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

[2/3] Refactor InternVL-based processors (#37324)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-03-18 22:22:19 +08:00
committed by GitHub
parent 525f2eeb0b
commit 99267c23ca
18 changed files with 815 additions and 1199 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
tests/models/multimodal/generation/vlm_utils/model_utils.py
View File

@@ -489,13 +489,14 @@ def h2ovl_patch_hf_runner(hf_model: HfRunner) -> HfRunner:
self.image_size = self.vision_config.image_size
def __call__(self, text: str, images: Image | list[Image], **kwargs):
from vllm.model_executor.models.h2ovl import (
IMG_CONTEXT,
IMG_END,
IMG_START,
from vllm.transformers_utils.processors.h2ovl import (
image_to_pixel_values_h2ovl,
)
IMG_START = "<img>"
IMG_END = "</img>"
IMG_CONTEXT = "<IMG_CONTEXT>"
images = [images] if isinstance(images, Image) else images
pixel_values = [
image_to_pixel_values_h2ovl(
@@ -751,16 +752,17 @@ def skyworkr1v_patch_hf_runner(hf_model: HfRunner) -> HfRunner:
self.image_size = self.vision_config.image_size
def __call__(self, text: str, images: Image | list[Image], **kwargs):
from vllm.model_executor.models.skyworkr1v import (
IMG_CONTEXT,
IMG_END,
IMG_START,
image_to_pixel_values_skyworkr1v,
from vllm.transformers_utils.processors.internvl import (
image_to_pixel_values_internvl,
)
IMG_START = "<img>"
IMG_END = "</img>"
IMG_CONTEXT = "<IMG_CONTEXT>"
images = [images] if isinstance(images, Image) else images
pixel_values = [
image_to_pixel_values_skyworkr1v(
image_to_pixel_values_internvl(
image,
input_size=self.image_size,
min_num=self.min_num,
@@ -815,14 +817,15 @@ def internvl_patch_hf_runner(hf_model: HfRunner) -> HfRunner:
videos: npt.NDArray | list[npt.NDArray] = None,
**kwargs,
):
from vllm.model_executor.models.internvl import (
IMG_CONTEXT,
IMG_END,
IMG_START,
from vllm.transformers_utils.processors.internvl import (
image_to_pixel_values_internvl,
video_to_pixel_values_internvl,
)
IMG_START = "<img>"
IMG_END = "</img>"
IMG_CONTEXT = "<IMG_CONTEXT>"
images = [images] if isinstance(images, Image) else images
videos = [videos] if isinstance(videos, np.ndarray) else videos
if images is not None: