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

[Misc] Remove pad_for_cudagraphs from config (#30143)

Browse Source 
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Co-authored-by: Matthew Bonanni <mbonanni@redhat.com>
This commit is contained in:
Lucas Wilkinson
2026-01-20 13:05:48 -07:00
committed by GitHub
parent 86c69dc54c
commit 2261340806
9 changed files with 157 additions and 136 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
tests/models/language/generation/test_hybrid.py
View File

@@ -9,6 +9,7 @@ from tests.models.registry import HF_EXAMPLE_MODELS
from tests.utils import multi_gpu_test
from vllm.engine.arg_utils import EngineArgs
from vllm.sampling_params import SamplingParams
from vllm.v1.cudagraph_dispatcher import CudagraphDispatcher
from ...utils import check_logprobs_close, check_outputs_equal
@@ -172,7 +173,14 @@ def test_mamba_cache_cg_padding(
tensor dimensions aren't compatible.
"""
vllm_config = EngineArgs(model=model, trust_remote_code=True).create_engine_config()
while len(example_prompts) == vllm_config.pad_for_cudagraph(len(example_prompts)):
cudagraph_dispatcher = CudagraphDispatcher(vllm_config)
cudagraph_dispatcher.initialize_cudagraph_keys(
vllm_config.compilation_config.cudagraph_mode
)
while (
len(example_prompts)
== cudagraph_dispatcher.dispatch(len(example_prompts))[1].num_tokens
):
example_prompts.append(example_prompts[0])
try: