tanghuikang
|
dac64f30ee
|
Support ms_function + heterogenous
|
5 years ago |
caifubi
|
171b468bb3
|
PyNative AllReduce Bucket
|
5 years ago |
zuochuanyong
|
3fa26683ac
|
nlp perf(Pynative): change memory sync mode from synchronous to asynchronous in SyncHostToDevice
|
5 years ago |
lizhenyu
|
f17534af08
|
ps cache support sparse
|
5 years ago |
yanghaitao1
|
8d147deb07
|
profiler memory
|
5 years ago |
yanghaoran
|
b8345d03b6
|
Synchronize latest Ascend software 18 Dec 2020, with profiler fixes
|
5 years ago |
limingqi107
|
a844d52b42
|
add ps cache check
|
5 years ago |
lizhenyu
|
e3f7ae61db
|
add ps cache manager
|
5 years ago |
HulkTang
|
c36b477568
|
Run ops one by one in pynative bp graph
|
5 years ago |
mindspore-ci-bot
|
0d49650bd5
|
!8888 Add PyNative Device Profiling
From: @jojobugfree
Reviewed-by: @kisnwang,@jjfeing
Signed-off-by: @jjfeing
|
5 years ago |
caifubi
|
12643a33dd
|
Add profiling reporter
|
5 years ago |
kswang
|
62ae6802dc
|
fix context null error
|
5 years ago |
Harshvardhan Gupta
|
744355a005
|
remove dbg args from runtime and remove needless argument from IsWatchpoint
|
5 years ago |
caifubi
|
9b76f4ed57
|
get physical device for gpu
|
5 years ago |
caifubi
|
d3b978147f
|
Ascend Dynamic Shape
|
5 years ago |
kswang
|
11989b5e30
|
enable async run
|
5 years ago |
Harshvardhan Gupta
|
7c5e0541ba
|
load inputs before suspending execution in dbg
|
5 years ago |
laiyongqiang
|
bd8aeefd95
|
disable memory reuse for selected op in e2e dump
|
5 years ago |
mindspore-ci-bot
|
c543db0585
|
!6180 clean codex warning in memreuse
Merge pull request !6180 from laiyongqiang/codex
|
5 years ago |
laiyongqiang
|
4063a69846
|
clean codex warning
|
5 years ago |
caifubi
|
372c2e7951
|
Combine Async Dump and E2E Dump
|
5 years ago |
lichen_101010
|
dffa61b228
|
send info when training is done
|
5 years ago |
John Tzanakakis
|
b0a7ebdeb0
|
enable debugger by default and set correct log message severity
|
5 years ago |
mindspore-ci-bot
|
0c316e522d
|
!5866 clean idle mem at proper time
Merge pull request !5866 from liangzelang/fix_global_step_error
|
5 years ago |
liangzelang
|
4c7291078c
|
clean idle mem in the beginning of ascend session
|
5 years ago |
Zhang Qinghua
|
c0070d3d49
|
Use the unified Execute function to run Graph or Single Op Graph.
|
5 years ago |
Zhang Qinghua
|
5ac60ff202
|
Move ascend dependent functions to ascend kernel runtime.
|
5 years ago |
limingqi107
|
7ec2f6a550
|
clear graph output address in graph destructor
|
5 years ago |
limingqi107
|
7029a861d7
|
add kernel release resource
|
5 years ago |
liubuyu
|
d81862a916
|
decoupling core and context
|
5 years ago |
z00505269
|
87668d6ea2
|
remove predict
|
5 years ago |
lvchangquan
|
fdbe4c19ba
|
use kernel_runtime::mem_manager to reduce rtMalloc and rtFree time in trans data format
|
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 |
John Tzanakakis
|
b3c0eb61d5
|
GPU debugger - milestone 1 and GPU dump
Additonal Authors: Adel Shafiei, Harshvardhan Gupta
|
5 years ago |
mindspore-ci-bot
|
72a2b7d496
|
!3117 not reuse ref node input's memory
Merge pull request !3117 from laiyongqiang/refnode_input
|
5 years ago |
mindspore-ci-bot
|
bae2f964e5
|
!3213 Unified code style
Merge pull request !3213 from liubuyu/dev
|
5 years ago |
laiyongqiang
|
acba03b191
|
not reuse ref node input's memory
|
5 years ago |
liubuyu
|
76dc80e7b7
|
Unified code style
|
5 years ago |
lvchangquan
|
7b48a122dd
|
insert trans_data to reduce time in print process
|
5 years ago |
laiyongqiang
|
68c78ab6bb
|
reuse communication op output's memory
|
5 years ago |
liubuyu
|
43c79eb853
|
mindspore path adjust
|
5 years ago |