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

[VLM] Initialize video input support for InternVL models (#18499)

Browse Source 
Signed-off-by: Isotr0py <2037008807@qq.com>
Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
This commit is contained in:
Isotr0py
2025-05-25 12:51:25 +08:00
committed by GitHub
parent 6ab681bcbe
commit 75f81750f3
10 changed files with 596 additions and 62 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
tests/models/multimodal/generation/test_common.py
View File

@@ -349,6 +349,17 @@ VLM_TEST_SETTINGS = {
use_tokenizer_eos=True,
patch_hf_runner=model_utils.internvl_patch_hf_runner,
),
"intern_vl-video": VLMTestInfo(
models=[
"OpenGVLab/InternVL3-1B",
],
test_type=VLMTestType.VIDEO,
prompt_formatter=lambda img_prompt: f"<|im_start|>User\n{img_prompt}<|im_end|>\n<|im_start|>Assistant\n", # noqa: E501
video_idx_to_prompt=lambda idx: "<video>",
max_model_len=8192,
use_tokenizer_eos=True,
patch_hf_runner=model_utils.internvl_patch_hf_runner,
),
"kimi_vl": VLMTestInfo(
models=["moonshotai/Kimi-VL-A3B-Instruct"],
test_type=(VLMTestType.IMAGE, VLMTestType.MULTI_IMAGE),