[Bugfix] Fix prefill suspension caused by delayed negotiate_should_allow_prefill invocation (#32389)
This commit is contained in:
@@ -1000,7 +1000,11 @@ class PrefillAdder:
|
|||||||
|
|
||||||
if (self.prefill_delayer_single_pass is not None) and (
|
if (self.prefill_delayer_single_pass is not None) and (
|
||||||
not self.prefill_delayer_single_pass.negotiate_should_allow_prefill(
|
not self.prefill_delayer_single_pass.negotiate_should_allow_prefill(
|
||||||
local_prefillable=True
|
local_prefillable=True,
|
||||||
|
running_batch=self.running_batch.batch_size(),
|
||||||
|
max_prefill_bs=self.max_prefill_bs,
|
||||||
|
max_running_requests=self.max_running_requests,
|
||||||
|
waiting_queue_len=self.waiting_queue_len,
|
||||||
)
|
)
|
||||||
):
|
):
|
||||||
return AddReqResult.OTHER
|
return AddReqResult.OTHER
|
||||||
|
|||||||
Reference in New Issue
Block a user