fix(pd): disable overlap for spec+grammar in disagg decode loop (#28039)
This commit is contained in:
@@ -10,6 +10,7 @@ import requests
|
||||
from transformers import AutoTokenizer
|
||||
|
||||
from sglang.test.ci.ci_register import register_cuda_ci
|
||||
from sglang.test.kits.json_constrained_kit import JSONConstrainedMixin
|
||||
from sglang.test.kits.pause_generation_kit import PauseResumeInPlaceMixin
|
||||
from sglang.test.run_eval import run_eval
|
||||
from sglang.test.server_fixtures.disaggregation_fixture import (
|
||||
@@ -21,7 +22,7 @@ from sglang.test.test_utils import (
|
||||
DEFAULT_TARGET_MODEL_EAGLE3,
|
||||
)
|
||||
|
||||
register_cuda_ci(est_time=509, stage="base-b", runner_config="2-gpu-large")
|
||||
register_cuda_ci(est_time=560, stage="base-b", runner_config="2-gpu-large")
|
||||
|
||||
|
||||
class TestDisaggregationAccuracy(PauseResumeInPlaceMixin, PDDisaggregationServerBase):
|
||||
@@ -215,7 +216,7 @@ class TestDisaggregationMooncakeFailure(PDDisaggregationServerBase):
|
||||
raise e from health_check_error
|
||||
|
||||
|
||||
class TestDisaggregationMooncakeSpec(PDDisaggregationServerBase):
|
||||
class TestDisaggregationMooncakeSpec(JSONConstrainedMixin, PDDisaggregationServerBase):
|
||||
@classmethod
|
||||
def setUpClass(cls):
|
||||
super().setUpClass()
|
||||
|
||||
Reference in New Issue
Block a user