biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Core] Async scheduling + structured outputs compatibility (#26866)

Browse Source 
Signed-off-by: Nick Hill <nhill@redhat.com>
This commit is contained in:
Nick Hill
2025-10-31 17:35:04 -07:00
committed by GitHub
parent df334868ca
commit 0cdbe7b744
25 changed files with 419 additions and 191 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
tests/v1/worker/test_gpu_model_runner.py
View File

@@ -152,8 +152,6 @@ def _schedule_new_request(*req_ids: str) -> SchedulerOutput:
num_common_prefix_blocks=[],
finished_req_ids=set(),
free_encoder_mm_hashes=[],
structured_output_request_ids=[],
grammar_bitmask=None,
)
@@ -269,8 +267,6 @@ def test_update_states_request_finished(model_runner, dist_init):
num_common_prefix_blocks=[],
finished_req_ids={req_id},
free_encoder_mm_hashes=[],
structured_output_request_ids=[],
grammar_bitmask=None,
)
metadata_before = model_runner.input_batch.sampling_metadata
@@ -301,8 +297,6 @@ def test_update_states_request_resumed(model_runner, dist_init):
num_common_prefix_blocks=[],
finished_req_ids=set(),
free_encoder_mm_hashes=[],
structured_output_request_ids=[],
grammar_bitmask=None,
)
model_runner._update_states(scheduler_output)
@@ -330,8 +324,6 @@ def test_update_states_request_resumed(model_runner, dist_init):
num_common_prefix_blocks=[],
finished_req_ids=set(),
free_encoder_mm_hashes=[],
structured_output_request_ids=[],
grammar_bitmask=None,
)
metadata_before = model_runner.input_batch.sampling_metadata
@@ -423,8 +415,6 @@ def test_update_states_no_changes(model_runner, dist_init):
num_common_prefix_blocks=[],
finished_req_ids=set(),
free_encoder_mm_hashes=[],
structured_output_request_ids=[],
grammar_bitmask=None,
)
metadata_before = model_runner.input_batch.sampling_metadata
@@ -460,8 +450,6 @@ def test_update_states_request_unscheduled(model_runner, dist_init):
num_common_prefix_blocks=[],
finished_req_ids=set(),
free_encoder_mm_hashes=[],
structured_output_request_ids=[],
grammar_bitmask=None,
)
metadata_before = model_runner._update_states(scheduler_output)