Skip to content

[https://nvbugs/5477730][fix] Fix the alltoall case when tp_size larger than ep_size #7331

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 1 commit into from
Sep 4, 2025
Merged

[https://nvbugs/5477730][fix] Fix the alltoall case when tp_size larger than ep_size #7331

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
10 changes: 6 additions & 4 deletions tensorrt_llm/_mnnvl_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -104,7 +104,9 @@ def get_comm(mapping: Mapping):
if MnnvlMemory.comm is not None:
return MnnvlMemory.comm
comm = mpi_comm().Split(
mapping.pp_rank * mapping.cp_size + mapping.cp_rank, mapping.tp_rank
(mapping.pp_rank * mapping.cp_size + mapping.cp_rank) * mapping.moe_tp_size
+ mapping.moe_tp_rank,
mapping.tp_rank,
)
MnnvlMemory.comm = comm
return comm
Expand Down Expand Up @@ -362,12 +364,12 @@ def get_moe_workspaces(mapping: Mapping):

MnnvlMoe.moe_mapping = mapping
workspace_size_per_rank = torch.ops.trtllm.get_moe_commworkspace_size_per_rank(
mapping.tp_size
mapping.moe_ep_size
)
MnnvlMoe.moe_workspace = MnnvlMemory(mapping, workspace_size_per_rank)
MnnvlMoe.moe_workspace_tensor = MnnvlMoe.moe_workspace.as_torch_strided_tensor(torch.uint64)
torch.ops.trtllm.moe_initialize_workspace(
MnnvlMoe.moe_workspace_tensor, mapping.tp_rank, mapping.tp_size
MnnvlMoe.moe_workspace_tensor, mapping.moe_ep_rank, mapping.moe_ep_size
)
MnnvlMoe.moe_workspace.comm.barrier()
return MnnvlMoe.moe_workspace_tensor
Expand All @@ -378,7 +380,7 @@ def get_moe_prepare_workspace(mapping: Mapping):
assert mapping == MnnvlMoe.moe_mapping, "only one moe mapping supported now"
return MnnvlMoe.moe_prepare_workspace_tensor
workspace_size_per_rank = torch.ops.trtllm.get_moe_prepare_workspace_size_per_rank(
mapping.tp_size
mapping.moe_ep_size
)
MnnvlMoe.moe_prepare_workspace = MnnvlMemory(mapping, workspace_size_per_rank)
MnnvlMoe.moe_prepare_workspace_tensor = (
Expand Down
Loading