[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
+5 -7
View File
@@ -12,7 +12,7 @@ from transformers import (
)
from vllm.config import ModelConfig, SpeechToTextConfig, VllmConfig
from vllm.config.multimodal import AudioDummyOptions, BaseDummyOptions
from vllm.config.multimodal import MultiModalDummyOptions
from vllm.config.speech_to_text import SpeechToTextParams
from vllm.inputs import MultiModalDataDict, PromptType
from vllm.logger import init_logger
@@ -207,20 +207,18 @@ class FireRedASR2DummyInputsBuilder(BaseDummyInputsBuilder[FireRedASR2Processing
self,
seq_len: int,
mm_counts: Mapping[str, int],
mm_options: Mapping[str, BaseDummyOptions],
mm_options: MultiModalDummyOptions,
) -> MultiModalDataDict:
feature_extractor = self.info.get_feature_extractor()
sampling_rate = feature_extractor.sampling_rate
audio_len = feature_extractor.chunk_length * sampling_rate
num_audios = mm_counts.get("audio", 0)
audio_overrides = mm_options.get("audio")
assert audio_overrides is None or isinstance(audio_overrides, AudioDummyOptions)
ret = {
"audio": self._get_dummy_audios(
length=audio_len, num_audios=num_audios, overrides=audio_overrides
length=audio_len,
num_audios=mm_counts.get("audio", 0),
overrides=mm_options.get("audio"),
)
}
return ret