[Bugfix] Missing NIXL metadata for handshake initialization if instance spans multi-node (#26338)

Signed-off-by: Guan Luo <gluo@nvidia.com>
Signed-off-by: GuanLuo <41310872+GuanLuo@users.noreply.github.com>
Signed-off-by: Guan Luo <41310872+GuanLuo@users.noreply.github.com>
Co-authored-by: Nicolò Lucchesi <nlucches@redhat.com>
This commit is contained in:
GuanLuo
2025-10-31 10:16:00 -07:00
committed by GitHub
co-authored by Nicolò Lucchesi
parent 7e06c40e63
commit d6517be3cd
7 changed files with 321 additions and 95 deletions
+8
View File
@@ -9,6 +9,9 @@ from typing import TYPE_CHECKING, Literal, TypeVar, overload
from vllm.config import VllmConfig
from vllm.distributed.kv_transfer.kv_connector.utils import KVOutputAggregator
from vllm.distributed.kv_transfer.kv_connector.v1.base import (
KVConnectorHandshakeMetadata,
)
from vllm.logger import init_logger
from vllm.lora.request import LoRARequest
from vllm.tasks import SupportedTask
@@ -177,6 +180,11 @@ class Executor(ABC):
):
raise NotImplementedError
def get_kv_connector_handshake_metadata(
self,
) -> list[dict[int, KVConnectorHandshakeMetadata]]:
return self.collective_rpc("get_kv_connector_handshake_metadata")
@overload
def execute_model(
self,