Skip to content

[Feature] Impl v1 disaggregated prefill in ascend scheduler #852

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
May 23, 2025
Merged

[Feature] Impl v1 disaggregated prefill in ascend scheduler #852

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
13 changes: 13 additions & 0 deletions vllm_ascend/core/scheduler.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,6 +51,11 @@ def __init__(
self.scheduled_req_ids: set[str] = set()
self.running: list[Request] = []

if self.vllm_config.kv_transfer_config is not None and \
self.vllm_config.kv_transfer_config.is_kv_consumer:
raise ValueError(
"AscendScheduler cannot be used for decode nodes. ")

def schedule(self) -> SchedulerOutput:
if self.scheduler_config.chunked_prefill_enabled:
return super().schedule()
Expand Down Expand Up @@ -287,6 +292,14 @@ def skip_cur_request():
grammar_bitmask=None,
)

# NOTE(Kuntai): this function is designed for multiple purposes:
# 1. Plan the KV cache store
# 2. Wrap up all the KV cache load / save ops into an opaque object
# 3. Clear the internal states of the connector
if self.connector is not None:
meta = self.connector.build_connector_meta(scheduler_output)
scheduler_output.kv_connector_metadata = meta

# Advance the number of computed tokens for the request AFTER
# the request is scheduled.
# 1. The scheduler_output of the current step has to include the
Expand Down