[Bug fix] need record start time in pd mode (#5425)
This commit is contained in:
@@ -959,14 +959,12 @@ class Scheduler(
|
|||||||
self._add_request_to_queue(req)
|
self._add_request_to_queue(req)
|
||||||
|
|
||||||
def _add_request_to_queue(self, req: Req):
|
def _add_request_to_queue(self, req: Req):
|
||||||
|
req.queue_time_start = time.time()
|
||||||
if self.disaggregation_mode == DisaggregationMode.PREFILL:
|
if self.disaggregation_mode == DisaggregationMode.PREFILL:
|
||||||
self.disagg_prefill_pending_queue.add(req)
|
self.disagg_prefill_pending_queue.add(req)
|
||||||
|
|
||||||
elif self.disaggregation_mode == DisaggregationMode.DECODE:
|
elif self.disaggregation_mode == DisaggregationMode.DECODE:
|
||||||
self.disagg_decode_prealloc_queue.add(req)
|
self.disagg_decode_prealloc_queue.add(req)
|
||||||
|
|
||||||
else:
|
else:
|
||||||
req.queue_time_start = time.time()
|
|
||||||
self.waiting_queue.append(req)
|
self.waiting_queue.append(req)
|
||||||
|
|
||||||
def _extend_requests_to_queue(self, reqs: List[Req], is_retracted: bool = False):
|
def _extend_requests_to_queue(self, reqs: List[Req], is_retracted: bool = False):
|
||||||
|
|||||||
Reference in New Issue
Block a user