[Model] support new model ovis2.5 (#23084)

Signed-off-by: myselvess <244285088@qq.com>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: Isotr0py <2037008807@qq.com>
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>
This commit is contained in:
myselvess
2025-08-19 21:12:59 +08:00
committed by GitHub
parent f856c33ce9
commit b87cb97a53
12 changed files with 1787 additions and 1 deletions

View File

@@ -11,6 +11,7 @@ from pathlib import PosixPath
import pytest
from transformers import (AutoModel, AutoModelForImageTextToText,
AutoModelForTextToWaveform, AutoModelForVision2Seq)
from transformers.utils import is_flash_attn_2_available
from vllm.platforms import current_platform
from vllm.utils import identity
@@ -621,6 +622,26 @@ VLM_TEST_SETTINGS = {
hf_model_kwargs={"llm_attn_implementation": "sdpa"},
patch_hf_runner=model_utils.ovis_patch_hf_runner,
),
"ovis2_5": VLMTestInfo(
models=["AIDC-AI/Ovis2.5-2B"],
test_type=(
VLMTestType.IMAGE,
VLMTestType.MULTI_IMAGE,
VLMTestType.VIDEO
),
prompt_formatter=lambda img_prompt: f"<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n<|im_start|>user\n{img_prompt}<|im_end|>\n<|im_start|>assistant\n", # noqa: E501
img_idx_to_prompt=lambda idx: "<image>\n", # noqa: E501
video_idx_to_prompt=lambda idx: "<video>\n",
max_model_len=4096,
max_num_seqs=2,
dtype="half",
num_logprobs=10,
patch_hf_runner=model_utils.ovis2_5_patch_hf_runner,
marks=[pytest.mark.skipif(
not is_flash_attn_2_available(),
reason="HF model needs `flash_attn` installed"
)],
),
"phi3v": VLMTestInfo(
models=["microsoft/Phi-3.5-vision-instruct"],
test_type=(VLMTestType.IMAGE, VLMTestType.MULTI_IMAGE),