提交 · 668a0a41325b0ff2b47677b1c800cefa62b6a161 · PaddlePaddle / Paddle

10 5月, 2022 19 次提交

[Eager] Refactor several sharding test (#42608) · 668a0a41

由 Weilong Wu 提交于 5月 10, 2022

* [Eager] fix sharding under eager mode

* [Eager] fix several sharding test under eager mode

* Recover using _test_eager_guard

* Ensured fleet.init under legacy

* Ensured fleet.init under legacy

* Fix CI issue, re-definition strategy and call fleet.init() in stage2_offload

* Modified dygraph_group_sharded_api.py, move fleet.init to a better line

668a0a41

fix switch client multithread bug (#42600) · e2540c17

由 ziyoujiyi 提交于 5月 10, 2022

* back fl

* delete ssl cert

* .

* make warning

* .

* unittest paral degree

* solve unittest

* heter & multi cloud commm ready

* .

* .

* arm_brpc compile

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* only output is ok

* base is ok

* .

* .

* .

* .

* .

* .

* .

* .

* add switch server bin

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* adapt brpc ssl

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* .

* fix heter_server & heter_client

* .

* .

* int->int64_t

* .

* safe map in multithread

* fix heter unitest

* .

* fix code_style

* .

* fix bug

* .

e2540c17

shape mkldnn kernel adapted to NHWC (#42548) · d47690b2

由 Jacek Czaja 提交于 5月 10, 2022

* - shape mkldnn adapted to NHWC

- NHWC shape mkldnn ut

- fixes to UT

- Fix to UT

- Fixes to UT

- Fix of compilation

* - lint candidate fix

d47690b2

J

update RandomErasing document (#42627) · 08158e93
由 JYChen 提交于 5月 10, 2022

08158e93

[Video detection] Added fill_constant FP32 FWD oneDNN kernel (#37216) · 66a10f36

由 jakpiase 提交于 5月 10, 2022

* added fill_constant kernel

* CI fix

* ci fix

* switched from nan to zero memory

* CI FIX

* ci fixes

* CI rerun

* ci fix

* minor change

* CI rerun

66a10f36

【PaddlePaddle Hackathon 2】18、为 Paddle 新增 paddle.heaviside 和 paddle.Tensor.heaviside API (#41872) · 4892d592

由 BrilliantYuKaimin 提交于 5月 10, 2022

* Create elementwise_heaviside_op.cc

* add ElementwiseHeavisideFunctor

* Create test_elementwise_heaviside_op.py

* 增加heaviside的python接口

* add heaviside in white list

* 增加heaviside的签名

* 增加heaviside的核函数

* 增加heaviside梯度的核函数

* 增加heaviside梯度的注册

* 调整代码格式

* Update elementwise_sig.cc

* add heaviside in __all__

* Update heaviside docs

* Update math.py

* Update math.py

* Update math.py

4892d592

W
[Eager] print gpu mem info (#42616) · 81644145
由 wanghuancoder 提交于 5月 10, 2022
```
* print mem

* refine

* refine

* refine

* refine
```
81644145
L

add int8 for cast (#42634) · 8a100774
由 lilong12 提交于 5月 10, 2022

8a100774
C

update base of cost model (#42601) · 6ac08db5
由 caozhou 提交于 5月 10, 2022

6ac08db5
Q

[MLU]add adam, adamw op of mlu device (#42557) · cc077693
由 qipengh 提交于 5月 10, 2022

cc077693
F

[MLU] add layernorm mlu kernel (#42356) · ecd6db43
由 fwenguang 提交于 5月 10, 2022

ecd6db43
Q

[MLU]add assign op of mlu device (#42591) · 4e5fb733
由 qipengh 提交于 5月 10, 2022

4e5fb733
Z
fix adamw unittest (#42593) · c6f49f0b
由 zhaoyingli 提交于 5月 10, 2022
```
* fix adamw unittest

* tiny fix

* fix param name
```
c6f49f0b
J
pdnode_compare (#42597) · 30234dd7
由 JingZhuangzhuang 提交于 5月 10, 2022
```
* pdnode_compare

* panode compare

* pdnode_compare
```
30234dd7
Z

merge develop. test=develop (#42624) · 0ce42fb0
由 zmxdream 提交于 5月 10, 2022

0ce42fb0
L

fix bug for heter (#42590) · 21b35167
由 lilong12 提交于 5月 10, 2022

21b35167
S

fix sample error (#42595) · df96d1ed
由 Siming Dai 提交于 5月 10, 2022

df96d1ed
A
fix random cache (#723) (#42621) · be87caf2
由 Allen Guo 提交于 5月 10, 2022
```
Co-authored-by: Nyaozhixin <522190855@qq.com>
```
be87caf2
S

broadcast_add kp performance optimization (#42097) · c7855125
由 shixingbo 提交于 5月 10, 2022

c7855125

09 5月, 2022 13 次提交

A
[Eager]Fix tensor.name is empty behavior (#42587) · 81078a88
由 Aurelius84 提交于 5月 09, 2022
```
* [Eager]Fix tensor.name is empty behavior

* fix unittest
```
81078a88
W
refine pylayer (#42572) · c22c2c58
由 wanghuancoder 提交于 5月 09, 2022
```
* refine pylayer

* refine
```
c22c2c58
L
fix docs of auto_cast, cuda_places, static.save (#42107) · c3b7bc61
由 Liyulingyue 提交于 5月 09, 2022
```
* auto_cast; test=document_fix

* static.save; test=document_fix

* cuda_places; test=document_fix
```
c3b7bc61
J
[Need approval] Add AdamW-CPU FP32 JIT assembly kernel (#42522) · 766c50ac
由 joanna.wozna.intel 提交于 5月 09, 2022
```
* Add AdamW jit kernel

* Second implementation

* Add missing header

* Correct number of jit kernels in the test
```
766c50ac

[Ready to merge] oneDNN NHWC matmul & elementwise kernels fixes (#42506) · bf481550

由 Jacek Czaja 提交于 5月 09, 2022

* - fix to crash

- more fixes

- added diagnostic

- matmul output fixes.

- compilation fix

- stop rotating too small shapes

* - Added enabling of matmul_V2 onednn test

bf481550

N

Modified reduce for xpu2 (#42439) · ae4d1ec1
由 niuliling123 提交于 5月 09, 2022

ae4d1ec1

double grad yaml and test case (#42553) · 8b546f1c

由 chentianyu03 提交于 5月 09, 2022

* add abs double grad yaml and test case

* add pool2d double grad yaml

* add pool2d dygraph double grad test case

8b546f1c

C
[Eager] Polish grad code details (#42536) · 778ea4ec
由 Chen Weihang 提交于 5月 09, 2022
```
* polish grad details

* polish detail by comment
```
778ea4ec
W
[Eager] Fix several sharding test under eager mode (#42573) · 13bcb7cd
由 Weilong Wu 提交于 5月 09, 2022
```
* [Eager] fix sharding under eager mode

* [Eager] fix several sharding test under eager mode
```
13bcb7cd
C

fix split stride_numel may be 0 (#42537) · d878f971
由 chentianyu03 提交于 5月 09, 2022

d878f971

【PaddlePaddle Hackathon 2】3、为 Paddle 新增 corrcoef(皮尔逊积矩相关系数) API (#40690) · 95a502a2

由 liqitong-a 提交于 5月 09, 2022

* corrcoef commit

* corrcoef commit

* Update test_corr.py

* Update linalg.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update linalg.py

* Update linalg.py

* Update linalg.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

* Update test_corr.py

95a502a2

Q

[ROCm] fix rocksdb on ROCm version 40020496, test=develop (#42563) · bba5e083
由 Qi Li 提交于 5月 09, 2022

bba5e083
J
[Eager] Support Gradient Accumulation for sr (#42371) · 1cddcd70
由 Jiabin Yang 提交于 5月 09, 2022
```
* Support Gradient Accumulation for sr

* add ut

* change ut to fit small vector
```
1cddcd70

07 5月, 2022 8 次提交
- Z
  
  fix bug of optional_tensor in amp logic (#42561) · 4e66010b
  由 zhangbo9674 提交于 5月 07, 2022
  
  4e66010b
- C
  put_record_event_in_python_on_timeline_python (#42555) · 80015c06
  由 chenjian 提交于 5月 07, 2022
```
* put_record_event_in_python_on_timeline_python

* fix
```
  80015c06
- Z
  
  fix the problem of slice infer shape (#42568) · c1e45a11
  由 zyfncg 提交于 5月 07, 2022
  
  c1e45a11
- Q
  
  [ROCm] add gfx908 support for AMD MI100, test=develop (#42560) · d1aedd58
  由 Qi Li 提交于 5月 07, 2022
  
  d1aedd58
- Z
  [Phi] Change sync copy to async for gpu_pinned to gpu place in data transform (#41966) · 6583a8d2
  由 zyfncg 提交于 5月 07, 2022
```
* the copy type of data transform for gpu_pinned to gpu change from syna to async

* refactor code
```
  6583a8d2
- W
  
  add some no need buff (#42556) · 3f372814
  由 wanghuancoder 提交于 5月 07, 2022
  
  3f372814
- P
  [dockerfile] add cuda11.6、cuda11.5、cuda11.4、cuda11.3 manylinux docker (#41251) · efa21b12
  由 pangyoki 提交于 5月 07, 2022
```
* add cuda11.5 manylinux docker

* build.sh

* fix

* fix

* support cu113 and cu114

* add --no-check-certificate when wget sqlite-autoconf-3250300

* change cuda11.4.2 to cuda11.4.3

* add cu116
```
  efa21b12
- Y
  [Auto Parallel] Improve the codes of the completion and distributed context (#40671) · bed9aaea
  由 Yulong Ao 提交于 5月 07, 2022
```
* [Auto Parallel] Replace the old planner by the new partition tuner

* [Auto Parallel] Improve the completion and distributed context

* [Auto Parallel] Fix some bugs of the compatible check of some dist ops

* [Auto Parallel] Fix some bugs
```
  bed9aaea

PaddlePaddle / Paddle 大约 1 年 前同步成功

PaddlePaddle / Paddle
大约 1 年前同步成功