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

[Multi Modal] Configurable MM Profiling (#25631)

Browse Source 
Signed-off-by: wwl2755 <wangwenlong2755@gmail.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Wenlong Wang
2025-10-03 03:59:10 -07:00
committed by GitHub
parent 2ed3f20dba
commit 79aa244678
60 changed files with 654 additions and 99 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
vllm/model_executor/models/cohere2_vision.py
View File

@@ -16,6 +16,7 @@ from transformers.models.cohere2_vision.processing_cohere2_vision import (
Cohere2VisionProcessor)
from vllm.config import VllmConfig
from vllm.config.multimodal import BaseDummyOptions
from vllm.model_executor.layers.activation import MulAndSilu
from vllm.model_executor.layers.linear import (MergedColumnParallelLinear,
RowParallelLinear)
@@ -209,16 +210,20 @@ class Cohere2VisionDummyInputsBuilder(
self,
seq_len: int,
mm_counts: Mapping[str, int],
mm_options: Optional[Mapping[str, BaseDummyOptions]] = None,
) -> MultiModalDataDict:
num_images = mm_counts.get("image", 0)
image_size = \
self.info.get_image_size_with_most_features()
image_overrides = mm_options.get("image") if mm_options else None
return {
"image":
self._get_dummy_images(width=image_size.width,
height=image_size.height,
num_images=num_images)
num_images=num_images,
overrides=image_overrides)
}