lingyunli63
8b3823b22c
optimizeMatmul
5 years ago
mindspore-ci-bot
ce38856029
!14451 fusion AtomicAddrClean node
From: @zhaosida_hw
Reviewed-by: @zhoufeng54,@jjfeing
Signed-off-by: @jjfeing
5 years ago
hwjiaorui
dac67cbabb
clean code
5 years ago
yuchaojie
dc9a8eb1cc
fix codedex warning
5 years ago
mindspore-ci-bot
dfe505a7f0
!14481 fix transdata attr not matched with build info when it has been spilted
From: @lianliguang
Reviewed-by: @kisnwang,@zhoufeng54
Signed-off-by: @zhoufeng54
5 years ago
mindspore-ci-bot
609b8edb9f
!14466 cleaning codex warnings
From: @nicholas_yhr
Reviewed-by: @kisnwang,@zhoufeng54
Signed-off-by: @zhoufeng54
5 years ago
LianLiguang
9c8d016d66
fix transdata's dst format && src format is unmatched with build info when transdata has been spilted
5 years ago
mindspore-ci-bot
36dbb2690e
!14442 fix codex
From: @lianliguang
Reviewed-by: @zhoufeng54,@chujinjin
Signed-off-by: @chujinjin
5 years ago
mindspore-ci-bot
47c6d91762
!14369 refactor processor setting
From: @wenfangpei
Reviewed-by: @gaoxiong1,@anyrenwei
Signed-off-by: @anyrenwei
5 years ago
mindspore-ci-bot
d0dd75c4b5
!14364 [GraphKernel]Remove depend stuff because of its order implication.
From: @tronzhang
Reviewed-by: @gaoxiong1,@anyrenwei
Signed-off-by: @anyrenwei
5 years ago
yanghaoran
50e81568b5
cleaning codex warnings
5 years ago
zhaosida
1740aac860
fusion atomic clear node
5 years ago
LianLiguang
9b94b6f8ed
fix codex && reviewbot
5 years ago
wenfangpei
4af448bdd3
refactor processor setting
5 years ago
mindspore-ci-bot
c50bdbeea8
!14264 add BatchMatMul&FusedMulAdd, BatchMatmul&ConfusionTranpose UB fusion pass
From: @yuchaojie
Reviewed-by: @zhoufeng54,@jjfeing
Signed-off-by: @jjfeing
5 years ago
mindspore-ci-bot
69526df01e
!14314 [GraphKernel] unify graph kernel pass add_atomic_clean on Ascend and GPU back-end
From: @looop5
Reviewed-by: @gaoxiong1,@gaoxiong1,@dylangeng
Signed-off-by: @dylangeng
5 years ago
mindspore-ci-bot
ddf75da542
!14085 [GraphKernel] add some expander ops
From: @chenlei_autodiff
Reviewed-by:
Signed-off-by:
5 years ago
mindspore-ci-bot
75fdaaa6aa
!14304 [GraphKernel] Dump GraphKernel split info as text; dump akg kernel launch fail message
From: @dayschan
Reviewed-by: @gaoxiong1,@gaoxiong1,@anyrenwei
Signed-off-by: @anyrenwei
5 years ago
looop5
76d322464d
unify graph kernel pass add_atomic_clean on Ascend and GPU back-end
refactor CanActivateAtomicAdd
use smart pointer
5 years ago
chenlei_autodiff
f4289d40f3
add graph kernel expander ops.
5 years ago
dayschan
3c6c30024c
dump graph_kernel_split info
5 years ago
mindspore-ci-bot
993d00c4ba
!13774 The input and output of batchnorm reverse operator increase pass in ascend platform under the mode of pynitve
From: @ding_fei_fei
Reviewed-by: @kingxian,@zhunaipan
Signed-off-by: @kingxian
5 years ago
yuchaojie
50f7f6b3de
add BatchMatMul&FusedMulAdd, BatchMatmul&ConfusionTranpose UB fusion pass
5 years ago
tronzhang
4182c1f02a
delete depend format for its order use
5 years ago
mindspore-ci-bot
7149e8c2c9
!14045 [Graph Kernel] add compare test case
From: @zengzitao
Reviewed-by: @gaoxiong1
Signed-off-by:
5 years ago
mindspore-ci-bot
15ed2daa31
!14255 [GraphKernel] open Assign fusion and fix OptimizeAssign
From: @lingyunli63
Reviewed-by: @gaoxiong1,@dylangeng
Signed-off-by: @dylangeng
5 years ago
mindspore-ci-bot
e644a66973
!14237 [GraphKernel] infer processor from ms_context inside function CreateCNode
From: @looop5
Reviewed-by: @gaoxiong1,@dylangeng
Signed-off-by: @dylangeng
5 years ago
mindspore-ci-bot
7f4994af7c
!14186 Support while bprop
From: @liangzelang
Reviewed-by: @kisnwang,@jjfeing
Signed-off-by: @jjfeing
5 years ago
lingyunli63
932027253f
fuse Assign and fix optimizeAssign
5 years ago
mindspore-ci-bot
a06234ed13
!13834 [GraphKernel] Remove two graph kernel frontend pass
From: @tronzhang
Reviewed-by: @gaoxiong1,@gaoxiong1,@dylangeng
Signed-off-by: @dylangeng
5 years ago
mindspore-ci-bot
203106c133
!14198 [GraphKernel] Refactor GraphKernelExpander (4th submission)
From: @dayschan
Reviewed-by: @gaoxiong1,@dylangeng
Signed-off-by: @dylangeng
5 years ago
dingpeifei
3c9d8cb073
The input and output of batchnorm reverse operator increase pass in ascend platform under the mode of pynitve
5 years ago
looop5
dd43972839
infer processor from ms_context inside function CreateCNode
5 years ago
zengzitao
72c6dad4ba
add compare_test case in gpu ci and update akg submodule
5 years ago
mindspore-ci-bot
72bad339e7
!14065 New add grad accumulation for network.
From: @linqingke
Reviewed-by: @guoqi1024,@xu-yfei
Signed-off-by: @guoqi1024
5 years ago
linqingke
7b48d059c5
new add grad accumulation for network.
5 years ago
tronzhang
87bf1ec80f
delete mark_interface_fusion and tensor reuse frontend pass for graph kernel
5 years ago
dayschan
56c4145cc5
Refactor GraphKernelExpander (4th submission)
Decoupled the process of expanding node from GraphKernelExpander pass,
so that it can be rolled back when error occurs.
By the way, supported controlling cluster ops by flags.
5 years ago
mindspore-ci-bot
ad140a8bf4
!14084 [GraphKernel] support matmul on D
From: @lingyunli63
Reviewed-by:
Signed-off-by:
5 years ago
liangzelang
ba65fb9f3c
Support non-tail recursive graphs
5 years ago
lingyunli63
4b966ed40d
support matmul on D
5 years ago
mindspore-ci-bot
7f92400ebd
!14068 [GraphKernel] fix graph_split and checkcircle
From: @lingyunli63
Reviewed-by: @gaoxiong1,@dylangeng
Signed-off-by: @dylangeng
5 years ago
mindspore-ci-bot
0909d69b4e
!14067 fix graph compile time over 3h
From: @zyli2020
Reviewed-by: @cristoval,@limingqi107
Signed-off-by: @limingqi107
5 years ago
lingyunli63
8832248205
fix abstract null and bug in checkcircle
5 years ago
lizhenyu
d2598d73be
fix graph compiling cost time over 3h
5 years ago
mindspore-ci-bot
946e403df8
!13701 dynamic memory pool support multi-thread
From: @limingqi107
Reviewed-by: @cristoval,@kisnwang
Signed-off-by: @kisnwang
5 years ago
mindspore-ci-bot
18e98c6a0b
!13720 【GraphKernel】Add context graph_kernel_flags
From: @dayschan
Reviewed-by: @gaoxiong1
Signed-off-by:
5 years ago
dayschan
11ee3b1624
add context graph_kernel_flags
used the flag "opt_level" to control GraphKernel,
0 means disabled while non-zero value means enabled.
the default value is controlled by context "enable_graph_kernel",
but if it's also set in "graph_kernel_flags", then the flag will prevail.
supported the whitelist and blacklist operators for GraphKernelExpander.
"enable_expand_ops", "enable_expand_ops_only", "disable_expand_ops".
5 years ago
liuxiao93
723bbac438
revert nn.BatchNorm3d.
5 years ago
yuchaojie
e769ee95ce
reduce cast number in syncbn
5 years ago