提交 · 7babb3d231f30678e47c82b866deb21635552a58 · Crayon鑫 / Paddle

16 9月, 2021 10 次提交
- W
  
  c and go api support tuned trt. (#35771) · 7babb3d2
  由 Wilber 提交于 9月 16, 2021
  
  7babb3d2
- L
  remove distributed attributes at the last stage for auto parallel (#35605) · a3790606
  由 lilong12 提交于 9月 16, 2021
```
* update
```
  a3790606
- L
  
  rename the auto parallel suffix (#35765) · 2c70b844
  由 lilong12 提交于 9月 16, 2021
  
  2c70b844
- Z
  
  add a fusion op: fused_dropout_act_bias (#35129) · cee70434
  由 zhangkaihuo 提交于 9月 16, 2021
  
  cee70434
- W
  Python support register pass via PassDesc (#35602) · bab39eb2
  由 wuhuanzhou 提交于 9月 16, 2021
```
PR主要功能：针对fusion等子图替换场景，支持Python侧开发并注册Pass。

背景
Pass是指输入一个深度学习计算图Graph，依照一定条件进行修改，输出修改后的Graph的过程；
当前PaddlePadle框架编写Pass代码存在以下问题：
用户需要手写Graph的条件匹配、在Graph上的修改代码；
对Graph操作需要深入底层框架代码，了解Graph的结构，并且知道相关Pass写法；
我们提出了针对fusion等子图替换类Pass的优化方案以支持用户在Python侧开发注册Pass，提升二次开发体验：
用户只需要输入匹配和替换的子图描述，由深度学习框架编写的代码来生成匹配和替换的逻辑，不需要用户对Graph进行匹配和替换操作；
API级别的替换，用户可以通过Paddle的Python API构造子图，从而不需要知道Graph的结构，也能写Paddle的Graph Pass代码
```
  bab39eb2
- W
  
  add run interface for standalone executor, test=develop (#35761) · 29ef7cc9
  由 wanghuancoder 提交于 9月 15, 2021
  
  29ef7cc9
- W
  
  [hybrid] remove scale op in insert_scale_loss_grad_ops (#35775) · 02b0be08
  由 WangXi 提交于 9月 16, 2021
  
  02b0be08
- W
  
  fix pass quickly for PR-CI-OP-Benchmark, test=develop (#35753) · 1dcf22a8
  由 wuhuanzhou 提交于 9月 16, 2021
  
  1dcf22a8
- S
  
  fix py api for paddle.inference.contrib.utils (#35769) · 554771dd
  由 Shang Zhizhou 提交于 9月 16, 2021
  
  554771dd
- Z
  
  Add a new op: paddle.linalg.multi_dot (#35224) · c9f7cff0
  由 zhangkaihuo 提交于 9月 16, 2021
  
  c9f7cff0
15 9月, 2021 30 次提交
- S
  Fix docs in stream_guard API · 72b07726
  由 Siming Dai 提交于 9月 15, 2021
```
Fix docs in stream_guard API 
```
  72b07726
- S
  
  set proxy off when download vis_demo (#35754) · 572021aa
  由 Sing_chan 提交于 9月 15, 2021
  
  572021aa
- J
  Fix for slice OneDNN kernel in solov2 and ppyolo models (#35706) · 9d996cdd
  由 jakpiase 提交于 9月 15, 2021
```
* fixed slice error

* added handling of StartsTensor+List and EndsTensor+List

* fix for ppyolo model
```
  9d996cdd
- J
  
  added fix for matmul and support for 6 rank tensor (#35740) · e80acff3
  由 jakpiase 提交于 9月 15, 2021
  
  e80acff3
- W
  add inplace logic into new_executor (#35618) · bd79ae09
  由 wanghuancoder 提交于 9月 15, 2021
```
* add inplace logic into new_executor, test=develop

* check shape and add inplace FLAGS, test=develop

* refine, test=develop

* refine, test=develop
```
  bd79ae09
- 王
  clip op extra information when export model. (#35447) · 4d236354
  由王明冬提交于 9月 15, 2021
```
* clip op extra information when export model,test=ocr

* rename clip_extra parameter to kwargs in save_inference_model, test=ocr
```
  4d236354
- Z
  Change the invoking method of settiem from numpy to set_value op when value isn't tensor (#35701) · 86d4af39
  由 zyfncg 提交于 9月 15, 2021
```
* Change the invoking method of settiem from numpy to set_value op when value is not tensor

* fix the check logic for inplace in setitem

* fix the unittest problem caused by setitem doesn't support fp16

* modify some code format in setitem
```
  86d4af39
- Z
  add dist_attr for dist op and var (#35585) · fc5fb2a1
  由 zhaoyingli 提交于 9月 15, 2021
```
* add dist_attr for dist op

* add unitest

* update inputname

* update function name

* add unitest

* update CMakeLists.txt for CI

* fix dis_matmul

* fix compile error

* update matmul to matmul_v2
```
  fc5fb2a1
- H
  
  fix bugs of PR 35401 (#35746) · 09eaa7d7
  由 Haohongxiang 提交于 9月 15, 2021
  
  09eaa7d7
- L
  add nvidia cusparse library, test=develop (#35675) · 7e18106a
  由 Liu-xiandong 提交于 9月 15, 2021
```
Put Nvidia's cusparse library into paddle.
```
  7e18106a
- P
  [NPU] add beam_search npu op (#34860) · 3760be06
  由 pangyoki 提交于 9月 15, 2021
```
* add beam_search npu op

* fix CMakeList and add unittest

* fix bug of beam search npu op

* fix unittest

* let input ids become int64

* set output ids to int64_t

* delete check_dygraph

* fix beam_width=1
```
  3760be06
- W
  support numpy.ndarray index. (#35748) · 9f588cc2
  由 WeiXin 提交于 9月 15, 2021
```
* support numpy.ndarray index.

* polish code.
```
  9f588cc2
- S
  
  upgrade dice_loss (#35734) · 46ec5b3e
  由 shangliang Xu 提交于 9月 15, 2021
  
  46ec5b3e
- H
  
  add set-xpu-device-id function for inference config. (#35572) · a74d7fb6
  由 houj04 提交于 9月 15, 2021
  
  a74d7fb6
- S
  
  disable mkldnn_tester_ppyolo_mbv3 for it always fail in cuda11.1 windows build pipe (#35729) · c192127b
  由 Sing_chan 提交于 9月 15, 2021
  
  c192127b
- Q
  [NPU] fix depthwise_conv2d_grad, test=develop (#35626) · d3e06a51
  由 Qi Li 提交于 9月 15, 2021
```
* [NPU] fix depthwise_conv2d_grad, test=develop

* remove debug files, test=develop
```
  d3e06a51
- Y
  
  Unify the functor definition of elementwise add, sub, mul, div, floordiv, max, min. (#35684) · 2367cca6
  由 Yiqun Liu 提交于 9月 15, 2021
  
  2367cca6
- J
  Add gelu convert test (#35529) · 39dcfc6c
  由 JingZhuangzhuang 提交于 9月 15, 2021
```
Co-authored-by: Nxiaoxiaohehe001 <hiteezsf@163.com>
```
  39dcfc6c
- Y
  Add New OP: gumbel_softmax (#35506) · 18eda6c3
  由 YuanRisheng 提交于 9月 15, 2021
```
* Add New Op: gumbel_softmax

* Add New Op: gumbel_softmax

* Add New Op: gumbel_softmax (amend)

* add __main__ function in unit test

* fix bugs when test in windows ci

* update en docs

* delete reletive error in unit test

* delete relative error in unit test

* set hard=True in unit test
```
  18eda6c3
- S
  Add paddle.cuda.device.stream_guard API (#35623) · 3218075d
  由 Siming Dai 提交于 9月 15, 2021
```
Add paddle.cuda.device.stream_guard API 
```
  3218075d
- G
  
  fix dim check of class center sample (#35733) · a9577347
  由 Guoxia Wang 提交于 9月 15, 2021
  
  a9577347
- W
  
  [hybrid] out data parallel as optimizer sharding parallel (#35593) · 78465703
  由 WangXi 提交于 9月 15, 2021
  
  78465703
- L
  
  Refactor dropout cuda impl for code reuse. (#35621) · 2b88057f
  由 Li Min 提交于 9月 15, 2021
  
  2b88057f
- X
  [Paddle Inference]Add split op TRT converter unittest. (#35127) · e26a2504
  由 xiaoxiaohehe001 提交于 9月 15, 2021
```
* add_split_op

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller

* add_split_teller
```
  e26a2504
- X
  [Paddle Inference]Add Transpose op TRT converter unittest (#35138) · 09f920a2
  由 xiaoxiaohehe001 提交于 9月 15, 2021
```
* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller

* add_transpose_teller
```
  09f920a2
- X
  [Paddle Inference]Add scale TRT converter unittest. (#35225) · c563609a
  由 xiaoxiaohehe001 提交于 9月 15, 2021
```
* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller

* add_scale_teller
```
  c563609a
- A
  Enhance Check mechanism and Support single tuple/list of fetch_list in Executor (#35726) · cae050e8
  由 Aurelius84 提交于 9月 15, 2021
```
* Enhance Check mechanism of fetch_list in Executor

* support single tuple

* fix typo

* fix typo
```
  cae050e8
- 津
  
  add test (#35316) · 9f4d201a
  由津提交于 9月 15, 2021
  
  9f4d201a
- 津
  
  add test (#35405) · 4cf1255c
  由津提交于 9月 15, 2021
  
  4cf1255c
- 津
  
  add test (#35100) · 09a48965
  由津提交于 9月 15, 2021
  
  09a48965

Crayon鑫 / Paddle 与 Fork 源项目一致

Crayon鑫 / Paddle
与 Fork 源项目一致