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
816a83e3
提交
816a83e3
authored
3月 07, 2013
作者:
nouiz
浏览文件
操作
浏览文件
下载
差异文件
Merge pull request #1271 from lamblin/arch_in_cache
Always put arch bitwidth in cache
上级
3805afa3
1d3ba7ec
隐藏空白字符变更
内嵌
并排
正在显示
10 个修改的文件
包含
172 行增加
和
119 行删除
+172
-119
config.txt
doc/library/config.txt
+1
-1
test_pfunc.py
theano/compile/tests/test_pfunc.py
+1
-1
test_shared.py
theano/compile/tests/test_shared.py
+1
-1
__init__.py
theano/gof/__init__.py
+3
-2
cmodule.py
theano/gof/cmodule.py
+40
-60
compiledir.py
theano/gof/compiledir.py
+38
-9
utils.py
theano/gof/utils.py
+75
-31
nvcc_compiler.py
theano/sandbox/cuda/nvcc_compiler.py
+2
-4
extra_ops.py
theano/tensor/extra_ops.py
+4
-4
test_extra_ops.py
theano/tensor/tests/test_extra_ops.py
+7
-6
没有找到文件。
doc/library/config.txt
浏览文件 @
816a83e3
...
...
@@ -364,7 +364,7 @@ import theano and print the config variable, as in:
.. attribute:: compiledir_format
Default: "compiledir_%(platform)s-%(processor)s-%(python_version)s"
Default: "compiledir_%(platform)s-%(processor)s-%(python_version)s
-%(python_bitwidth)s
"
This is a Python format string that specifies the subdirectory
of ``config.base_compiledir`` in which to store platform-dependent
...
...
theano/compile/tests/test_pfunc.py
浏览文件 @
816a83e3
...
...
@@ -548,7 +548,7 @@ class Test_pfunc(unittest.TestCase):
def
test_default_updates_input
(
self
):
x
=
shared
(
0
)
y
=
shared
(
1
)
if
theano
.
gof
.
cmodule
.
python_int_bitwidth
()
==
32
:
if
theano
.
gof
.
python_int_bitwidth
()
==
32
:
a
=
iscalar
(
'a'
)
else
:
a
=
lscalar
(
'a'
)
...
...
theano/compile/tests/test_shared.py
浏览文件 @
816a83e3
...
...
@@ -18,7 +18,7 @@ class Test_SharedVariable(unittest.TestCase):
assert
shared
(
7
,
dtype
=
'float64'
)
.
type
==
Scalar
(
'float64'
)
else
:
if
theano
.
gof
.
cmodule
.
python_int_bitwidth
()
==
32
:
if
theano
.
gof
.
python_int_bitwidth
()
==
32
:
assert
shared
(
7
)
.
type
==
theano
.
tensor
.
iscalar
,
shared
(
7
)
.
type
else
:
assert
shared
(
7
)
.
type
==
theano
.
tensor
.
lscalar
,
shared
(
7
)
.
type
...
...
theano/gof/__init__.py
浏览文件 @
816a83e3
...
...
@@ -38,7 +38,9 @@ e-mail thread "What is gof?"
from
theano.gof.cc
import
\
CLinker
,
OpWiseCLinker
,
DualLinker
import
theano.gof.compiledir
# adds config vars
# Also adds config vars
from
theano.gof.compiledir
import
\
local_bitwidth
,
python_int_bitwidth
from
theano.gof.fg
import
\
InconsistencyError
,
MissingInputError
,
FunctionGraph
...
...
@@ -77,4 +79,3 @@ from theano.gof.type import \
from
theano.gof.utils
import
\
object2
,
MethodNotDefined
theano/gof/cmodule.py
浏览文件 @
816a83e3
...
...
@@ -8,7 +8,6 @@ import os
import
shutil
import
stat
import
StringIO
import
struct
import
subprocess
import
sys
import
tempfile
...
...
@@ -27,7 +26,7 @@ from theano.misc.windows import call_subprocess_Popen
# we will abuse the lockfile mechanism when reading and writing the registry
from
theano.gof
import
compilelock
from
theano.gof.compiledir
import
gcc_version_str
from
theano.gof.compiledir
import
gcc_version_str
,
local_bitwidth
from
theano.configparser
import
AddConfigVar
,
BoolParam
...
...
@@ -55,29 +54,6 @@ AddConfigVar('cmodule.compilation_warning',
BoolParam
(
False
))
def
local_bitwidth
():
"""
Return 32 for 32bit arch, 64 for 64bit arch
By "architecture", we mean the size of memory pointers (size_t in C),
*not* the size of long int, as it can be different.
"""
# Note that according to Python documentation, `platform.architecture()` is
# not reliable on OS X with universal binaries.
# Also, sys.maxsize does not exist in Python < 2.6.
# 'P' denotes a void*, and the size is expressed in bytes.
return
struct
.
calcsize
(
'P'
)
*
8
def
python_int_bitwidth
():
"""
Return the bit width of Python int (C long int).
Note that it can be different from the size of a memory pointer.
"""
# 'l' denotes a C long int, and the size is expressed in bytes.
return
struct
.
calcsize
(
'l'
)
*
8
_logger
=
logging
.
getLogger
(
"theano.gof.cmodule"
)
_logger
.
setLevel
(
logging
.
WARNING
)
...
...
@@ -176,14 +152,14 @@ static struct PyModuleDef moduledef = {{
}};
"""
.
format
(
name
=
self
.
name
)
print
>>
stream
,
"PyMODINIT_FUNC PyInit_
%
s(void) {"
%
self
.
name
for
b
in
self
.
init_blocks
:
print
>>
stream
,
' '
,
b
for
b
lock
in
self
.
init_blocks
:
print
>>
stream
,
' '
,
b
lock
print
>>
stream
,
" PyObject *m = PyModule_Create(&moduledef);"
print
>>
stream
,
" return m;"
else
:
print
>>
stream
,
"PyMODINIT_FUNC init
%
s(void){"
%
self
.
name
for
b
in
self
.
init_blocks
:
print
>>
stream
,
' '
,
b
for
b
lock
in
self
.
init_blocks
:
print
>>
stream
,
' '
,
b
lock
print
>>
stream
,
' '
,
(
'(void) Py_InitModule("
%
s", MyMethods);'
%
self
.
name
)
print
>>
stream
,
"}"
...
...
@@ -1564,7 +1540,8 @@ class GCC_compiler(object):
lines
=
stdout
+
stderr
return
lines
# The '-' at the end is needed. Otherwise, g++ do not output enough information.
# The '-' at the end is needed. Otherwise, g++ do not output
# enough information.
native_lines
=
get_lines
(
"g++ -march=native -E -v -"
)
_logger
.
info
(
"g++ -march=native selected lines:
%
s"
,
native_lines
)
if
len
(
native_lines
)
!=
1
:
...
...
@@ -1619,6 +1596,39 @@ class GCC_compiler(object):
cxxflags
.
append
(
"-D NPY_ARRAY_UPDATE_ALL=NPY_UPDATE_ALL"
)
cxxflags
.
append
(
"-D NPY_ARRAY_C_CONTIGUOUS=NPY_C_CONTIGUOUS"
)
cxxflags
.
append
(
"-D NPY_ARRAY_F_CONTIGUOUS=NPY_F_CONTIGUOUS"
)
# Platform-specific flags.
# We put them here, rather than in compile_str(), so they en up
# in the key of the compiled module, avoiding potential conflicts.
# Figure out whether the current Python executable is 32
# or 64 bit and compile accordingly.
n_bits
=
local_bitwidth
()
cxxflags
.
append
(
'-m
%
d'
%
n_bits
)
_logger
.
debug
(
"Compiling for
%
s bit architecture"
,
n_bits
)
if
sys
.
platform
!=
'win32'
:
# Under Windows it looks like fPIC is useless. Compiler warning:
# '-fPIC ignored for target (all code is position independent)'
cxxflags
.
append
(
'-fPIC'
)
if
sys
.
platform
==
'win32'
and
local_bitwidth
()
==
64
:
# Under 64-bit Windows installation, sys.platform is 'win32'.
# We need to define MS_WIN64 for the preprocessor to be able to
# link with libpython.
cxxflags
.
append
(
'-DMS_WIN64'
)
#DSE Patch 1 for supporting OSX frameworks; add -framework Python
if
sys
.
platform
==
'darwin'
:
cxxflags
.
extend
([
'-undefined'
,
'dynamic_lookup'
])
python_inc
=
distutils
.
sysconfig
.
get_python_inc
()
# link with the framework library *if specifically requested*
# config.mac_framework_link is by default False, since on some mac
# installs linking with -framework causes a Bus Error
if
(
python_inc
.
count
(
'Python.framework'
)
>
0
and
config
.
cmodule
.
mac_framework_link
):
cxxflags
.
extend
([
'-framework'
,
'Python'
])
return
cxxflags
@staticmethod
...
...
@@ -1744,40 +1754,10 @@ class GCC_compiler(object):
else
:
preargs
=
list
(
preargs
)
if
sys
.
platform
!=
'win32'
:
# Under Windows it looks like fPIC is useless. Compiler warning:
# '-fPIC ignored for target (all code is position independent)'
preargs
.
append
(
'-fPIC'
)
if
sys
.
platform
==
'win32'
and
local_bitwidth
()
==
64
:
# Under 64-bit Windows installation, sys.platform is 'win32'.
# We need to define MS_WIN64 for the preprocessor to be able to
# link with libpython.
preargs
.
append
(
'-DMS_WIN64'
)
# We also add "-m64", in case the installed gcc is 32-bit
preargs
.
append
(
'-m64'
)
include_dirs
=
include_dirs
+
std_include_dirs
()
libs
=
std_libs
()
+
libs
lib_dirs
=
std_lib_dirs
()
+
lib_dirs
#DSE Patch 1 for supporting OSX frameworks; add -framework Python
if
sys
.
platform
==
'darwin'
:
preargs
.
extend
([
'-undefined'
,
'dynamic_lookup'
])
python_inc
=
distutils
.
sysconfig
.
get_python_inc
()
# link with the framework library *if specifically requested*
# config.mac_framework_link is by default False, since on some mac
# installs linking with -framework causes a Bus Error
if
(
python_inc
.
count
(
'Python.framework'
)
>
0
and
config
.
cmodule
.
mac_framework_link
):
preargs
.
extend
([
'-framework'
,
'Python'
])
# Figure out whether the current Python executable is 32
# or 64 bit and compile accordingly.
n_bits
=
local_bitwidth
()
preargs
.
extend
([
'-m
%
s'
%
n_bits
])
_logger
.
debug
(
"OS X: compiling for
%
s bit architecture"
,
n_bits
)
# sometimes, the linker cannot find -lpython so we need to tell it
# explicitly where it is located
# this returns somepath/lib/python2.x
...
...
theano/gof/compiledir.py
浏览文件 @
816a83e3
...
...
@@ -4,6 +4,7 @@ import os
import
platform
import
re
import
shutil
import
struct
import
subprocess
import
sys
import
textwrap
...
...
@@ -32,16 +33,44 @@ except OSError:
del
p
del
dummy_err
compiledir_format_dict
=
{
"platform"
:
platform
.
platform
(),
"processor"
:
platform
.
processor
(),
"python_version"
:
platform
.
python_version
(),
"theano_version"
:
theano
.
__version__
,
"numpy_version"
:
numpy
.
__version__
,
"gxx_version"
:
gcc_version_str
.
replace
(
" "
,
"_"
),
}
def
local_bitwidth
():
"""
Return 32 for 32bit arch, 64 for 64bit arch
By "architecture", we mean the size of memory pointers (size_t in C),
*not* the size of long int, as it can be different.
"""
# Note that according to Python documentation, `platform.architecture()` is
# not reliable on OS X with universal binaries.
# Also, sys.maxsize does not exist in Python < 2.6.
# 'P' denotes a void*, and the size is expressed in bytes.
return
struct
.
calcsize
(
'P'
)
*
8
def
python_int_bitwidth
():
"""
Return the bit width of Python int (C long int).
Note that it can be different from the size of a memory pointer.
"""
# 'l' denotes a C long int, and the size is expressed in bytes.
return
struct
.
calcsize
(
'l'
)
*
8
compiledir_format_dict
=
{
"platform"
:
platform
.
platform
(),
"processor"
:
platform
.
processor
(),
"python_version"
:
platform
.
python_version
(),
"python_bitwidth"
:
local_bitwidth
(),
"python_int_bitwidth"
:
python_int_bitwidth
(),
"theano_version"
:
theano
.
__version__
,
"numpy_version"
:
numpy
.
__version__
,
"gxx_version"
:
gcc_version_str
.
replace
(
" "
,
"_"
),
}
compiledir_format_keys
=
", "
.
join
(
sorted
(
compiledir_format_dict
.
keys
()))
default_compiledir_format
=
\
"compiledir_
%(platform)
s-
%(processor)
s-
%(python_version)
s"
default_compiledir_format
=
(
"compiledir_
%(platform)
s-
%(processor)
s-"
"
%(python_version)
s-
%(python_bitwidth)
s"
)
AddConfigVar
(
"compiledir_format"
,
textwrap
.
fill
(
textwrap
.
dedent
(
"""
\
...
...
theano/gof/utils.py
浏览文件 @
816a83e3
# import op
# import variable
import
re
import
traceback
from
theano
import
config
import
re
,
traceback
def
add_tag_trace
(
thing
):
"""Add tag.trace to an node or variable.
...
...
@@ -11,15 +10,18 @@ def add_tag_trace(thing):
The argument is returned after being affected (inplace).
"""
limit
=
config
.
traceback
.
limit
if
limit
==
-
1
:
limit
=
None
if
limit
==
-
1
:
limit
=
None
thing
.
tag
.
trace
=
traceback
.
extract_stack
(
limit
=
limit
)[:
-
1
]
return
thing
def
hashgen
():
hashgen
.
next
+=
1
return
hashgen
.
next
hashgen
.
next
=
0
class
MethodNotDefined
(
Exception
):
"""
To be raised by functions defined as part of an interface.
...
...
@@ -28,6 +30,7 @@ class MethodNotDefined(Exception):
function has been left out of an implementation class.
"""
class
object2
(
object
):
__slots__
=
[]
if
0
:
...
...
@@ -36,23 +39,30 @@ class object2(object):
if
hasattr
(
self
,
'__eq__'
)
or
hasattr
(
self
,
'__cmp__'
):
raise
TypeError
(
"unhashable object:
%
s"
%
self
)
return
id
(
self
)
def
__ne__
(
self
,
other
):
return
not
self
==
other
class
scratchpad
:
def
clear
(
self
):
self
.
__dict__
.
clear
()
def
__update__
(
self
,
other
):
self
.
__dict__
.
update
(
other
.
__dict__
)
return
self
def
__str__
(
self
):
return
"scratchpad"
+
str
(
self
.
__dict__
)
def
__repr__
(
self
):
return
"scratchpad"
+
str
(
self
.
__dict__
)
def
info
(
self
):
print
"<theano.gof.utils.scratchpad instance at
%
i>"
%
id
(
self
)
for
k
,
v
in
self
.
__dict__
.
items
():
print
"
%
s:
%
s"
%
(
k
,
v
)
print
"<theano.gof.utils.scratchpad instance at
%
i>"
%
id
(
self
)
for
k
,
v
in
self
.
__dict__
.
items
():
print
"
%
s:
%
s"
%
(
k
,
v
)
class
D
:
def
__init__
(
self
,
**
d
):
...
...
@@ -63,6 +73,7 @@ def memoize(f):
"""Cache the return value for each tuple of arguments
(which must be hashable) """
cache
=
{}
def
rval
(
*
args
,
**
kwargs
):
kwtup
=
tuple
(
kwargs
.
items
())
key
=
(
args
,
kwtup
)
...
...
@@ -72,8 +83,8 @@ def memoize(f):
else
:
val
=
cache
[
key
]
return
val
return
rval
return
rval
def
deprecated
(
filename
,
msg
=
''
):
...
...
@@ -92,6 +103,7 @@ def deprecated(filename, msg=''):
"""
def
_deprecated
(
f
):
printme
=
[
True
]
def
g
(
*
args
,
**
kwargs
):
if
printme
[
0
]:
print
'WARNING:
%
s.
%
s deprecated.
%
s'
\
...
...
@@ -99,19 +111,23 @@ def deprecated(filename, msg=''):
printme
[
0
]
=
False
return
f
(
*
args
,
**
kwargs
)
return
g
return
_deprecated
def
uniq
(
seq
):
#TODO: consider building a set out of seq so that the if condition is constant time -JB
#TODO: consider building a set out of seq so that the if condition
#is constant time -JB
return
[
x
for
i
,
x
in
enumerate
(
seq
)
if
seq
.
index
(
x
)
==
i
]
def
difference
(
seq1
,
seq2
):
"""
Returns all elements in seq1 which are not in seq2: i.e seq1
\
seq2
"""
try
:
# try to use O(const * len(seq1)) algo
if
len
(
seq2
)
<
4
:
# I'm guessing this threshold -JB
if
len
(
seq2
)
<
4
:
# I'm guessing this threshold -JB
raise
Exception
(
'not worth it'
)
set2
=
set
(
seq2
)
return
[
x
for
x
in
seq1
if
x
not
in
set2
]
...
...
@@ -132,13 +148,16 @@ def partition(f, seq):
seqf
.
append
(
elem
)
return
seqt
,
seqf
def
attr_checker
(
*
attrs
):
def
f
(
candidate
):
for
attr
in
attrs
:
if
not
hasattr
(
candidate
,
attr
):
return
False
return
True
f
.
__doc__
=
"Checks that the candidate has the following attributes:
%
s"
%
", "
.
join
([
"'
%
s'"
%
attr
for
attr
in
attrs
])
f
.
__doc__
=
(
"Checks that the candidate has the following attributes:
%
s"
%
", "
.
join
([
"'
%
s'"
%
attr
for
attr
in
attrs
]))
return
f
...
...
@@ -149,11 +168,10 @@ def all_bases(cls, accept):
return
[
cls
for
cls
in
rval
if
accept
(
cls
)]
def
all_bases_collect
(
cls
,
raw_name
):
rval
=
set
()
name
=
"__
%
s__"
%
raw_name
if
name
in
cls
.
__dict__
:
# don't use hasattr
if
name
in
cls
.
__dict__
:
# don't use hasattr
rval
.
add
(
getattr
(
cls
,
name
))
cut
=
"__
%
s_override__"
%
raw_name
if
not
cls
.
__dict__
.
get
(
cut
,
False
):
...
...
@@ -162,7 +180,7 @@ def all_bases_collect(cls, raw_name):
return
rval
def
camelcase_to_separated
(
string
,
sep
=
"_"
):
def
camelcase_to_separated
(
string
,
sep
=
"_"
):
return
re
.
sub
(
'(.)([A-Z])'
,
'
\\
1
%
s
\\
2'
%
sep
,
string
)
.
lower
()
...
...
@@ -172,6 +190,7 @@ def to_return_values(values):
else
:
return
values
def
from_return_values
(
values
):
if
isinstance
(
values
,
(
list
,
tuple
)):
return
values
...
...
@@ -186,7 +205,8 @@ class ClsInit(type):
Validate and initialize the L{Op} subclass 'cls'
This function:
- changes class attributes input_names and output_names to be lists if they are single strings.
- changes class attributes input_names and output_names to be lists
if they are single strings.
"""
type
.
__init__
(
cls
,
name
,
bases
,
dct
)
...
...
@@ -195,8 +215,10 @@ class ClsInit(type):
def
toposort
(
prereqs_d
):
"""
Sorts prereqs_d.keys() topologically. prereqs_d[x] contains all the elements
that must come before x in the ordering.
Sorts prereqs_d.keys() topologically.
prereqs_d[x] contains all the elements that must come before x
in the ordering.
"""
# all1 = set(prereqs_d.keys())
...
...
@@ -223,19 +245,26 @@ def toposort(prereqs_d):
if
not
prereqs_d
[
postreq
]
.
difference
(
done
):
next
.
add
(
postreq
)
if
len
(
prereqs_d
)
!=
len
(
seq
):
raise
Exception
(
"Cannot sort topologically: there might be cycles, "
+
\
"prereqs_d does not have a key for each element or "
+
\
raise
Exception
(
"Cannot sort topologically: there might be cycles, "
"prereqs_d does not have a key for each element or "
"some orderings contain invalid elements."
)
return
seq
def
print_for_dot
(
self
):
#TODO: popen2("dot -Tpng | display") and actually make the graph window pop up
print
"digraph unix { size = '6,6'; node [color = lightblue2; style = filled];"
for
op
in
self
.
order
:
for
input
in
op
.
inputs
:
if
input
.
owner
:
print
input
.
owner
.
__class__
.
__name__
+
str
(
abs
(
id
(
input
.
owner
))),
" -> "
,
op
.
__class__
.
__name__
+
str
(
abs
(
id
(
op
))),
";"
#TODO: popen2("dot -Tpng | display") and actually make the graph window
#pop up
print
(
"digraph unix { size = '6,6'; node [color = lightblue2;"
"style = filled];"
)
for
op
in
self
.
order
:
for
input
in
op
.
inputs
:
if
input
.
owner
:
print
' '
.
join
((
input
.
owner
.
__class__
.
__name__
+
str
(
abs
(
id
(
input
.
owner
))),
" -> "
,
op
.
__class__
.
__name__
+
str
(
abs
(
id
(
op
))),
";"
))
class
Keyword
:
...
...
@@ -263,9 +292,11 @@ simple_types = (int, float, str, bool, None.__class__, Keyword)
ANY_TYPE
=
Keyword
(
"ANY_TYPE"
)
FALL_THROUGH
=
Keyword
(
"FALL_THROUGH"
)
def
comm_guard
(
type1
,
type2
):
def
wrap
(
f
):
old_f
=
f
.
func_globals
[
f
.
__name__
]
def
new_f
(
arg1
,
arg2
,
*
rest
):
if
(
type1
is
ANY_TYPE
or
isinstance
(
arg1
,
type1
))
\
and
(
type2
is
ANY_TYPE
or
isinstance
(
arg2
,
type2
)):
...
...
@@ -283,6 +314,7 @@ def comm_guard(type1, type2):
return
variable
new_f
.
__name__
=
f
.
__name__
def
typename
(
type
):
if
isinstance
(
type
,
Keyword
):
return
str
(
type
)
...
...
@@ -290,14 +322,19 @@ def comm_guard(type1, type2):
return
"("
+
", "
.
join
([
x
.
__name__
for
x
in
type
])
+
")"
else
:
return
type
.
__name__
new_f
.
__doc__
=
str
(
old_f
.
__doc__
)
+
"
\n
"
+
", "
.
join
([
typename
(
type
)
for
type
in
(
type1
,
type2
)])
+
"
\n
"
+
str
(
f
.
__doc__
or
""
)
new_f
.
__doc__
=
(
str
(
old_f
.
__doc__
)
+
"
\n
"
+
", "
.
join
([
typename
(
type
)
for
type
in
(
type1
,
type2
)])
+
"
\n
"
+
str
(
f
.
__doc__
or
""
))
return
new_f
return
wrap
def
type_guard
(
type1
):
def
wrap
(
f
):
old_f
=
f
.
func_globals
[
f
.
__name__
]
def
new_f
(
arg1
,
*
rest
):
if
(
type1
is
ANY_TYPE
or
isinstance
(
arg1
,
type1
)):
variable
=
f
(
arg1
,
*
rest
)
...
...
@@ -308,8 +345,8 @@ def type_guard(type1):
else
:
return
old_f
(
arg1
,
*
rest
)
new_f
.
__name__
=
f
.
__name__
def
typename
(
type
):
if
isinstance
(
type
,
Keyword
):
return
str
(
type
)
...
...
@@ -317,8 +354,12 @@ def type_guard(type1):
return
"("
+
", "
.
join
([
x
.
__name__
for
x
in
type
])
+
")"
else
:
return
type
.
__name__
new_f
.
__doc__
=
str
(
old_f
.
__doc__
)
+
"
\n
"
+
", "
.
join
([
typename
(
type
)
for
type
in
(
type1
,)])
+
"
\n
"
+
str
(
f
.
__doc__
or
""
)
new_f
.
__doc__
=
(
str
(
old_f
.
__doc__
)
+
"
\n
"
+
", "
.
join
([
typename
(
type
)
for
type
in
(
type1
,)])
+
"
\n
"
+
str
(
f
.
__doc__
or
""
))
return
new_f
return
wrap
...
...
@@ -331,15 +372,18 @@ def flatten(a):
else
:
return
[
a
]
def
unique
(
x
):
return
len
(
set
(
x
))
==
len
(
x
)
def
hist
(
coll
):
counts
=
{}
for
elem
in
coll
:
counts
[
elem
]
=
counts
.
get
(
elem
,
0
)
+
1
return
counts
def
give_variables_names
(
variables
):
""" Gives unique names to an iterable of variables. Modifies input.
...
...
@@ -349,10 +393,10 @@ def give_variables_names(variables):
bad_var
=
lambda
var
:
not
var
.
name
or
h
[
var
.
name
]
>
1
for
i
,
var
in
enumerate
(
filter
(
bad_var
,
variables
)):
var
.
name
=
(
var
.
name
or
""
)
+
"_
%
d"
%
i
var
.
name
=
(
var
.
name
or
""
)
+
"_
%
d"
%
i
if
not
unique
(
map
(
str
,
variables
)):
raise
ValueError
(
"Not all variables have unique names."
"Maybe you've named some of the variables identically"
)
"Maybe you've named some of the variables identically"
)
return
variables
theano/sandbox/cuda/nvcc_compiler.py
浏览文件 @
816a83e3
...
...
@@ -9,11 +9,11 @@ import warnings
import
numpy
import
theano
from
theano.gof
import
local_bitwidth
from
theano.gof.cc
import
hash_from_file
from
theano.gof.cmodule
import
(
std_libs
,
std_lib_dirs
,
std_include_dirs
,
dlimport
,
get_lib_extension
,
local_bitwidth
)
get_lib_extension
)
from
theano.gof.python25
import
any
from
theano.misc.windows
import
call_subprocess_Popen
...
...
@@ -245,8 +245,6 @@ class NVCC_compiler(object):
cppfile
=
file
(
cppfilename
,
'w'
)
_logger
.
debug
(
'Writing module C++ code to
%
s'
,
cppfilename
)
ofiles
=
[]
rval
=
None
cppfile
.
write
(
src_code
)
cppfile
.
close
()
...
...
theano/tensor/extra_ops.py
浏览文件 @
816a83e3
...
...
@@ -114,12 +114,12 @@ class BinCountOp(theano.Op):
# Some dtypes are not supported by numpy's implementation of bincount.
# Until another one is available, we should fail at graph construction
# time, not wait for execution.
int_bitwidth
=
theano
.
gof
.
cmodule
.
python_int_bitwidth
()
int_bitwidth
=
theano
.
gof
.
python_int_bitwidth
()
if
int_bitwidth
==
64
:
numpy_unsupported_dtypes
=
(
'uint64'
,)
if
int_bitwidth
==
32
:
numpy_unsupported_dtypes
=
(
'uint32'
,
'int64'
,
'uint64'
)
intp_bitwidth
=
theano
.
gof
.
cmodule
.
local_bitwidth
()
intp_bitwidth
=
theano
.
gof
.
local_bitwidth
()
if
intp_bitwidth
==
32
:
out_type
=
basic
.
ivector
()
elif
intp_bitwidth
==
64
:
...
...
@@ -246,7 +246,7 @@ class RepeatOp(theano.Op):
# Some dtypes are not supported by numpy's implementation of repeat.
# Until another one is available, we should fail at graph construction
# time, not wait for execution.
int_bitwidth
=
theano
.
gof
.
cmodule
.
python_int_bitwidth
()
int_bitwidth
=
theano
.
gof
.
python_int_bitwidth
()
if
int_bitwidth
==
64
:
numpy_unsupported_dtypes
=
(
'uint64'
,)
if
int_bitwidth
==
32
:
...
...
@@ -259,7 +259,7 @@ class RepeatOp(theano.Op):
%
numpy_unsupported_dtypes
),
repeats
.
dtype
)
if
self
.
axis
is
None
:
broadcastable
=
[
False
]
broadcastable
=
[
False
]
else
:
try
:
const_reps
=
basic
.
get_scalar_constant_value
(
repeats
)
...
...
theano/tensor/tests/test_extra_ops.py
浏览文件 @
816a83e3
...
...
@@ -13,6 +13,7 @@ from theano import config, tensor, function
numpy_ver
=
[
int
(
n
)
for
n
in
numpy
.
__version__
.
split
(
'.'
)[:
2
]]
numpy_16
=
bool
(
numpy_ver
>=
[
1
,
6
])
class
TestBinCountOp
(
utt
.
InferShapeTester
):
def
setUp
(
self
):
super
(
TestBinCountOp
,
self
)
.
setUp
()
...
...
@@ -25,7 +26,7 @@ class TestBinCountOp(utt.InferShapeTester):
'uint8'
,
'uint16'
,
'uint32'
,
'uint64'
):
# uint64 always fails
# int64 and uint32 also fail if python int are 32-bit
int_bitwidth
=
theano
.
gof
.
cmodule
.
python_int_bitwidth
()
int_bitwidth
=
theano
.
gof
.
python_int_bitwidth
()
if
int_bitwidth
==
64
:
numpy_unsupported_dtypes
=
(
'uint64'
,)
if
int_bitwidth
==
32
:
...
...
@@ -57,7 +58,7 @@ class TestBinCountOp(utt.InferShapeTester):
for
dtype
in
tensor
.
discrete_dtypes
:
# uint64 always fails
# int64 and uint32 also fail if python int are 32-bit
int_bitwidth
=
theano
.
gof
.
cmodule
.
python_int_bitwidth
()
int_bitwidth
=
theano
.
gof
.
python_int_bitwidth
()
if
int_bitwidth
==
64
:
numpy_unsupported_dtypes
=
(
'uint64'
,)
if
int_bitwidth
==
32
:
...
...
@@ -188,7 +189,6 @@ class SqueezeTester(utt.InferShapeTester):
def
test_grad
(
self
):
for
shape
,
broadcast
in
zip
(
self
.
shape_list
,
self
.
broadcast_list
):
data
=
numpy
.
random
.
random
(
size
=
shape
)
.
astype
(
theano
.
config
.
floatX
)
variable
=
tensor
.
TensorType
(
theano
.
config
.
floatX
,
broadcast
)()
utt
.
verify_grad
(
self
.
op
,
[
data
])
...
...
@@ -203,7 +203,7 @@ class TestRepeatOp(utt.InferShapeTester):
self
.
op
=
RepeatOp
()
# uint64 always fails
# int64 and uint32 also fail if python int are 32-bit
int_bitwidth
=
theano
.
gof
.
cmodule
.
python_int_bitwidth
()
int_bitwidth
=
theano
.
gof
.
python_int_bitwidth
()
if
int_bitwidth
==
64
:
self
.
numpy_unsupported_dtypes
=
(
'uint64'
,)
if
int_bitwidth
==
32
:
...
...
@@ -287,11 +287,12 @@ class TestRepeatOp(utt.InferShapeTester):
x
=
T
.
TensorType
(
config
.
floatX
,
[
False
,
True
,
False
])()
r
=
RepeatOp
(
axis
=
1
)(
x
,
2
)
self
.
assertEqual
(
r
.
broadcastable
,
(
False
,
False
,
False
))
r
=
RepeatOp
(
axis
=
1
)(
x
,
1
)
r
=
RepeatOp
(
axis
=
1
)(
x
,
1
)
self
.
assertEqual
(
r
.
broadcastable
,
(
False
,
True
,
False
))
r
=
RepeatOp
(
axis
=
0
)(
x
,
2
)
r
=
RepeatOp
(
axis
=
0
)(
x
,
2
)
self
.
assertEqual
(
r
.
broadcastable
,
(
False
,
True
,
False
))
class
TestBartlett
(
utt
.
InferShapeTester
):
def
setUp
(
self
):
...
...
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