[Bugfix][Mooncake] Fix heterogeneous PP transfer completion (#56033)

Signed-off-by: wangyicong <wangyicong@bytedance.com>
Co-authored-by: OpenAI Codex <codex@openai.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
Yicong Wang
2026-09-11 10:57:19 +08:00
committed by GitHub
co-authored by OpenAI Codex mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent 07b7553465
commit ce08bb5b34
2 changed files with 85 additions and 3 deletions
@@ -307,6 +307,8 @@ def _validate_asymmetric_region_lengths(
def _align_transfer_regions(
local_regions: list[TransferRegion],
remote_regions: list[TransferRegion],
*,
allow_partial_layers: bool = False,
) -> tuple[list[TransferRegion], list[TransferRegion], str | None]:
"""Align KV transfer regions by registered layer-name occurrence.
@@ -335,6 +337,11 @@ def _align_transfer_regions(
for key, local_region in local_keyed:
remote_region = remote_by_key.get(key)
if remote_region is None:
if (
allow_partial_layers
and (local_region.layer_name, 0) not in remote_by_key
):
continue
return (
[],
[],
@@ -393,6 +400,7 @@ class MooncakeXferMetadata(
registered_layer_names: list[str] = msgspec.field(default_factory=list)
registered_layer_indices: list[int] = msgspec.field(default_factory=list)
registered_group_indices: list[int] = msgspec.field(default_factory=list)
remote_pp_size: int = 1
class MooncakeXferResponseStatus(IntEnum):
@@ -1232,7 +1240,11 @@ class MooncakeConnectorWorker:
meta.registered_group_indices,
)
local_regions, remote_regions, align_err = _align_transfer_regions(
local_regions, remote_regions
local_regions,
remote_regions,
allow_partial_layers=(
meta.remote_pp_size > 1 and meta.remote_pp_size != self.pp_size
),
)
if align_err is not None:
response = MooncakeXferResponse(
@@ -1322,7 +1334,9 @@ class MooncakeConnectorWorker:
# Mark it sending to avoid expiration.
send_meta.sending += 1
if not send_meta.need_send:
self.resolve_need_send(send_meta, remote_tp_ranks)
self.resolve_need_send(
send_meta, remote_tp_ranks, meta.remote_pp_size
)
ready_reqs.append((d_req_id, send_meta))
else:
# Otherwise (expired, very unlikely), just forget it.
@@ -1398,11 +1412,16 @@ class MooncakeConnectorWorker:
self,
send_meta: SendBlockMeta,
remote_tp_ranks: list[int],
remote_pp_size: int = 1,
):
# Prepare for heterogeneous TP (one P pairs to multiple D)
send_meta.need_send = len(remote_tp_ranks)
if remote_pp_size > 1 and remote_pp_size != self.pp_size:
# Each consumer PP stage pulls every producer stage, including
# peers with no shared layers.
send_meta.need_send *= remote_pp_size
logger.debug(
"Mooncake request %s will be served by %d consumer TP workers: TP ranks=%s",
"Mooncake request %s will be served by %d consumer workers: TP ranks=%s",
send_meta.transfer_id,
send_meta.need_send,
remote_tp_ranks,
@@ -1851,6 +1870,7 @@ class MooncakeConnectorWorker:
remote_port=self.rpc_port,
remote_tp_size=self.tp_size,
remote_tp_rank=self.tp_rank,
remote_pp_size=self.pp_size,
req_blocks={
req_id: (pull_meta.transfer_id, pull_meta.local_block_ids)
for req_id, pull_meta in pull_metas.items()