[Chore] Remove SupportsV0Only interface and update supported models docs (#26783)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
@@ -76,9 +76,6 @@ class _HfExamplesInfo:
|
||||
trust_remote_code: bool = False
|
||||
"""The ``trust_remote_code`` level required to load the model."""
|
||||
|
||||
v0_only: bool = False
|
||||
"""The model is only available with the vLLM V0 engine."""
|
||||
|
||||
hf_overrides: dict[str, Any] = field(default_factory=dict)
|
||||
"""The ``hf_overrides`` required to load the model."""
|
||||
|
||||
@@ -694,7 +691,6 @@ _MULTIMODAL_EXAMPLE_MODELS = {
|
||||
"MiniMaxVL01ForConditionalGeneration": _HfExamplesInfo(
|
||||
"MiniMaxAI/MiniMax-VL-01",
|
||||
trust_remote_code=True,
|
||||
v0_only=True,
|
||||
),
|
||||
"Mistral3ForConditionalGeneration": _HfExamplesInfo(
|
||||
"mistralai/Mistral-Small-3.1-24B-Instruct-2503",
|
||||
|
||||
@@ -88,13 +88,15 @@ def can_initialize(
|
||||
# gpu_blocks (> 0), cpu_blocks, scheduler_kv_cache_config
|
||||
return 1, 0, scheduler_kv_cache_config
|
||||
|
||||
if model_arch == "MiniMaxVL01ForConditionalGeneration":
|
||||
pytest.skip(
|
||||
"pickle error when loading `transformers.models.auto.CONFIG_MAPPING`"
|
||||
)
|
||||
|
||||
with (
|
||||
patch.object(V1EngineCore, "_initialize_kv_caches", _initialize_kv_caches_v1),
|
||||
monkeypatch.context() as m,
|
||||
):
|
||||
if model_info.v0_only:
|
||||
# NOTE(woosuk): skip the test for V0-only models
|
||||
return
|
||||
if model_arch == "GptOssForCausalLM":
|
||||
# FIXME: A hack to bypass FA3 assertion because our CI's L4 GPU
|
||||
# has cc==8.9 which hasn't supported FA3 yet. Remove this hack when
|
||||
@@ -132,8 +134,6 @@ def can_initialize(
|
||||
@pytest.mark.parametrize("model_arch", MINIMAL_MODEL_ARCH_LIST)
|
||||
def test_can_initialize_small_subset(model_arch: str, monkeypatch: pytest.MonkeyPatch):
|
||||
"""Test initializing small subset of supported models"""
|
||||
if model_arch == "Lfm2ForCausalLM":
|
||||
pytest.skip("Skipping until test supports V1-only models")
|
||||
can_initialize(model_arch, monkeypatch, HF_EXAMPLE_MODELS)
|
||||
|
||||
|
||||
@@ -144,8 +144,6 @@ def test_can_initialize_large_subset(model_arch: str, monkeypatch: pytest.Monkey
|
||||
This test covers the complement of the tests covered in the "small subset"
|
||||
test.
|
||||
"""
|
||||
if model_arch == "Lfm2ForCausalLM":
|
||||
pytest.skip("Skipping until test supports V1-only models")
|
||||
can_initialize(model_arch, monkeypatch, HF_EXAMPLE_MODELS)
|
||||
|
||||
|
||||
|
||||
Reference in New Issue
Block a user