Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in
Toggle navigation
Menu
Open sidebar
OpenDAS
DeepEP
Commits
905e9e87
Commit
905e9e87
authored
Jun 02, 2026
by
lijian
Browse files
modify deepep release version to 1.1.0.
Signed-off-by:
lijian
<
34831075+lijian0711@users.noreply.github.com
>
parent
dcf58e2a
Changes
2
Hide whitespace changes
Inline
Side-by-side
Showing
2 changed files
with
11 additions
and
66 deletions
+11
-66
build.sh
build.sh
+8
-63
setup.py
setup.py
+3
-3
No files found.
build.sh
View file @
905e9e87
...
@@ -59,75 +59,20 @@ for arg in "$@"; do
...
@@ -59,75 +59,20 @@ for arg in "$@"; do
done
done
detect_offload_arch
()
{
detect_offload_arch
()
{
# 获取当前硬件的 gfx 版本(例如 gfx936)
if
!
command
-v
rocm_agent_enumerator
>
/dev/null 2>&1
;
then
current_gfx
=
$(
rocminfo 2>/dev/null |
grep
-E
'Name:.*gfx[0-9]+'
|
head
-n1
|
grep
-oE
'gfx[0-9]+'
|
cut
-c4-
)
if
[
-z
"
$current_gfx
"
]
;
then
# 如果无法获取当前硬件版本,回退到原逻辑(选择最大的架构)
if
command
-v
rocm_agent_enumerator
>
/dev/null 2>&1
;
then
arch
=
$(
rocm_agent_enumerator 2>/dev/null |
grep
-E
'^gfx[0-9]+'
|
sort
-r
|
head
-n1
)
if
[
-n
"
$arch
"
]
;
then
echo
"--offload-arch=
$arch
"
return
0
fi
fi
return
1
return
1
fi
fi
# 转换为整数,以便比较(如 936)
supported_archs
=
$(
rocm_agent_enumerator 2>/dev/null |
grep
-E
'^gfx[0-9]+'
|
sort
-r
)
current_gfx_int
=
$((
current_gfx
))
[
-z
"
$supported_archs
"
]
&&
return
1
# 获取所有支持的 gfx 版本(降序排列)
if
command
-v
rocm_agent_enumerator
>
/dev/null 2>&1
;
then
supported_archs
=
$(
rocm_agent_enumerator 2>/dev/null |
grep
-E
'^gfx[0-9]+'
|
sort
-r
)
if
[
-n
"
$supported_archs
"
]
;
then
# 取前2个最大的架构作为基础
top2
=
""
count
=
0
for
arch
in
$supported_archs
;
do
top2
=
"
$top2
--offload-arch=
$arch
"
count
=
$((
count
+
1
))
[
$count
-ge
2
]
&&
break
done
# 检查当前 GPU 是否已经在前2个中
found
=
0
for
arch
in
$supported_archs
;
do
arch_int
=
${
arch
:3
}
if
[
"
$arch_int
"
-eq
"
$current_gfx_int
"
]
;
then
count2
=
0
for
a
in
$supported_archs
;
do
count2
=
$((
count2
+
1
))
[
$count2
-gt
2
]
&&
break
a_int
=
${
a
:3
}
if
[
"
$a_int
"
-eq
"
$current_gfx_int
"
]
;
then
found
=
1
break
fi
done
break
fi
done
# 如果当前 GPU 不在前2个中,追加它
if
[
"
$found
"
-eq
0
]
;
then
top2
=
"
$top2
--offload-arch=gfx
${
current_gfx_int
}
"
fi
echo
"
$top2
"
top2_gfx
=
$(
echo
"
$supported_archs
"
|
head
-n2
|
sed
's/^/--offload-arch=/'
|
tr
'\n'
' '
)
return
0
current_gfx
=
$(
rocminfo 2>/dev/null |
grep
-Eo
'gfx[0-9]+'
|
head
-n1
)
fi
fi
# 回退逻辑:如果没有匹配的架构,选择最大的架构
if
[
-n
"
$current_gfx
"
]
&&
!
echo
"
$top2_gfx
"
|
grep
-q
"
$current_gfx
"
;
then
if
command
-v
rocm_agent_enumerator
>
/dev/null 2>&1
;
then
top2_gfx
=
"
$top2_gfx
--offload-arch=
$current_gfx
"
arch
=
$(
rocm_agent_enumerator 2>/dev/null |
grep
-E
'^gfx[0-9]+'
|
sort
-r
|
head
-n1
)
if
[
-n
"
$arch
"
]
;
then
echo
"--offload-arch=
$arch
"
return
0
fi
fi
fi
echo
"
$top2_gfx
"
return
1
}
}
DETECTED_ARCH
=
$(
detect_offload_arch
)
DETECTED_ARCH
=
$(
detect_offload_arch
)
echo
"Current
$DETECTED_ARCH
"
echo
"Current
$DETECTED_ARCH
"
...
...
setup.py
View file @
905e9e87
...
@@ -56,9 +56,9 @@ def get_version_add(sha: Optional[str] = None) -> str:
...
@@ -56,9 +56,9 @@ def get_version_add(sha: Optional[str] = None) -> str:
new_version_content
=
f
"""
new_version_content
=
f
"""
try:
try:
__version__ = "1.
0
.0"
__version__ = "1.
1
.0"
__version_tuple__ = (1,
0
, 0)
__version_tuple__ = (1,
1
, 0)
__hcu_version__ = f'1.
0
.0+
{
version
}
'
__hcu_version__ = f'1.
1
.0+
{
version
}
'
from deep_ep.version import __version__, __version_tuple__, __hcu_version__
from deep_ep.version import __version__, __version_tuple__, __hcu_version__
except Exception as e:
except Exception as e:
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment