Skip to content
项目
群组
代码片段
帮助
当前项目
正在载入...
登录 / 注册
切换导航面板
P
pytensor
项目
项目
详情
活动
周期分析
仓库
仓库
文件
提交
分支
标签
贡献者
图表
比较
统计图
议题
0
议题
0
列表
看板
标记
里程碑
合并请求
0
合并请求
0
CI / CD
CI / CD
流水线
作业
日程
统计图
Wiki
Wiki
代码片段
代码片段
成员
成员
折叠边栏
关闭边栏
活动
图像
聊天
创建新问题
作业
提交
问题看板
Open sidebar
testgroup
pytensor
Commits
b6f3c469
提交
b6f3c469
authored
6月 11, 2014
作者:
abergeron
浏览文件
操作
浏览文件
下载
差异文件
Merge pull request #1911 from nouiz/cycle
Fix crash du to opt that introduce cycle in the graph
上级
34824493
a52b4e2e
全部展开
隐藏空白字符变更
内嵌
并排
正在显示
5 个修改的文件
包含
29 行增加
和
8 行删除
+29
-8
destroyhandler.py
theano/gof/destroyhandler.py
+0
-0
opt.py
theano/gof/opt.py
+20
-1
optdb.py
theano/gof/optdb.py
+6
-6
check_blas.py
theano/misc/check_blas.py
+2
-0
opt.py
theano/sandbox/cuda/opt.py
+1
-1
没有找到文件。
theano/gof/destroyhandler.py
浏览文件 @
b6f3c469
差异被折叠。
点击展开。
theano/gof/opt.py
浏览文件 @
b6f3c469
...
...
@@ -593,8 +593,27 @@ class MergeOptimizer(Optimizer):
pairs_list
=
sched
.
pop
()
success
=
True
for
pairs
in
pairs_list
:
# We must check again the equivalence, as the graph
# can have changed. If so, doing the replacement can
# introduce node that depend on itself. Doing the
# full check of such cycle everytimes is very time
# consumming. I think this double check is faster then
# doing the full cycle check. The full cycle check is
# skipped by validate() if the graph don't contain
# destroyers.
node
=
pairs
[
0
][
0
]
candidate
=
pairs
[
0
][
1
]
if
node
.
owner
and
candidate
.
owner
:
node
=
node
.
owner
candidate
=
candidate
.
owner
inputs_match
=
all
(
node_in
is
cand_in
for
node_in
,
cand_in
in
zip
(
node
.
inputs
,
candidate
.
inputs
))
# No need to compare the op again, as it don't change.
if
not
inputs_match
:
continue
try
:
fgraph
.
replace_all_validate
(
pairs
,
'Merge'
)
fgraph
.
replace_all_validate
(
pairs
,
'Merge
Optimizer
'
)
except
InconsistencyError
:
success
=
False
nb_fail
+=
1
...
...
theano/gof/optdb.py
浏览文件 @
b6f3c469
...
...
@@ -41,19 +41,19 @@ class DB(object):
raise
ValueError
(
'The name of the object cannot be an existing'
' tag or the name of another existing object.'
,
obj
,
name
)
if
self
.
name
is
not
None
:
tags
=
tags
+
(
self
.
name
,)
obj
.
name
=
name
# This restriction is there because in many place we suppose that
# something in the DB is there only once.
if
getattr
(
obj
,
'name'
,
""
)
in
self
.
__db__
:
if
obj
.
name
in
self
.
__db__
:
raise
ValueError
(
'''You can
\'
t register the same optimization
multiple time in a DB. Tryed to register "
%
s" again under the new name "
%
s".
Use theano.gof.ProxyDB to work around that'''
%
(
obj
.
name
,
name
))
if
self
.
name
is
not
None
:
tags
=
tags
+
(
self
.
name
,)
obj
.
name
=
name
self
.
__db__
[
name
]
=
set
([
obj
])
self
.
_names
.
add
(
name
)
self
.
__db__
[
obj
.
__class__
.
__name__
]
.
add
(
obj
)
self
.
add_tags
(
name
,
*
tags
)
def
add_tags
(
self
,
name
,
*
tags
):
...
...
theano/misc/check_blas.py
浏览文件 @
b6f3c469
...
...
@@ -204,6 +204,7 @@ if __name__ == "__main__":
cuda version 6.0 5.5 5.0 4.2 4.1 4.0 3.2 3.0 # note
gpu
K6000/NOECC 0.06s
K40 0.07s
K20m/ECC 0.07s
K20/NOECC 0.07s
M2090 0.19s
...
...
@@ -213,6 +214,7 @@ if __name__ == "__main__":
M2070-Q 0.48s 0.27s 0.32s
M2050(Amazon) 0.25s
C1060 0.46s
K600 1.04s
GTX Titan Black 0.05s
GTX Titan(D15U-50) 0.06s 0.06s don't work
...
...
theano/sandbox/cuda/opt.py
浏览文件 @
b6f3c469
...
...
@@ -115,7 +115,7 @@ class InputToGpuOptimizer(Optimizer):
if
new_input
.
type
==
input
.
type
:
fgraph
.
replace_validate
(
input
,
new_input
,
"InputToGpuOptimizer"
)
"InputToGpuOptimizer"
)
except
TypeError
:
#as we currently only support float32, this can fail.
#Using try except make that we won't need
...
...
编写
预览
Markdown
格式
0%
重试
或
添加新文件
添加附件
取消
您添加了
0
人
到此讨论。请谨慎行事。
请先完成此评论的编辑!
取消
请
注册
或者
登录
后发表评论