[V1] V1 Enablement Oracle (#13726)

Signed-off-by: rshaw@neuralmagic.com <rshaw@neuralmagic.com>
Co-authored-by: rshaw@neuralmagic.com <rshaw@neuralmagic.com>
Co-authored-by: Nicolò Lucchesi <nlucches@redhat.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Michael Goin <michael@neuralmagic.com>
This commit is contained in:
Robert Shaw
2025-03-14 22:02:20 -07:00
committed by GitHub
co-authored by rshaw@neuralmagic.com <rshaw@neuralmagic.com> Nicolò Lucchesi Tyler Michael Smith Michael Goin
parent 8c0d15d5c5
commit d4d93db2c5
96 changed files with 1537 additions and 512 deletions
@@ -34,7 +34,10 @@ def test_disable_sliding_window(model_len_len, ):
del vllm_disabled_model
cleanup_dist_env_and_memory()
vllm_enabled_model = LLM(model, disable_sliding_window=False)
vllm_enabled_model = LLM(model,
enforce_eager=True,
disable_sliding_window=False,
enable_prefix_caching=False)
vllm_enabled_model.generate("Hi my name is")
model_config = vllm_enabled_model.llm_engine.model_config
assert model_config.max_model_len == full_len, (
@@ -16,6 +16,15 @@ from vllm.platforms import current_platform
from ..models.utils import check_outputs_equal
@pytest.fixture(scope="function", autouse=True)
def use_v0_only(monkeypatch):
"""
This module relies on V0 internals, so set VLLM_USE_V1=0.
"""
monkeypatch.setenv('VLLM_USE_V1', '0')
MODELS = [
"distilbert/distilgpt2",
]