[Multi Modal] Configurable MM Profiling (#25631)
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:
co-authored by
Harry Mellor
parent
2ed3f20dba
commit
79aa244678
@@ -13,6 +13,7 @@ from vllm.attention.backends.registry import _Backend
|
||||
from vllm.attention.layer import (check_upstream_fa_availability,
|
||||
maybe_get_vit_flash_attn_backend)
|
||||
from vllm.config import VllmConfig
|
||||
from vllm.config.multimodal import BaseDummyOptions
|
||||
from vllm.distributed import utils as dist_utils
|
||||
from vllm.distributed.parallel_state import (
|
||||
get_tensor_model_parallel_rank, get_tensor_model_parallel_world_size)
|
||||
@@ -91,17 +92,21 @@ class DotsOCRDummyInputsBuilder(Qwen2VLDummyInputsBuilder):
|
||||
self,
|
||||
seq_len: int,
|
||||
mm_counts: Mapping[str, int],
|
||||
mm_options: Optional[Mapping[str, BaseDummyOptions]] = None,
|
||||
) -> MultiModalDataDict:
|
||||
num_images = mm_counts.get("image", 0)
|
||||
|
||||
target_width, target_height = self.info.get_image_size_with_most_features( # noqa: E501
|
||||
)
|
||||
|
||||
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),
|
||||
}
|
||||
|
||||
|
||||
|
||||
Reference in New Issue
Block a user