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/qwen_vl.py
View File

@@ -24,6 +24,7 @@ from transformers.image_utils import ImageInput
from transformers.tokenization_utils_base import TextInput
from vllm.config import VllmConfig
from vllm.config.multimodal import BaseDummyOptions
from vllm.model_executor.layers.activation import get_act_fn
from vllm.model_executor.layers.linear import (ColumnParallelLinear,
ReplicatedLinear,
@@ -567,6 +568,7 @@ class QwenVLDummyInputsBuilder(BaseDummyInputsBuilder[QwenVLProcessingInfo]):
self,
seq_len: int,
mm_counts: Mapping[str, int],
mm_options: Optional[Mapping[str, BaseDummyOptions]] = None,
) -> MultiModalDataDict:
hf_config = self.info.get_hf_config()
vision_config = hf_config.visual
@@ -574,11 +576,14 @@ class QwenVLDummyInputsBuilder(BaseDummyInputsBuilder[QwenVLProcessingInfo]):
target_width = target_height = vision_config["image_size"]
num_images = mm_counts.get("image", 0)
image_overrides = mm_options.get("image") if mm_options else None
return {
"image":
self._get_dummy_images(width=target_width,
height=target_height,
num_images=num_images)
num_images=num_images,
overrides=image_overrides)
}