biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Docs] Improve V1 KVConnector interface documentation (#19172)

Browse Source 
Signed-off-by: Nick Hill <nhill@redhat.com>
This commit is contained in:
Nick Hill
2025-06-06 01:22:45 -07:00
committed by GitHub
parent 94870359cd
commit 65c69444b1
2 changed files with 32 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
vllm/v1/core/sched/scheduler.py
View File

@@ -101,7 +101,7 @@ class Scheduler(SchedulerInterface):
# This is flushed at the end of each scheduling step.
self.finished_req_ids: set[str] = set()
# P/D: requests in process of recving KV transfers
# KV Connector: requests in process of async KV loading or recving
self.finished_recving_kv_req_ids: set[str] = set()
# OPTIMIZATION: Cache the CachedRequestData objects to avoid creating
@@ -822,7 +822,7 @@ class Scheduler(SchedulerInterface):
if not stopped:
new_running.append(request)
# P/D: update state for finished KV Transfers.
# KV Connector: update state for finished KV Transfers.
self._update_from_kv_xfer_finished(model_runner_output)
# Return the cached request data to the queue so they can be reused.
@@ -969,7 +969,7 @@ class Scheduler(SchedulerInterface):
self.kv_event_publisher.shutdown()
########################################################################
# P/D Related Methods
# KV Connector Related Methods
########################################################################
def get_kv_connector(self) -> Optional[KVConnectorBase_V1]:
@@ -992,7 +992,7 @@ class Scheduler(SchedulerInterface):
def _update_waiting_for_remote_kv(self, request: Request) -> bool:
"""
P/D: check if the request_id is finished_recving.
KV Connector: check if the request_id is finished_recving.
The finished_recving_kv_req_ids list is populated
on the previous steps()'s update_from_output based
@@ -1029,7 +1029,7 @@ class Scheduler(SchedulerInterface):
def _update_from_kv_xfer_finished(self,
model_runner_output: ModelRunnerOutput):
"""
P/D: update the scheduler state based on the output.
KV Connector: update the scheduler state based on the output.
The Worker side connectors add finished_recving and
finished_sending reqs to the output.
@@ -1037,7 +1037,7 @@ class Scheduler(SchedulerInterface):
# if finished_recving: add to state so we can
scheduler the request during the next step.
"""
# P/D: update recv and send status from last step.
# KV Connector:: update recv and send status from last step.
for req_id in (model_runner_output.finished_recving or ()):
logger.debug("Finished recving KV transfer for request %s", req_id)
self.finished_recving_kv_req_ids.add(req_id)