huangbingjian
2a85af5d83
1. if Switch/SwitchLayer, do not replace Load or remove UpdateState; 2. add control flow testcases; 3. fix codedex problem
4 years ago
huangbingjian
302746d6a3
fix code check
4 years ago
huangbingjian
e5d32c9ff8
add switch_defer_inline
4 years ago
dayschan
771e3f61f3
Clean GraphKernel's codes from frontend
1. set class GraphKernel as deprecated, and treat it as Cell
2. set class InplaceAssign as deprecated, suggested using Assign instead.
3. set op_selector as deprecated, removed the _selected_ops and _selected_grad_ops, replaced with real operations
4. removed the two passes of GraphKernel from frontend
5. removed the GraphKernel's codes from other modules
5 years ago
zhangzhaoju
a1db9c4959
Pybind11::object && PrimitivePy recycle optimize
5 years ago
huangbingjian
1c2430b2bd
update FindPrimalJPair
5 years ago
zhangzhaoju
ef4481c4a3
Memory leak fix, clear using cnt map of graph while funcgraph manager
free
5 years ago
huangbingjian
c8f9d93c2d
modify findPrimalJPair
5 years ago
zhousiyi
74d500a756
If a fv cnode is not mapped but belong to primal_graph_, then don't propagate this fv to caller, it should mapped instead
5 years ago
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
Zhang Qinghua
c621a48533
Refactor grad operation implementation.
5 years ago
lvliang
1490947ff0
adjust-the-location-of-cleaning-unuseless-memory-in-value-node
5 years ago
zjun
67866474f9
Add pynative second derivative function
Signed-off-by: zjun <zhangjun0@huawei.com>
5 years ago
tanghuikang
82450afa9e
Optimize memory using in pynative mode
5 years ago
Zhang Qinghua
829ab4492b
Optimize the performance after Grad due to mapping Primitive to K.
5 years ago
mindspore-ci-bot
0e6d06c463
!9230 Set trace info of Primitive CNode only for equiv out node, not the whole fprop function.
From: @zh_qh
Reviewed-by: @ginfung
Signed-off-by: @ginfung
5 years ago
Zhang Qinghua
8310236ff1
Set trace info of Primitive CNode only for equiv out node, not the whole fprop function.
5 years ago
mindspore-ci-bot
ea191ed5da
!8463 Run ops one by one in pynative bp graph
From: @HulkTang
Reviewed-by:
Signed-off-by:
5 years ago
HulkTang
c36b477568
Run ops one by one in pynative bp graph
5 years ago
huanghui
bda74495f3
Add trace mangager around front opt
5 years ago
mindspore-ci-bot
56a731e48a
!8406 [AutoParallel]Add auto parallel pipeline
From: @lichen666
Reviewed-by:
Signed-off-by:
5 years ago
lichenever
2e1c43483e
add auto parallel pipeline
5 years ago
Zhang Qinghua
bcfa1f72b1
Add debug and trace info for grad operation.
5 years ago
Yi Huaijie
d7faa77b5e
support int64 shape
5 years ago
Zhang Qinghua
e3d62123e5
Bind the fvs with single fg bprop for switch layer.
5 years ago
Zhang Qinghua
2aeff5ba95
Replace primal graph with fprop part of k graph after grad operation.
5 years ago
es chow
93cd019855
add env_item cache for fv_node
5 years ago
root
4e85071055
redundant codes clean
5 years ago
kpy
4338dd266e
optimizer pynative graph memory
5 years ago
buxue
3fd73f9d08
raise exception when use HookBackward in graph mode
5 years ago
mindspore-ci-bot
88f5cbe5db
!5692 Add requires_grad option for python pass
Merge pull request !5692 from BowenK/pre_ad
5 years ago
kingfo
cfda024336
fix op id issue in pynative
5 years ago
BowenK
1bdb26f9e8
Warming up python pass by adding inline passes before it
5 years ago
mindspore-ci-bot
65819fb50c
!5488 heck arg is tensor when graph output is a const or parameter
Merge pull request !5488 from zhangbuxue/add_check_tensor_before_compile_graph
5 years ago
buxue
4168c51f0d
add check tensor when graph output is a const or parameter
5 years ago
fary86
fcbb3e0edc
Refactor ms_context implementation
5 years ago
kingfo
1de829aeee
resolve output twice out of memory
5 years ago
zhoufeng
22ec976ef7
fix compilation warnings
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
5 years ago
zhoufeng
663278112f
optimize code compile performance
Signed-off-by: zhoufeng <zhoufeng54@huawei.com>
5 years ago
mindspore-ci-bot
f875bf21bc
!2948 fix control flow
Merge pull request !2948 from amongo/FixControlFlow
5 years ago
huangdongrun
2a6d346d2f
support if by if grad parameter
add join for ref
adjust env eliminate to eliminate all env ops
add partial app cache
resolve while endless
fix env eliminate
support for "for while" cases
fix join shape error
5 years ago
panyifeng
1c296b96c9
fix switch layer sigle prim cell
5 years ago
liubuyu
45b6600001
mv cc file that end with py to pybind
5 years ago
liubuyu
d81862a916
decoupling core and context
5 years ago
liubuyu
a499d4e47c
decoupling core and debug
5 years ago
liubuyu
f4bc0bc9fe
move the dependency of utils to core
5 years ago
kpy
570da089a8
set output value for dynamic graph
5 years ago
panyifeng
4d4e23fd9e
Add bprop for sparse_tensor
5 years ago
liubuyu
76dc80e7b7
Unified code style
5 years ago
panyifeng
5a10383cc3
add coo_tensor
5 years ago