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

[V1][Spec Decode] Ngram Spec Decode (#12193)

Browse Source 
Signed-off-by: LiuXiaoxuanPKU <lilyliupku@gmail.com>
This commit is contained in:
Lily Liu
2025-02-15 18:05:11 -08:00
committed by GitHub
parent 367cb8ce8c
commit 80f63a3966
21 changed files with 1023 additions and 82 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tests/v1/worker/test_gpu_model_runner.py
View File

@@ -66,6 +66,7 @@ def _schedule_new_request(*req_ids: str) -> SchedulerOutput:
scheduled_cached_reqs=[],
num_scheduled_tokens=num_scheduled_tokens,
total_num_scheduled_tokens=total_num_scheduled_tokens,
scheduled_spec_decode_tokens={},
scheduled_encoder_inputs={},
num_common_prefix_blocks=0,
finished_req_ids=set(),
@@ -109,6 +110,7 @@ def test_update_states_request_finished(model_runner):
scheduled_cached_reqs=[],
num_scheduled_tokens={},
total_num_scheduled_tokens=0,
scheduled_spec_decode_tokens={},
scheduled_encoder_inputs={},
num_common_prefix_blocks=0,
finished_req_ids={req_id},
@@ -137,6 +139,7 @@ def test_update_states_request_resumed(model_runner):
scheduled_cached_reqs=[],
num_scheduled_tokens={},
total_num_scheduled_tokens=0,
scheduled_spec_decode_tokens={},
scheduled_encoder_inputs={},
num_common_prefix_blocks=0,
finished_req_ids={},
@@ -160,6 +163,7 @@ def test_update_states_request_resumed(model_runner):
scheduled_cached_reqs=[cached_req_data],
num_scheduled_tokens={req_id: 1},
total_num_scheduled_tokens=1,
scheduled_spec_decode_tokens={},
scheduled_encoder_inputs={},
num_common_prefix_blocks=0,
finished_req_ids=set(),
@@ -188,6 +192,7 @@ def test_update_states_no_changes(model_runner):
scheduled_cached_reqs=[],
num_scheduled_tokens={req_id: 1},
total_num_scheduled_tokens=1,
scheduled_spec_decode_tokens={},
scheduled_encoder_inputs={},
num_common_prefix_blocks=0,
finished_req_ids=set(),
@@ -220,6 +225,7 @@ def test_update_states_request_unscheduled(model_runner):
scheduled_cached_reqs=[],
num_scheduled_tokens={req_ids[0]: 1},
total_num_scheduled_tokens=1,
scheduled_spec_decode_tokens={},
scheduled_encoder_inputs={},
num_common_prefix_blocks=0,
finished_req_ids=set(),