Skip to content

[Draft] merge_allreduce #7039

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
fxyfxy777 wants to merge 2 commits into
base: develop
Choose a base branch
from
+14 −3
Open

[Draft] merge_allreduce #7039

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3ad41c9
1 reduce
fxyfxy777 Mar 16, 2026
292401d
Merge branch 'develop' into merge_allreduce
zhoutianzi666 Mar 30, 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
17 changes: 14 additions & 3 deletions fastdeploy/model_executor/models/glm4_moe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,6 +26,7 @@
from paddleformers.utils.log import logger

from fastdeploy.config import FDConfig
from fastdeploy.distributed.communication import tensor_model_parallel_all_reduce
from fastdeploy.model_executor.forward_meta import ForwardMeta
from fastdeploy.model_executor.graph_optimization.decorator import (
support_graph_optimization,
Expand Down Expand Up @@ -160,8 +161,16 @@ def __init__(
default_initializer=paddle.nn.initializer.Constant(0),
)

# In pure-TP mode (tp>1, ep=1) both branches return partial sums, so we
# defer the all-reduce to after combining them — saving one collective.
# In all other modes (EP, EP+attn-TP, no parallelism) each branch handles
# its own reduction internally (reduce_results default=True), so we must
# NOT add an extra all-reduce here.
self._pure_tp = self.use_tp and not self.use_ep
Copy link
Copy Markdown
Collaborator

@zhoutianzi666 zhoutianzi666 Mar 30, 2026

Choose a reason for hiding this comment

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

这个变量应该叫 self.merge_ffn_tp 是不是更好点

Copy link
Copy Markdown
Contributor Author

@fxyfxy777 fxyfxy777 Mar 30, 2026

Choose a reason for hiding this comment

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

这个变量应该叫 self.merge_ffn_tp 是不是更好点

是的，更容易理解一些


self.experts = FusedMoE(
fd_config,
reduce_results=not self._pure_tp,
renormalize=self.norm_topk_prob,
moe_intermediate_size=fd_config.model_config.moe_intermediate_size,
num_experts=fd_config.model_config.n_routed_experts,
Expand All @@ -182,14 +191,16 @@ def __init__(
intermediate_size=shared_experts_intermediate_size,
layer_id=layer_id,
prefix=f"{prefix}.shared_experts",
reduce_results=not self._pure_tp,
)

def forward(self, x, forward_meta: ForwardMeta = None):
out = self.experts(x, self.gate, forward_meta)
if self.n_shared_experts > 0:
shared_experts_out = self.shared_experts(x)
out = out + shared_experts_out

out = out + self.shared_experts(x)
if self._pure_tp:
# Both branches produced partial sums; combine first, then single all-reduce.
out = tensor_model_parallel_all_reduce(out, self.tp_group)
return out


Expand Down
Loading