Skip to content

Commit

Permalink
[Misc] Fix typos in scheduler.py (vllm-project#7285)
Browse files Browse the repository at this point in the history
Signed-off-by: Rui Qiao <[email protected]>
  • Loading branch information
ruisearch42 authored Aug 8, 2024
1 parent e53dfd3 commit 7467096
Showing 1 changed file with 3 additions and 3 deletions.
6 changes: 3 additions & 3 deletions vllm/core/scheduler.py
Original file line number Diff line number Diff line change
Expand Up @@ -678,7 +678,7 @@ def _schedule_prefills(
all tokens.
Returns:
SchedulerSwappedInOutputs.
SchedulerPrefillOutputs.
"""
ignored_seq_groups: List[SequenceGroup] = []
seq_groups: List[SequenceGroup] = []
Expand Down Expand Up @@ -851,7 +851,7 @@ def _schedule_default(self) -> SchedulerOutputs:
preempted=preempted,
)

def _schedule_chunked_prefill(self):
def _schedule_chunked_prefill(self) -> SchedulerOutputs:
"""Schedule queued requests.
Chunked prefill allows to chunk prefill requests, batch them together
Expand All @@ -862,7 +862,7 @@ def _schedule_chunked_prefill(self):
The policy can sustain the high GPU utilization because it can put
prefill and decodes requests to the same batch, while it improves
inter token latency because decodes requests don't need to blocked
inter token latency because decodes requests don't need to be blocked
by prefill requests.
"""
budget = SchedulingBudget(
Expand Down

0 comments on commit 7467096

Please sign in to comment.