[Refactor] Simplify dummy data generation (#35025)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

vllm/model_executor/models/llava.py

@@ -232,14 +232,13 @@ class LlavaDummyInputsBuilder(BaseDummyInputsBuilder[_I]):
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_images = mm_counts.get("image", 0)
 
         target_width, target_height = self.info.get_image_size_with_most_features()
 
-        image_overrides = mm_options.get("image") if mm_options else None
+        image_overrides = mm_options.get("image")
 
         return {
             "image": self._get_dummy_images(

vllm/model_executor/models/llava_next_video.py

@@ -165,8 +165,7 @@ class LlavaNextVideoDummyInputsBuilder(
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_videos = mm_counts.get("video", 0)
 
@@ -175,7 +174,7 @@ class LlavaNextVideoDummyInputsBuilder(
             seq_len, mm_counts
         )
 
-        video_overrides = mm_options.get("video") if mm_options else None
+        video_overrides = mm_options.get("video")
 
         return {
             "video": self._get_dummy_videos(

vllm/model_executor/models/llava_onevision.py

@@ -276,8 +276,7 @@ class LlavaOnevisionDummyInputsBuilder(
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_images = mm_counts.get("image", 0)
         num_videos = mm_counts.get("video", 0)
@@ -287,8 +286,8 @@ class LlavaOnevisionDummyInputsBuilder(
             seq_len, mm_counts
         )
 
-        image_overrides = mm_options.get("image") if mm_options else None
-        video_overrides = mm_options.get("video") if mm_options else None
+        image_overrides = mm_options.get("image")
+        video_overrides = mm_options.get("video")
 
         return {
             "image": self._get_dummy_images(

vllm/model_executor/models/midashenglm.py

@@ -565,12 +565,11 @@ class MiDashengLMDummyInputsBuilder(BaseDummyInputsBuilder[MiDashengLMProcessing
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_audios = mm_counts.get("audio", 0)
 
-        audio_overrides = mm_options.get("audio") if mm_options else None
+        audio_overrides = mm_options.get("audio")
 
         return {
             "audio": self._get_dummy_audios(

vllm/model_executor/models/minicpmo.py

@@ -301,8 +301,7 @@ class MiniCPMODummyInputsBuilder(MiniCPMVDummyInputsBuilder[MiniCPMOProcessingIn
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_audios = mm_counts.get("audio", 0)
         audio_len = (
@@ -310,11 +309,13 @@ class MiniCPMODummyInputsBuilder(MiniCPMVDummyInputsBuilder[MiniCPMOProcessingIn
             * self.info.get_default_audio_sampling_rate()
         )
 
-        audio_overrides = mm_options.get("audio") if mm_options else None
+        audio_overrides = mm_options.get("audio")
 
         audio_mm_data = {
             "audio": self._get_dummy_audios(
-                length=audio_len, num_audios=num_audios, overrides=audio_overrides
+                length=audio_len,
+                num_audios=num_audios,
+                overrides=audio_overrides,
             )
         }
 

vllm/model_executor/models/minicpmv.py

@@ -707,8 +707,7 @@ class MiniCPMVDummyInputsBuilder(BaseDummyInputsBuilder[_I]):
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_images = mm_counts.get("image", 0)
         num_videos = mm_counts.get("video", 0)
@@ -719,8 +718,8 @@ class MiniCPMVDummyInputsBuilder(BaseDummyInputsBuilder[_I]):
             seq_len, mm_counts
         )
 
-        image_overrides = mm_options.get("image") if mm_options else None
-        video_overrides = mm_options.get("video") if mm_options else None
+        image_overrides = mm_options.get("image")
+        video_overrides = mm_options.get("video")
 
         return {
             "image": self._get_dummy_images(

vllm/model_executor/models/mistral3.py

@@ -236,14 +236,13 @@ class Mistral3DummyInputsBuilder(BaseDummyInputsBuilder[_I]):
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_images = mm_counts.get("image", 0)
 
         target_width, target_height = self.info.get_image_size_with_most_features()
 
-        image_overrides = mm_options.get("image") if mm_options else None
+        image_overrides = mm_options.get("image")
 
         return {
             "image": self._get_dummy_images(

vllm/model_executor/models/mllama4.py

@@ -707,14 +707,13 @@ class Mllama4DummyInputsBuilder(BaseDummyInputsBuilder[Mllama4ProcessingInfo]):
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_images = mm_counts.get("image", 0)
 
         (target_width, target_height) = self.info.get_image_size_with_most_features()
 
-        image_overrides = mm_options.get("image") if mm_options else None
+        image_overrides = mm_options.get("image")
 
         return {
             "image": self._get_dummy_images(

vllm/model_executor/models/molmo.py

@@ -1274,13 +1274,12 @@ class MolmoDummyInputsBuilder(BaseDummyInputsBuilder[MolmoProcessingInfo]):
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         target_width, target_height = self.info.get_image_size_with_most_features()
         num_images = mm_counts.get("image", 0)
 
-        image_overrides = mm_options.get("image") if mm_options else None
+        image_overrides = mm_options.get("image")
 
         return {
             "image": self._get_dummy_images(

vllm/model_executor/models/molmo2.py

@@ -2082,8 +2082,7 @@ class Molmo2DummyInputsBuilder(BaseDummyInputsBuilder[Molmo2ProcessingInfo]):
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_images = mm_counts.get("image", 0)
         num_videos = mm_counts.get("video", 0)
@@ -2094,7 +2093,7 @@ class Molmo2DummyInputsBuilder(BaseDummyInputsBuilder[Molmo2ProcessingInfo]):
         if num_images > 0:
             target_width, target_height = self.info.get_image_size_with_most_features()
 
-            image_overrides = mm_options.get("image") if mm_options else None
+            image_overrides = mm_options.get("image")
 
             dummy_images = self._get_dummy_images(
                 width=target_width,
@@ -2110,7 +2109,7 @@ class Molmo2DummyInputsBuilder(BaseDummyInputsBuilder[Molmo2ProcessingInfo]):
                 seq_len, mm_counts
             )
 
-            video_overrides = mm_options.get("video") if mm_options else None
+            video_overrides = mm_options.get("video")
 
             if video_overrides:
                 assert isinstance(video_overrides, VideoDummyOptions)

vllm/model_executor/models/nano_nemotron_vl.py

@@ -1388,8 +1388,7 @@ class NanoNemotronVLDummyInputsBuilder(BaseDummyInputsBuilder[_I]):
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_images = mm_counts.get("image", 0)
         processor = self.info.get_hf_processor()
@@ -1404,7 +1403,7 @@ class NanoNemotronVLDummyInputsBuilder(BaseDummyInputsBuilder[_I]):
                 max_num_tiles
             )
 
-        image_overrides = mm_options.get("image") if mm_options else None
+        image_overrides = mm_options.get("image")
 
         return {
             "image": self._get_dummy_images(
@@ -1461,12 +1460,9 @@ class NanoNemotronVLDummyInputsBuilder(
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
-        dummy_image = super().get_dummy_mm_data(
-            seq_len=seq_len, mm_counts=mm_counts, mm_options=mm_options
-        )
+        dummy_image = super().get_dummy_mm_data(seq_len, mm_counts, mm_options)
         if self.info.supports_video:
             config = self.info.get_hf_config()
             image_size: int = config.force_image_size
@@ -1474,7 +1470,7 @@ class NanoNemotronVLDummyInputsBuilder(
                 seq_len, mm_counts
             )
             num_videos = mm_counts.get("video", 0)
-            video_overrides = mm_options.get("video") if mm_options else None
+            video_overrides = mm_options.get("video")
             dummy_video = {
                 "video": self._get_dummy_videos(
                     width=image_size,

vllm/model_executor/models/nemotron_parse.py

@@ -645,8 +645,7 @@ class NemotronParseDummyInputsBuilder(
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_images = mm_counts.get("image", 0)
 

vllm/model_executor/models/nvlm_d.py

@@ -92,13 +92,12 @@ class NVLMDummyInputsBuilder(BaseInternVLDummyInputsBuilder[NVLMProcessingInfo])
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         target_width, target_height = self.info.get_image_size_with_most_features()
         num_images = mm_counts.get("image", 0)
 
-        image_overrides = mm_options.get("image") if mm_options else None
+        image_overrides = mm_options.get("image")
 
         return {
             "image": self._get_dummy_images(

vllm/model_executor/models/ovis.py

@@ -306,14 +306,13 @@ class OvisDummyInputsBuilder(BaseDummyInputsBuilder[OvisProcessingInfo]):
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_images = mm_counts.get("image", 0)
 
         target_width, target_height = self.info.get_image_size_with_most_features()
 
-        image_overrides = mm_options.get("image") if mm_options else None
+        image_overrides = mm_options.get("image")
 
         mm_data = {
             "image": self._get_dummy_images(

vllm/model_executor/models/ovis2_5.py

@@ -287,8 +287,7 @@ class Ovis2_5DummyInputsBuilder(BaseDummyInputsBuilder[Ovis2_5ProcessingInfo]):
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_images = mm_counts.get("image", 0)
         num_videos = mm_counts.get("video", 0)
@@ -298,8 +297,8 @@ class Ovis2_5DummyInputsBuilder(BaseDummyInputsBuilder[Ovis2_5ProcessingInfo]):
             seq_len, mm_counts
         )
 
-        image_overrides = mm_options.get("image") if mm_options else None
-        video_overrides = mm_options.get("video") if mm_options else None
+        image_overrides = mm_options.get("image")
+        video_overrides = mm_options.get("video")
 
         mm_data = {
             "image": self._get_dummy_images(

vllm/model_executor/models/paddleocr_vl.py

@@ -206,13 +206,12 @@ class PaddleOCRVLDummyInputsBuilder(BaseDummyInputsBuilder[PaddleOCRVLProcessing
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_images = mm_counts.get("image", 0)
 
         max_image_size = self.info.get_image_size_with_most_features()
-        image_overrides = mm_options.get("image") if mm_options else None
+        image_overrides = mm_options.get("image")
 
         return {
             "image": self._get_dummy_images(

vllm/model_executor/models/paligemma.py

@@ -131,8 +131,7 @@ class PaliGemmaDummyInputsBuilder(BaseDummyInputsBuilder[PaliGemmaProcessingInfo
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         hf_config = self.info.get_hf_config()
         vision_config = hf_config.vision_config
@@ -140,7 +139,7 @@ class PaliGemmaDummyInputsBuilder(BaseDummyInputsBuilder[PaliGemmaProcessingInfo
 
         num_images = mm_counts.get("image", 0)
 
-        image_overrides = mm_options.get("image") if mm_options else None
+        image_overrides = mm_options.get("image")
 
         return {
             "image": self._get_dummy_images(

vllm/model_executor/models/phi3v.py

@@ -376,14 +376,13 @@ class Phi3VDummyInputsBuilder(BaseDummyInputsBuilder[Phi3VProcessingInfo]):
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_images = mm_counts.get("image", 0)
 
         target_width, target_height = self.info.get_image_size_with_most_features()
 
-        image_overrides = mm_options.get("image") if mm_options else None
+        image_overrides = mm_options.get("image")
 
         return {
             "image": self._get_dummy_images(

vllm/model_executor/models/phi4mm.py

@@ -822,16 +822,15 @@ class Phi4MMDummyInputsBuilder(BaseDummyInputsBuilder[Phi4MMProcessingInfo]):
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_audios = mm_counts.get("audio", 0)
         num_images = mm_counts.get("image", 0)
 
         target_width, target_height = self.info.get_image_size_with_most_features()
 
-        image_overrides = mm_options.get("image") if mm_options else None
-        audio_overrides = mm_options.get("audio") if mm_options else None
+        image_overrides = mm_options.get("image")
+        audio_overrides = mm_options.get("audio")
 
         mm_data = {
             "image": self._get_dummy_images(

vllm/model_executor/models/pixtral.py

@@ -249,14 +249,13 @@ class PixtralDummyInputsBuilder(BaseDummyInputsBuilder[PixtralProcessingInfo]):
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> MultiModalDataDict:
         num_images = mm_counts.get("image", 0)
 
         target_width, target_height = self.info.get_image_size_with_most_features()
 
-        image_overrides = mm_options.get("image") if mm_options else None
+        image_overrides = mm_options.get("image")
 
         return {
             "image": self._get_dummy_images(
@@ -271,8 +270,7 @@ class PixtralDummyInputsBuilder(BaseDummyInputsBuilder[PixtralProcessingInfo]):
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-        mm_options: Mapping[str, BaseDummyOptions] | None = None,
-        mm_processor_kwargs: Mapping[str, object] | None = None,
+        mm_options: Mapping[str, BaseDummyOptions],
     ) -> ProcessorInputs:
         tokenizer = self.info.get_tokenizer()