提交 · 0d45ac739a6f85afbb00cd83500d10a013bd0477 · PaddlePaddle / Paddle

11 5月, 2023 1 次提交
- 张
  
  昇腾和寒武纪相关代码退场 npu相关代码退场2 (#53568) · 0d45ac73
  由张春乔提交于 5月 11, 2023
  
  0d45ac73
10 5月, 2023 2 次提交

C
[LAUNCH] add log overwrite flag (#53608) · 7f39bcd1
由 Chitsing KUI 提交于 5月 10, 2023
```
* add log overwrite flag

* use strtobool
```
7f39bcd1

由 Roc 提交于 5月 10, 2023

* revert p2p communication for xpu

* pp for npu

* update

* update

* fix xpuplace

* add ut for sync send

* Revert "fix xpuplace"

This reverts commit f89c1d7622426686bc153a3414a42c39e0f4a647.

* add ut for pp sync send

* rm unusable ut

* update

f023d42f

09 5月, 2023 1 次提交
- [Zero-Dim] Support p_norm/reduce_sum_p output 0D (#53421) · e4bf1a8e
  由 zhouweiwei2014 提交于 5月 09, 2023
  
  e4bf1a8e
08 5月, 2023 1 次提交
- J
  
  hack 1-D tensor to Scalar (#53552) · 186f5e0f
  由 JYChen 提交于 5月 08, 2023
  
  186f5e0f
05 5月, 2023 1 次提交
- H
  
  [Dygraph] Fix bugs in dp_pp_comm_overlap for HybridParallel (#53384) · 0d9a23b4
  由 Haohongxiang 提交于 5月 05, 2023
  
  0d9a23b4
04 5月, 2023 1 次提交
- R
  
  revert p2p communication for xpu (#53496) · eda0c588
  由 Roc 提交于 5月 04, 2023
  
  eda0c588
30 4月, 2023 1 次提交
- [Zero-Dim] Support paddle.sum/mean/loss api output 0D,test=allcase (#52739) · ddf94ae4
  由 zhouweiwei2014 提交于 4月 30, 2023
  
  ddf94ae4
28 4月, 2023 2 次提交
- M
  
  replace varbase to relevant name or notes as per the context (#53431) · 96180fff
  由 Meteor Liu 提交于 4月 28, 2023
  
  96180fff
- R
  [KUNLUN] fix pp send /recv on xpu (#53427) · 6d9bbee3
  由 Roc 提交于 4月 28, 2023
```
To make it synchronized at the first recv operator.
If warping all send and recv operators with group start and end, the received tensor will be not complete.
```
  6d9bbee3
27 4月, 2023 1 次提交
- H
  [XPU] remove scale_loss in parallel.py (#53337) · 2e1ac529
  由 houj04 提交于 4月 27, 2023
```
* [XPU] remove scale_loss in parallel.py

* [XPU] throw Unimplemented when using Reducer
```
  2e1ac529
26 4月, 2023 3 次提交
- [Zero-Dim] distributed scatter/all_to_all support input 0D tensor (#53186) · 0b6dd535
  由 zhouweiwei2014 提交于 4月 26, 2023
  
  0b6dd535
- pp 策略调整后，模型转换，以便模型热启 (#52927) · 3650c4a8
  由 zhenhailiu 提交于 4月 26, 2023
```
* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish

* polish
```
  3650c4a8
- S
  
  [HybridParallel]Add segment methods for pipelineparallel (#53344) · c59debe2
  由 ShenLiang 提交于 4月 26, 2023
  
  c59debe2
25 4月, 2023 2 次提交
- W
  
  add mp_sync config. (#53254) · 503f422e
  由 wuhuachaocoding 提交于 4月 25, 2023
  
  503f422e
- C
  [DEBUG] print modifed flags (#53243) · 8d4b64e8
  由 Chitsing KUI 提交于 4月 25, 2023
```
* print modifed flags

* fix ref, opt print

* fix default getter

* fix ut
```
  8d4b64e8
24 4月, 2023 5 次提交

[Zero-Dim] Support paddle.max output 0D, test=allcase (#53242) · 9f9cd919
由 zhouweiwei2014 提交于 4月 24, 2023

9f9cd919

[BugFix] wrong match between depend and c_allreduce_sum (#53089) · f0f58665

由 kangguangli 提交于 4月 24, 2023

* fix bug: wrong match between depend and c_allreduce_sum

* fix codestyle

* fix bug

* add c_sync_calc_stream back

* fix

* revert

* use flag to control

* fix for code coverage

f0f58665

张

Update sharding_optimizer.py (#53164) · fb5fab07
由张春乔提交于 4月 24, 2023

fb5fab07
张

rm mlu (#53194) · 987fb2d8
由张春乔提交于 4月 24, 2023

987fb2d8

[Zero-Dim] Support output 0D for to_tensor. (#52741) · 81fb7df2

由 zqw_1997 提交于 4月 24, 2023

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* test=allcase

* fix doc erros, test=allcase

81fb7df2

23 4月, 2023 1 次提交
- C
  [LAUNCH] save env log for each worker (#53207) · 993bc412
  由 Chitsing KUI 提交于 4月 23, 2023
```
* save env log for each worker

* fix ut
```
  993bc412
22 4月, 2023 1 次提交
- [Zero-Dim] fix functool.reduce more safe with intial value, to support empty list (#53182) · 4da9b87b
  由 zhouweiwei2014 提交于 4月 22, 2023
  
  4da9b87b
21 4月, 2023 3 次提交

support 0-D output and 0-D as indice in __getitem__/__setitem__ (#52814) · 4e939c89

由 JYChen 提交于 4月 21, 2023

* support 0-D output and 0-D as indice in __getitem__

* fix tests

* fix inference and UT

* add unittest for setitem

* fix xpu test

* fix xpu 0-d

4e939c89

Revert "remove ASCEND* keyword" (#53131) · 0f99debd

由 ronnywang 提交于 4月 21, 2023

* Revert "remove ASCEND* keyword (#53046)"

This reverts commit 7fa415ca.

* Delete ascend_trigger_op.cc

* revert-53046-remove_ASCEND_keyword

* update

* update

0f99debd

张

rm _init_npu_pipeline_comm (#53150) · 07878a34
由张春乔提交于 4月 21, 2023

07878a34

20 4月, 2023 3 次提交
- C
  [FlashAttn] add flash randomness control (#52902) · 00ac8014
  由 Chitsing KUI 提交于 4月 20, 2023
```
* add flash randomness control

* fix VLOG undefied
```
  00ac8014
- [Zero-Dim] Support all/any/min/prod/logsumexp/amax/amin/some loss output 0D,test=allcase (#53051) · e6def1eb
  由 zhouweiwei2014 提交于 4月 20, 2023
  
  e6def1eb
- W
  remove ASCEND* keyword (#53046) · 7fa415ca
  由 Wang Xin 提交于 4月 20, 2023
```
* remove ASCEND* keyword

* update docstring

* bug fixed

* bug fixed
```
  7fa415ca
19 4月, 2023 2 次提交
- R
  [CustomDevice] add recompute support (#53044) · 3206fa80
  由 ronnywang 提交于 4月 19, 2023
```
* [CustomDevice] add recompute support

* update
```
  3206fa80
- K
  [Perf] fix static graph performance issue in amp mode with multicard (#52724) · 38ec37cd
  由 kangguangli 提交于 4月 19, 2023
```
* fix

* fix

* fix

* fix

* fix

* fix fuse group order
```
  38ec37cd
18 4月, 2023 2 次提交
- 张
  
  remove mlu(#53007) · 4d5a3ad6
  由张春乔提交于 4月 18, 2023
  
  4d5a3ad6
- M
  rename _varbase_creator as _create_tensor (#52938) · 240e13a2
  由 Meteor Liu 提交于 4月 18, 2023
```
* rename _varbase_creator as create_tensor

* rename _varbase_creator as create_tensor
```
  240e13a2
17 4月, 2023 7 次提交
- Y
  
  [Auto Parallel] Add the micro-bathsize config (#52912) · 94afa5ab
  由 Yulong Ao 提交于 4月 17, 2023
  
  94afa5ab
- L
  cherry-pick fleet executor from 2.4 (#52896) · bafe287a
  由 LiYuRio 提交于 4月 17, 2023
```
* cherry-pick fleet executor from 2.4

* fix test case
```
  bafe287a
- C
  [Fused] controlled randomness for fused dropout add (#52903) · e36f80c6
  由 Chitsing KUI 提交于 4月 17, 2023
```
* add random control for fused dropout add

* add __init__
```
  e36f80c6
- K
  
  rem cncl keyword in py (#52939) · ea04bef8
  由 Kim Yann 提交于 4月 17, 2023
  
  ea04bef8
- 张
  remove hccl in .py files (#52934) · 27a601e8
  由张春乔提交于 4月 17, 2023
```
* remove hccl in .py files

* remove ascend in setup.py.in

* remove ascend in setup.py
```
  27a601e8
- H
  
  [Dygraph] Support delaying div loss by accumulate_steps in PipelineLayer (#52848) · 0abdcff6
  由 Haohongxiang 提交于 4月 17, 2023
  
  0abdcff6
- C
  [Auto Parallel]Add o2 tune of rule based tuner (#52928) · 118a7415
  由 caozhou 提交于 4月 17, 2023
```
* add o2 tune

* add unittest

* fix error

* set unittest timeout
```
  118a7415

PaddlePaddle / Paddle 大约 1 年 前同步成功

PaddlePaddle / Paddle
大约 1 年前同步成功