Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
change
sglang
Commits
4a6e7a66
Unverified
Commit
4a6e7a66
authored
Aug 01, 2025
by
kk
Committed by
GitHub
Jul 31, 2025
Browse files
Fix nan value generated after custom all reduce (#8532)
parent
4b04998d
Changes
1
Hide whitespace changes
Inline
Side-by-side
Showing
1 changed file
with
7 additions
and
7 deletions
+7
-7
python/sglang/srt/distributed/device_communicators/custom_all_reduce.py
...srt/distributed/device_communicators/custom_all_reduce.py
+7
-7
No files found.
python/sglang/srt/distributed/device_communicators/custom_all_reduce.py
View file @
4a6e7a66
...
@@ -184,7 +184,7 @@ class CustomAllreduce:
...
@@ -184,7 +184,7 @@ class CustomAllreduce:
# 8*world_size bytes where world_size is at most 8. Allocating 8MB
# 8*world_size bytes where world_size is at most 8. Allocating 8MB
# is enough for 131072 such tuples. The largest model I've seen only
# is enough for 131072 such tuples. The largest model I've seen only
# needs less than 10000 of registered tuples.
# needs less than 10000 of registered tuples.
self
.
rank_data
=
torch
.
empty
(
self
.
rank_data
=
torch
.
zeros
(
8
*
1024
*
1024
,
dtype
=
torch
.
uint8
,
device
=
self
.
device
8
*
1024
*
1024
,
dtype
=
torch
.
uint8
,
device
=
self
.
device
)
)
self
.
_ptr
=
ops
.
init_custom_ar
(
self
.
_ptr
=
ops
.
init_custom_ar
(
...
@@ -194,14 +194,14 @@ class CustomAllreduce:
...
@@ -194,14 +194,14 @@ class CustomAllreduce:
else
:
else
:
# meta data buffers need to be "uncached" for signal on MI200
# meta data buffers need to be "uncached" for signal on MI200
self
.
meta
=
ops
.
allocate_meta_buffer
(
ops
.
meta_size
()
+
max_size
)
self
.
meta
=
ops
.
allocate_meta_buffer
(
ops
.
meta_size
()
+
max_size
)
self
.
buffer
=
torch
.
empty
(
max_size
,
dtype
=
torch
.
uint8
,
device
=
self
.
device
)
self
.
buffer
=
torch
.
zeros
(
max_size
,
dtype
=
torch
.
uint8
,
device
=
self
.
device
)
handle
=
ops
.
get_meta_buffer_ipc_handle
(
self
.
meta
)
handle
=
ops
.
get_meta_buffer_ipc_handle
(
self
.
meta
)
shard_data
=
(
shard_data
=
(
bytes
(
handle
),
# ipc handle to base ptr
bytes
(
handle
),
# ipc handle to base ptr
0
,
# offset of base ptr
0
,
# offset of base ptr
)
)
handles
,
offsets
=
self
.
_gather_ipc_meta
(
shard_data
)
handles
,
offsets
=
self
.
_gather_ipc_meta
(
shard_data
)
self
.
rank_data
=
torch
.
empty
(
self
.
rank_data
=
torch
.
zeros
(
8
*
1024
*
1024
,
dtype
=
torch
.
uint8
,
device
=
self
.
device
8
*
1024
*
1024
,
dtype
=
torch
.
uint8
,
device
=
self
.
device
)
)
self
.
_ptr
=
ops
.
init_custom_ar
(
self
.
_ptr
=
ops
.
init_custom_ar
(
...
@@ -350,14 +350,14 @@ class CustomAllreduce:
...
@@ -350,14 +350,14 @@ class CustomAllreduce:
# or, in the context of cuda graphs, register_graph_buffers
# or, in the context of cuda graphs, register_graph_buffers
def
all_reduce_reg
(
self
,
inp
:
torch
.
Tensor
,
out
:
torch
.
Tensor
=
None
):
def
all_reduce_reg
(
self
,
inp
:
torch
.
Tensor
,
out
:
torch
.
Tensor
=
None
):
if
out
is
None
:
if
out
is
None
:
out
=
torch
.
empty
_like
(
inp
)
out
=
torch
.
zeros
_like
(
inp
)
ops
.
all_reduce_reg
(
self
.
_ptr
,
inp
,
out
)
ops
.
all_reduce_reg
(
self
.
_ptr
,
inp
,
out
)
return
out
return
out
# all reduce, assuming inp tensor is NOT IPC registered
# all reduce, assuming inp tensor is NOT IPC registered
def
all_reduce_unreg
(
self
,
inp
:
torch
.
Tensor
,
out
:
torch
.
Tensor
=
None
):
def
all_reduce_unreg
(
self
,
inp
:
torch
.
Tensor
,
out
:
torch
.
Tensor
=
None
):
if
out
is
None
:
if
out
is
None
:
out
=
torch
.
empty
_like
(
inp
)
out
=
torch
.
zeros
_like
(
inp
)
ops
.
all_reduce_unreg
(
self
.
_ptr
,
inp
,
self
.
buffer
,
out
)
ops
.
all_reduce_unreg
(
self
.
_ptr
,
inp
,
self
.
buffer
,
out
)
return
out
return
out
...
@@ -375,7 +375,7 @@ class CustomAllreduce:
...
@@ -375,7 +375,7 @@ class CustomAllreduce:
buffer.
buffer.
"""
"""
if
out
is
None
:
if
out
is
None
:
out
=
torch
.
empty
_like
(
inp
)
out
=
torch
.
zeros
_like
(
inp
)
if
registered
:
if
registered
:
ops
.
all_reduce
(
self
.
_ptr
,
inp
,
out
,
0
,
0
)
ops
.
all_reduce
(
self
.
_ptr
,
inp
,
out
,
0
,
0
)
else
:
else
:
...
@@ -398,7 +398,7 @@ class CustomAllreduce:
...
@@ -398,7 +398,7 @@ class CustomAllreduce:
else
:
else
:
# If warm up, mimic the allocation pattern since custom
# If warm up, mimic the allocation pattern since custom
# allreduce is out-of-place.
# allreduce is out-of-place.
return
torch
.
empty
_like
(
input
)
return
torch
.
zeros
_like
(
input
)
else
:
else
:
if
_is_hip
:
if
_is_hip
:
# note: outside of cuda graph context,
# note: outside of cuda graph context,
...
...
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