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

[Chore] Rename SchedulerConfig.chunked_prefill_enabled (#28735)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-11-15 02:39:57 +08:00
committed by GitHub
parent 67187554dd
commit e2741f6cbc
9 changed files with 21 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
tests/v1/e2e/test_spec_decode.py
View File

@@ -272,7 +272,7 @@ def test_speculators_model_integration(
@pytest.mark.parametrize(
["model_setup", "mm_enabled", "chunked_prefill_enabled"],
["model_setup", "mm_enabled", "enable_chunked_prefill"],
[
(("eagle3", "Qwen/Qwen3-8B", "AngelSlim/Qwen3-8B_eagle3", 1), False, False),
pytest.param(
@@ -358,7 +358,7 @@ def test_eagle_correctness(
sampling_config: SamplingParams,
model_setup: tuple[str, str, str, int],
mm_enabled: bool,
chunked_prefill_enabled: bool,
enable_chunked_prefill: bool,
attn_backend: str,
):
if attn_backend == "TREE_ATTN":
@@ -396,9 +396,7 @@ def test_eagle_correctness(
method, model_name, spec_model_name, tp_size = model_setup
max_model_len = 2048
max_num_batched_tokens = max_model_len
if chunked_prefill_enabled:
max_num_batched_tokens = 128
max_num_batched_tokens = 128 if enable_chunked_prefill else max_model_len
ref_llm = LLM(
model=model_name, max_model_len=max_model_len, tensor_parallel_size=tp_size
@@ -420,7 +418,7 @@ def test_eagle_correctness(
},
max_model_len=max_model_len,
max_num_batched_tokens=max_num_batched_tokens,
enable_chunked_prefill=chunked_prefill_enabled,
enable_chunked_prefill=enable_chunked_prefill,
)
spec_outputs = spec_llm.chat(test_prompts, sampling_config)
matches = 0