ops, only 4 ops, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

* refine, test=develop

52f63cd2

Enabled Eager AutoCodeGen for 40+ more operators (#37910) · cf873c39

由 Zhanlue Yang 提交于 12月 08, 2021

* Rearranged Eager AutoCodeGen directory structure

* Removed USE_OP in Eager AutoCodeGen

* Enabled generation for Operators without Grad/Inputs/Outputs

* Resolved operators without input

* Fixed merge conflicts

* Enabled Eager AutoCodeGen for 10+ more operators

cf873c39

Y

fix softmax max dim (#37901) · b5dd12fb
由 Yanxing Shi 提交于 12月 08, 2021

b5dd12fb
S

add pyyaml needed by python\paddle\utils\code_gen\api_gen.py (#37897) · a8f009e4
由 Sing_chan 提交于 12月 08, 2021

a8f009e4
S
Fix CUDA Graph H2D bug by restore host memory (#37774) · a1ad3a63
由 sneaxiy 提交于 12月 08, 2021
```
* fix CUDA Graph H2D bug again

* fix no return bug
```
a1ad3a63

07 12月, 2021 12 次提交
- L
  
  [Fleet Executor] Add feed, fetch and check correctness (#37824) · b8793f70
  由 LiYuRio 提交于 12月 07, 2021
  
  b8793f70
- Y
  introduce INF-RT (#37669) · 70dea138
  由 Yan Chunwei 提交于 12月 07, 2021
```
* add infrt code

refined with Paddle's code style.

* rename CinnRtConfig to InfRtConfig

* rename CinnRt to InfRt of some code

* rename CINNRT to INFRT

* remove unnecessary code

* replace CINN to INFRT in the source code

* replace all "cinn" in code to "infrt"

* remove some const_cast
```
  70dea138
- Z
  Buf fix for reset grad inplace version (#37811) · cf586021
  由 Zhanlue Yang 提交于 12月 07, 2021
```
* Debug

* Fixed issue with reset_grad_inplace_version when used with clear_gradient & cross-batch accumulation

* Rearranged interfaces

* Fixed ci issues
```
  cf586021
- Z
  
  add cmake depend for api_gen.py (#37900) · 7e831b5a
  由 zyfncg 提交于 12月 07, 2021
  
  7e831b5a
- T
  Fix static git diff (#37914) · a754d907
  由 tianshuo78520a 提交于 12月 07, 2021
```
* fix static git diff check

* test=document_fix
```
  a754d907
- D
  
  fix filter_by_instag op for lod_level=0 without lod;test=develop (#37834) · b48545ee
  由 danleifeng 提交于 12月 07, 2021
  
  b48545ee
- J
  multithread memory optimize error fix (#37894) · 6b7b7677
  由 JingZhuangzhuang 提交于 12月 07, 2021
```
* multithread_memory_optimize
```
  6b7b7677
- H
  Set runtime_include_dir in Paddle.__init__.py (#37886) · e3cca8ac
  由 Huihuang Zheng 提交于 12月 07, 2021
```
Paddle don't have to set runtime_include_dir during run CINN.
```
  e3cca8ac
- T
  add some op to xpu2 op list && format xpu op list (#37832) · efd7a229
  由 TTerror 提交于 12月 07, 2021
```
* format xpu op list

* format xpu op list

* update xpu1 op list
```
  efd7a229
- W
  [Eager] fix cmake generate error, and fix circular import (#37871) · 79c25979
  由 wanghuancoder 提交于 12月 07, 2021
```
* refine a test case, test=develop

* rm python, test=develop

* refine, test=develop

* fix cmake generate error, and fix circular import, test=develop
```
  79c25979
- Y
  [Pten]Move func from kernel_context.h into kernel_context.cc (#37804) · bfa0d7f3
  由 YuanRisheng 提交于 12月 07, 2021
```
* add inplace op adaptation

* optimize inplace logic and fix bugs when run kernel that has args of vector<DenseTensor>

* move func in kernel_context.h into kernel_context.cc

* refactor logic that transform variable to densetensor

* fix bugs when compile

* update func name

* fix bugs when run windows-ci
```
  bfa0d7f3
- Z
  [heterps]fix heter service (#37860) · b3185296
  由 zmxdream 提交于 12月 07, 2021
```
* fix heter service. test=develop

* fix heter section worker in debug mode
```
  b3185296

PaddlePaddle / Paddle 大约 1 年 前同步成功

PaddlePaddle / Paddle
大约 1 年前同步成功