mirror of
https://git.datalinker.icu/vllm-project/vllm.git
synced 2025-12-10 01:24:54 +08:00
[V1][Minor] Minor enhancements on scheduler (#14732)
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
This commit is contained in:
parent
f53a0586b9
commit
01b3fd0af7
@ -587,9 +587,6 @@ class Scheduler:
|
|||||||
if spec_token_ids is not None:
|
if spec_token_ids is not None:
|
||||||
request.spec_token_ids = spec_token_ids[req_index]
|
request.spec_token_ids = spec_token_ids[req_index]
|
||||||
|
|
||||||
# Get prompt logprobs for this request.
|
|
||||||
prompt_logprobs_tensors = prompt_logprobs_dict.get(req_id)
|
|
||||||
|
|
||||||
stopped = False
|
stopped = False
|
||||||
new_logprobs = None
|
new_logprobs = None
|
||||||
new_token_ids: list[int] = []
|
new_token_ids: list[int] = []
|
||||||
@ -622,6 +619,8 @@ class Scheduler:
|
|||||||
new_token_ids,
|
new_token_ids,
|
||||||
)
|
)
|
||||||
|
|
||||||
|
# Get prompt logprobs for this request.
|
||||||
|
prompt_logprobs_tensors = prompt_logprobs_dict.get(req_id)
|
||||||
# Transmit partial if chunked prefill & prompt logprobs is enabled
|
# Transmit partial if chunked prefill & prompt logprobs is enabled
|
||||||
if new_token_ids or prompt_logprobs_tensors is not None:
|
if new_token_ids or prompt_logprobs_tensors is not None:
|
||||||
# Add EngineCoreOutput for this Request.
|
# Add EngineCoreOutput for this Request.
|
||||||
@ -693,8 +692,7 @@ class Scheduler:
|
|||||||
|
|
||||||
if request.status == RequestStatus.RUNNING:
|
if request.status == RequestStatus.RUNNING:
|
||||||
self.running.remove(request)
|
self.running.remove(request)
|
||||||
if request.request_id in self.scheduled_req_ids:
|
self.scheduled_req_ids.discard(request.request_id)
|
||||||
self.scheduled_req_ids.remove(request.request_id)
|
|
||||||
else:
|
else:
|
||||||
self.waiting.remove(request)
|
self.waiting.remove(request)
|
||||||
request.status = finished_status
|
request.status = finished_status
|
||||||
|
|||||||
Loading…
x
Reference in New Issue
Block a user