Skip to content

Fix Flashinfer Allreduce+Norm enable disable calculation based on fi_allreduce_fusion_max_token_num #21325

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

Merged
merged 3 commits into from
Jul 22, 2025
Merged

Fix Flashinfer Allreduce+Norm enable disable calculation based on fi_allreduce_fusion_max_token_num #21325

Changes from 2 commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
3fb686f
fix flashifner enable disable calculation
xinli-sw Jul 21, 2025
bf3fa63
address copilot feedback
xinli-sw Jul 21, 2025
8195e6c
address review feedback when world size is uncommon
xinli-sw Jul 21, 2025
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
14 changes: 8 additions & 6 deletions vllm/compilation/collective_fusion.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -173,12 +173,14 @@ def call_trtllm_fused_allreduce_norm(
max_token_num: int,
norm_out: Optional[torch.Tensor] = None,
) -> None:
use_flashinfer = allreduce_in.shape[0] * allreduce_in.shape[
1] * allreduce_in.element_size() <= min(
_FI_MAX_SIZES[world_size],
max_token_num * allreduce_in.shape[0] *
allreduce_in.element_size(),
)

num_tokens, hidden_size = allreduce_in.shape
element_size = allreduce_in.element_size()
current_tensor_size = num_tokens * hidden_size * element_size
max_fusion_size = max_token_num * hidden_size * element_size
use_flashinfer = current_tensor_size <= min(_FI_MAX_SIZES[world_size],
max_fusion_size)

if use_flashinfer:
assert (_FI_WORKSPACE_TENSOR is not None
), "Flashinfer must be enabled when using flashinfer"
Expand Down