He Wei
7d9a783993
[auto-monad] Support side-effects by auto-monad
The basic idea is: exploits data dependency to control the execution order
of side-effect operations, and keep the semantics of ANF unchanged.
The ControlDepend primitive is removed and there are two primitives added:
1. UpdateState:
```
a = Assign(para, value)
```
became:
```
a = Assign(para, value, u)
u = UpdateState(u, a)
```
2. Load:
```
x = Add(para, value)
```
became:
```
p = Load(para, u)
x = Add(p, value)
u = UpdateState(u, p)
```
5 years ago
l00591931
9ec100d069
Change TensorAdd to Add, from r1.1 to master
5 years ago
jjfeing
27257b9901
add trace when mindspore error
5 years ago
Yi Huaijie
d7faa77b5e
support int64 shape
5 years ago
huanghui
e6eb89e58c
fix MatMul's third input not in all_nodes after matmul_biasadd fusion
5 years ago
jzg
7cbd55e17d
add embedding layer.
5 years ago
caifubi
d3b978147f
Ascend Dynamic Shape
5 years ago
huanghui
90dfecfb00
fix bnupdate_eltwise_eltwise's cycle fusion
5 years ago
huanghui
d6944a70ca
fix cpu kernel:ScatterNdUpdate doesn't set output
5 years ago
mindspore-ci-bot
43c1147843
!5623 Fix run error when there is a Depend or ControlDepend on BatchNorm
Merge pull request !5623 from huanghui/bn-infer
5 years ago
mindspore-ci-bot
3048240f16
!5508 Add AdamApplyOneWithDecayAssign fusion pass
Merge pull request !5508 from YuJianfeng/adam_assign
5 years ago
huanghui
b8e737f66a
fix run error when there is a Depend or ControlDepend on BatchNorm
5 years ago
yujianfeng
4b77f6b53c
Add AdamApplyOneWithDecayAssign fusion pass
5 years ago
huanghui
998ff0399b
clear codex and reviewbot warning
5 years ago
WilliamLian
601b0b6e4d
remove convert datatype when updateoutputs &&
set parameter device dtype using it's infer dtype && set transdata's abstract
5 years ago
zhousiyi
d0e58dd765
remove ccsrc/common.h
replace frontend/operator/ops.h in backend with base/core_ops.h as
backend should not use any frontend-only primitive
5 years ago
mindspore-ci-bot
58523a41fe
!4392 use builtin float16 for arm
Merge pull request !4392 from xychow/use-float16-in-arm-neon
5 years ago
zhoufeng
663278112f
optimize code compile performance
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
5 years ago
zhousiyi
e1aa49a4b7
use built-in float16 in arm_neon.h for lite arm
5 years ago
yujianfeng
8a77751988
Add AdamApplyOneAssign and AdamApplyOneWithDecayAssign fusion pass
5 years ago
mindspore-ci-bot
52689a7dcf
!3938 decoupling core and context
Merge pull request !3938 from liubuyu/master
5 years ago
zhoufeng
ca7154a548
graph compile performance optimization
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
5 years ago
liubuyu
d81862a916
decoupling core and context
5 years ago
huanghui
b25e114840
add op mapping attr for those pass worked in LeNet
5 years ago
liubuyu
76dc80e7b7
Unified code style
5 years ago
He Wei
ee1510da41
Eliminate circular dependency between 'ir' and 'device/kernel'
5 years ago
liubuyu
43c79eb853
mindspore path adjust
5 years ago