[Multimodal] Type dummy options per modality (#57576)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Harry Mellor
2026-09-18 17:29:11 +01:00
committed by GitHub
co-authored by Claude Opus 5
parent 56cac80b14
commit 8a5cf54387
93 changed files with 461 additions and 720 deletions
+3 -5
View File
@@ -11,7 +11,7 @@ import torch.nn as nn
from transformers import BatchFeature, CLIPVisionConfig
from vllm.config import VllmConfig
from vllm.config.multimodal import BaseDummyOptions
from vllm.config.multimodal import MultiModalDummyOptions
from vllm.inputs import MultiModalDataDict
from vllm.model_executor.models.interfaces import (
MultiModalEmbeddings,
@@ -268,17 +268,15 @@ class DeepseekOCRDummyInputsBuilder(BaseDummyInputsBuilder[DeepseekOCRProcessing
self,
seq_len: int,
mm_counts: Mapping[str, int],
mm_options: Mapping[str, BaseDummyOptions],
mm_options: MultiModalDummyOptions,
) -> MultiModalDataDict:
num_images = mm_counts.get("image", 0)
max_image_size = self.info.get_image_size_with_most_features()
return {
"image": self._get_dummy_images(
width=max_image_size.width,
height=max_image_size.height,
num_images=num_images,
num_images=mm_counts.get("image", 0),
)
}