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

10
vllm/model_executor/models/minicpmo.py
View File

@@ -36,6 +36,7 @@ from transformers.models.whisper.modeling_whisper import (ACT2FN,
WhisperEncoder)
from vllm.config import VllmConfig
from vllm.config.multimodal import BaseDummyOptions
from vllm.multimodal import MULTIMODAL_REGISTRY, MultiModalKwargsItems
from vllm.multimodal.inputs import (MultiModalDataDict, MultiModalFieldConfig,
NestedTensors)
@@ -237,18 +238,23 @@ class MiniCPMODummyInputsBuilder(
self,
seq_len: int,
mm_counts: Mapping[str, int],
mm_options: Optional[Mapping[str, BaseDummyOptions]] = None,
) -> MultiModalDataDict:
num_audios = mm_counts.get("audio", 0)
audio_len = self.info.get_max_audio_chunks_with_most_features() * \
self.info.get_default_audio_sampling_rate()
audio_overrides = mm_options.get("audio") if mm_options else None
audio_mm_data = {
"audio":
self._get_dummy_audios(length=audio_len, num_audios=num_audios)
self._get_dummy_audios(length=audio_len,
num_audios=num_audios,
overrides=audio_overrides)
}
return {
**super().get_dummy_mm_data(seq_len, mm_counts),
**super().get_dummy_mm_data(seq_len, mm_counts, mm_options),
**audio_mm_data,
}