Skip to content

Bugfix: fix symm not enabled due to incorrect registration of comm #19329

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wangfakang wants to merge 10 commits into
base: main
Choose a base branch
from
+24 −15
Open

Bugfix: fix symm not enabled due to incorrect registration of comm #19329

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
6bff9ba
Bugfix: fix symm not enabled due to incorrect registration of comm.
wangfakang Feb 25, 2026
11e5bbc
Merge branch 'main' into bugfix_symm
wangfakang Feb 25, 2026
87fc74d
fix style.
wangfakang Feb 25, 2026
7396483
Merge branch 'main' into bugfix_symm
wangfakang Feb 26, 2026
5750fd4
Merge branch 'main' into bugfix_symm
wangfakang Feb 27, 2026
ded5f45
Merge branch 'main' into bugfix_symm
wangfakang Mar 9, 2026
13b345a
Merge branch 'main' into bugfix_symm
wangfakang Mar 10, 2026
e482f98
Merge branch 'main' into bugfix_symm
wangfakang Mar 11, 2026
f648969
Merge branch 'main' into bugfix_symm
wangfakang Mar 13, 2026
f46710d
Merge branch 'main' into bugfix_symm
wangfakang Mar 13, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 10 additions & 5 deletions python/sglang/srt/layers/communicator.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,6 +43,7 @@
get_attention_cp_rank,
get_attention_cp_size,
get_attention_dp_size,
get_attention_tp_group,
get_attention_tp_rank,
get_attention_tp_size,
get_global_dp_buffer,
Expand Down Expand Up @@ -740,7 +741,7 @@ def _scattered_to_tp_attn_full(
return tuple(gathered_hidden_states)

hidden_states, local_hidden_states = (
get_local_dp_buffer(),
get_local_dp_buffer(get_attention_tp_group()),
hidden_states,
)
attn_tp_all_gather_into_tensor(
Expand Down Expand Up @@ -837,7 +838,7 @@ def _gather_hidden_states_and_residual(

if residual_input_mode == ScatterMode.SCATTERED and context.attn_tp_size > 1:
residual, local_residual = (
get_local_dp_buffer(),
get_local_dp_buffer(get_attention_tp_group()),
residual,
)
attn_tp_all_gather_into_tensor(residual, local_residual)
Expand All @@ -854,7 +855,7 @@ def _gather_hidden_states_and_residual(
hidden_states += residual

hidden_states, local_hidden_states = (
get_global_dp_buffer(),
get_global_dp_buffer(get_tp_group()),
hidden_states,
)
dp_gather_partial(hidden_states, local_hidden_states, forward_batch)
Expand Down Expand Up @@ -993,8 +994,12 @@ def _scatter_hidden_states(
context: CommunicateContext,
allow_reduce_scatter: bool = False,
):
if get_tensor_model_parallel_world_size() == get_attention_dp_size():
group = get_tp_group()
else:
group = get_attention_tp_group()
hidden_states, global_hidden_states = (
get_local_dp_buffer(),
get_local_dp_buffer(group),
hidden_states,
)
if allow_reduce_scatter and forward_batch.dp_padding_mode.is_max_len():
Expand All @@ -1015,7 +1020,7 @@ def _gather(
hidden_states += residual
residual = None
hidden_states, local_hidden_states = (
get_local_dp_buffer(),
get_local_dp_buffer(get_attention_tp_group()),
hidden_states,
)
attn_tp_all_gather_into_tensor(
Expand Down
3 changes: 2 additions & 1 deletion python/sglang/srt/layers/communicator_nsa_cp.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,6 +34,7 @@
from sglang.srt.layers.dp_attention import (
attn_cp_all_gather_into_tensor,
attn_cp_reduce_scatter_tensor,
get_attention_cp_group,
get_local_dp_buffer,
)
from sglang.srt.model_executor.forward_batch_info import ForwardBatch
Expand Down Expand Up @@ -154,7 +155,7 @@ def _gather_hidden_states_and_residual(
if nsa_use_prefill_cp(forward_batch):
assert context.attn_dp_size == 1
hidden_states, local_hidden_states = (
get_local_dp_buffer(),
get_local_dp_buffer(get_attention_cp_group()),
hidden_states,
)
attn_cp_all_gather_into_tensor(
Expand Down
16 changes: 8 additions & 8 deletions python/sglang/srt/layers/dp_attention.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -116,8 +116,8 @@ def set_dp_buffer_len(
cls._global_num_tokens = global_num_tokens

@classmethod
def get_global_dp_buffer(cls) -> torch.Tensor:
with use_symmetric_memory(get_tp_group(), disabled=not cls._dp_max_padding):
def get_global_dp_buffer(cls, group: GroupCoordinator) -> torch.Tensor:
with use_symmetric_memory(group, disabled=not cls._dp_max_padding):
buffer = torch.empty(
(cls._global_dp_buffer_len, cls._hidden_size),
dtype=cls._dtype,
Expand All @@ -126,8 +126,8 @@ def get_global_dp_buffer(cls) -> torch.Tensor:
return buffer

@classmethod
def get_local_dp_buffer(cls) -> torch.Tensor:
with use_symmetric_memory(get_tp_group(), disabled=not cls._dp_max_padding):
def get_local_dp_buffer(cls, group: GroupCoordinator) -> torch.Tensor:
with use_symmetric_memory(group, disabled=not cls._dp_max_padding):
Comment on lines -129 to +130
Copy link
Copy Markdown
Collaborator

@ShangmingCai ShangmingCai Feb 26, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why not make group: Optional[GroupCoordinator] = None, if group==None, then we still use get_tp_group() by default.

Copy link
Copy Markdown
Contributor Author

@wangfakang wangfakang Feb 26, 2026 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why not make group: Optional[GroupCoordinator] = None, if group==None, then we still use get_tp_group() by default.

Using default values can easily lead to group inconsistency, so it's necessary to explicitly declare the correct group to ensure consistency with the communication operator's context.

buffer = torch.empty(
(cls._local_dp_buffer_len, cls._hidden_size),
dtype=cls._dtype,
Expand Down Expand Up @@ -183,12 +183,12 @@ def set_dp_buffer_len(
)


def get_global_dp_buffer() -> torch.Tensor:
return _DpGatheredBufferWrapper.get_global_dp_buffer()
def get_global_dp_buffer(group: GroupCoordinator) -> torch.Tensor:
return _DpGatheredBufferWrapper.get_global_dp_buffer(group=group)


def get_local_dp_buffer() -> torch.Tensor:
return _DpGatheredBufferWrapper.get_local_dp_buffer()
def get_local_dp_buffer(group: GroupCoordinator) -> torch.Tensor:
return _DpGatheredBufferWrapper.get_local_dp_buffer(group=group)


def get_global_dp_buffer_len() -> int:
Expand Down
5 changes: 4 additions & 1 deletion python/sglang/srt/layers/moe/token_dispatcher/standard.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -188,7 +188,10 @@ def dispatch(
def combine(self, combine_input: StandardCombineInput) -> torch.Tensor:
(hidden_states,) = combine_input
if should_use_flashinfer_cutlass_moe_fp4_allgather():
hidden_states, global_hidden_states = get_local_dp_buffer(), hidden_states
hidden_states, global_hidden_states = (
get_local_dp_buffer(get_tp_group()),
hidden_states,
)
get_tp_group().reduce_scatterv(
global_hidden_states,
output=hidden_states,
Expand Down
Loading