Sync from v0.13

This commit is contained in:
2026-01-19 10:38:50 +08:00
parent b2ef04d792
commit 5aef6c175a
3714 changed files with 854317 additions and 89342 deletions

View File

@@ -0,0 +1,33 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import pytest
from ..conftest import IMAGE_ASSETS
HF_IMAGE_PROMPTS = IMAGE_ASSETS.prompts(
{
"stop_sign": "USER: <image>\nWhat's the content of the image?\nASSISTANT:",
"cherry_blossom": "USER: <image>\nWhat is the season?\nASSISTANT:",
}
)
models = ["llava-hf/llava-1.5-7b-hf"]
@pytest.mark.parametrize("model", models)
def test_context_length_too_short(vllm_runner, image_assets, model):
images = [asset.pil_image for asset in image_assets]
with pytest.raises(ValueError, match="longer than the maximum model length"):
vllm_model = vllm_runner(
model,
max_model_len=128, # LLaVA has a feature size of 576
enforce_eager=True,
load_format="dummy",
)
with vllm_model:
vllm_model.generate_greedy(
[HF_IMAGE_PROMPTS[0]], max_tokens=1, images=[images[0]]
)