Skip to content

Commit 62942e3

Browse files
DarkLight1337DarkLight1337
committed
Rename
Signed-off-by: DarkLight1337 <[email protected]>
1 parent 5f940fc commit 62942e3

File tree

18 files changed

+71
-68
lines changed

18 files changed

+71
-68
lines changed

‎tests/multimodal/test_processing.py

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -690,7 +690,7 @@ def _test_processing_cache_correctness(
690690

691691
baseline_processor = factories.build_processor(ctx, cache=None)
692692
cached_processor = factories.build_processor(ctx, cache=cache)
693-
dummy_data_builder = baseline_processor.dummy_data_builder
693+
dummy_inputs = baseline_processor.dummy_inputs
694694

695695
rng = np.random.RandomState(0)
696696

@@ -722,7 +722,7 @@ def _test_processing_cache_correctness(
722722
}
723723

724724
mm_counts = {k: len(vs) for k, vs in mm_data.items()}
725-
prompt = dummy_data_builder.get_dummy_processor_inputs(
725+
prompt = dummy_inputs.get_dummy_processor_inputs(
726726
model_config.max_model_len,
727727
mm_counts,
728728
).prompt_text

‎tests/plugins/vllm_add_dummy_model/vllm_add_dummy_model/my_llava.py

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -2,7 +2,7 @@
22

33
import torch
44

5-
from vllm.model_executor.models.llava import (LlavaDummyDataBuilder,
5+
from vllm.model_executor.models.llava import (LlavaDummyInputsBuilder,
66
LlavaForConditionalGeneration,
77
LlavaMultiModalProcessor,
88
LlavaProcessingInfo)
@@ -12,7 +12,7 @@
1212

1313
@MULTIMODAL_REGISTRY.register_processor(LlavaMultiModalProcessor,
1414
info=LlavaProcessingInfo,
15-
dummy_data=LlavaDummyDataBuilder)
15+
dummy=LlavaDummyInputsBuilder)
1616
class MyLlava(LlavaForConditionalGeneration):
1717

1818
def compute_logits(

‎vllm/model_executor/models/aria.py

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -26,7 +26,7 @@
2626
from vllm.multimodal.parse import MultiModalDataItems
2727
from vllm.multimodal.processing import BaseProcessingInfo, PromptReplacement
2828
from vllm.multimodal.processor import BaseMultiModalProcessor
29-
from vllm.multimodal.profiling import BaseDummyDataBuilder, ProcessorInputs
29+
from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
3030
from vllm.sequence import IntermediateTensors
3131
from vllm.transformers_utils.configs.aria import (AriaMoELMConfig,
3232
AriaVisionConfig)
@@ -464,7 +464,7 @@ def get_num_image_tokens(self) -> int:
464464
return max(hf_config.projector_patch_to_query_dict.values())
465465

466466

467-
class AriaDummyDataBuilder(BaseDummyDataBuilder[AriaProcessingInfo]):
467+
class AriaDummyInputsBuilder(BaseDummyInputsBuilder[AriaProcessingInfo]):
468468

469469
def get_dummy_processor_inputs(
470470
self,
@@ -526,7 +526,7 @@ def _get_prompt_replacements(
526526

527527
@MULTIMODAL_REGISTRY.register_processor(AriaMultiModalProcessor,
528528
info=AriaProcessingInfo,
529-
dummy_data=AriaDummyDataBuilder)
529+
dummy=AriaDummyInputsBuilder)
530530
class AriaForConditionalGeneration(nn.Module, SupportsMultiModal):
531531
"""
532532
Aria model for conditional generation tasks.

‎vllm/model_executor/models/blip2.py

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -20,7 +20,7 @@
2020
from vllm.multimodal.parse import MultiModalDataItems
2121
from vllm.multimodal.processing import BaseProcessingInfo, PromptReplacement
2222
from vllm.multimodal.processor import BaseMultiModalProcessor
23-
from vllm.multimodal.profiling import BaseDummyDataBuilder, ProcessorInputs
23+
from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
2424
from vllm.sequence import IntermediateTensors
2525

2626
from .blip import BlipVisionModel
@@ -413,7 +413,7 @@ def get_num_image_tokens(self) -> int:
413413
return hf_config.num_query_tokens
414414

415415

416-
class Blip2DummyDataBuilder(BaseDummyDataBuilder[Blip2ProcessingInfo]):
416+
class Blip2DummyInputsBuilder(BaseDummyInputsBuilder[Blip2ProcessingInfo]):
417417

418418
def get_dummy_processor_inputs(
419419
self,
@@ -490,7 +490,7 @@ def apply(
490490

491491
@MULTIMODAL_REGISTRY.register_processor(Blip2MultiModalProcessor,
492492
info=Blip2ProcessingInfo,
493-
dummy_data=Blip2DummyDataBuilder)
493+
dummy=Blip2DummyInputsBuilder)
494494
class Blip2ForConditionalGeneration(nn.Module, SupportsMultiModal, SupportsPP):
495495

496496
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):

‎vllm/model_executor/models/chameleon.py

Lines changed: 4 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -33,7 +33,7 @@
3333
from vllm.multimodal.parse import MultiModalDataItems
3434
from vllm.multimodal.processing import BaseProcessingInfo, PromptReplacement
3535
from vllm.multimodal.processor import BaseMultiModalProcessor
36-
from vllm.multimodal.profiling import BaseDummyDataBuilder, ProcessorInputs
36+
from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
3737
from vllm.sequence import IntermediateTensors
3838
from vllm.utils import print_warning_once
3939

@@ -68,7 +68,8 @@ def get_num_image_tokens(self) -> int:
6868
return processor.image_seq_length
6969

7070

71-
class ChameleonDummyDataBuilder(BaseDummyDataBuilder[ChameleonProcessingInfo]):
71+
class ChameleonDummyInputsBuilder(
72+
BaseDummyInputsBuilder[ChameleonProcessingInfo]):
7273

7374
def get_dummy_processor_inputs(
7475
self,
@@ -915,7 +916,7 @@ def forward(
915916

916917
@MULTIMODAL_REGISTRY.register_processor(ChameleonMultiModalProcessor,
917918
info=ChameleonProcessingInfo,
918-
dummy_data=ChameleonDummyDataBuilder)
919+
dummy=ChameleonDummyInputsBuilder)
919920
class ChameleonForConditionalGeneration(nn.Module, SupportsMultiModal,
920921
SupportsPP):
921922

‎vllm/model_executor/models/fuyu.py

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -37,7 +37,7 @@
3737
MultiModalDataItems)
3838
from vllm.multimodal.processing import BaseProcessingInfo, PromptReplacement
3939
from vllm.multimodal.processor import BaseMultiModalProcessor
40-
from vllm.multimodal.profiling import BaseDummyDataBuilder, ProcessorInputs
40+
from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
4141
from vllm.sequence import IntermediateTensors
4242

4343
from .interfaces import SupportsMultiModal, SupportsPP
@@ -117,7 +117,7 @@ def get_image_size_with_most_features(self) -> ImageSize:
117117
height=image_processor.size["height"])
118118

119119

120-
class FuyuDummyDataBuilder(BaseDummyDataBuilder[FuyuProcessingInfo]):
120+
class FuyuDummyInputsBuilder(BaseDummyInputsBuilder[FuyuProcessingInfo]):
121121

122122
def get_dummy_processor_inputs(
123123
self,
@@ -244,7 +244,7 @@ def apply(
244244

245245
@MULTIMODAL_REGISTRY.register_processor(FuyuMultiModalProcessor,
246246
info=FuyuProcessingInfo,
247-
dummy_data=FuyuDummyDataBuilder)
247+
dummy=FuyuDummyInputsBuilder)
248248
class FuyuForCausalLM(nn.Module, SupportsMultiModal, SupportsPP):
249249

250250
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):

‎vllm/model_executor/models/llava.py

Lines changed: 7 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -29,7 +29,7 @@
2929
from vllm.multimodal.processing import (BaseProcessingInfo, ProcessingCache,
3030
PromptReplacement)
3131
from vllm.multimodal.processor import BaseMultiModalProcessor
32-
from vllm.multimodal.profiling import BaseDummyDataBuilder, ProcessorInputs
32+
from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
3333
from vllm.sequence import IntermediateTensors
3434

3535
from .clip import CLIPVisionModel
@@ -170,7 +170,7 @@ def get_max_image_tokens(self) -> int:
170170
_I = TypeVar("_I", bound=BaseLlavaProcessingInfo)
171171

172172

173-
class LlavaDummyDataBuilder(BaseDummyDataBuilder[_I]):
173+
class LlavaDummyInputsBuilder(BaseDummyInputsBuilder[_I]):
174174

175175
def get_dummy_processor_inputs(
176176
self,
@@ -360,23 +360,23 @@ def _build_llava_or_pixtral_hf_info(
360360

361361
def _build_llava_or_pixtral_hf_processor(
362362
info: _I,
363-
dummy_data_builder: BaseDummyDataBuilder[_I],
363+
dummy_inputs: BaseDummyInputsBuilder[_I],
364364
*,
365365
cache: Optional[ProcessingCache] = None,
366366
enable_sanity_checks: bool = True,
367367
) -> BaseMultiModalProcessor:
368368
if isinstance(info, PixtralHFProcessingInfo):
369369
return PixtralHFMultiModalProcessor(
370370
info,
371-
dummy_data_builder, # type: ignore
371+
dummy_inputs, # type: ignore
372372
cache=cache,
373373
enable_sanity_checks=enable_sanity_checks,
374374
)
375375

376376
if isinstance(info, LlavaProcessingInfo):
377377
return LlavaMultiModalProcessor(
378378
info,
379-
dummy_data_builder, # type: ignore
379+
dummy_inputs, # type: ignore
380380
cache=cache,
381381
enable_sanity_checks=enable_sanity_checks,
382382
)
@@ -461,7 +461,7 @@ def init_vision_tower_for_llava(
461461

462462
@MULTIMODAL_REGISTRY.register_processor(_build_llava_or_pixtral_hf_processor,
463463
info=_build_llava_or_pixtral_hf_info,
464-
dummy_data=LlavaDummyDataBuilder)
464+
dummy=LlavaDummyInputsBuilder)
465465
class LlavaForConditionalGeneration(nn.Module, SupportsMultiModal, SupportsPP):
466466
# BitandBytes specific attributes
467467
bitsandbytes_stacked_params_mapping = {
@@ -793,6 +793,6 @@ def get_replacement_mantis(item_idx: int):
793793
# `--hf_overrides '{"architectures": ["MantisForConditionalGeneration"]}'`
794794
@MULTIMODAL_REGISTRY.register_processor(MantisMultiModalProcessor,
795795
info=LlavaProcessingInfo,
796-
dummy_data=LlavaDummyDataBuilder)
796+
dummy=LlavaDummyInputsBuilder)
797797
class MantisForConditionalGeneration(LlavaForConditionalGeneration):
798798
pass

‎vllm/model_executor/models/llava_next.py

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -21,7 +21,7 @@
2121
from .clip import CLIPVisionModel
2222
from .interfaces import SupportsMultiModal, SupportsPP
2323
from .llava import (BaseLlavaMultiModalProcessor, BaseLlavaProcessingInfo,
24-
LlavaDummyDataBuilder, LlavaLikeConfig,
24+
LlavaDummyInputsBuilder, LlavaLikeConfig,
2525
LlavaMultiModalProjector, init_vision_tower_for_llava)
2626
from .siglip import SiglipVisionModel
2727
from .utils import (AutoWeightsLoader, embed_multimodal, flatten_bn,
@@ -181,7 +181,7 @@ class LlavaNextMultiModalProcessor(
181181

182182
@MULTIMODAL_REGISTRY.register_processor(LlavaNextMultiModalProcessor,
183183
info=LlavaNextProcessingInfo,
184-
dummy_data=LlavaDummyDataBuilder)
184+
dummy=LlavaDummyInputsBuilder)
185185
class LlavaNextForConditionalGeneration(nn.Module, SupportsMultiModal,
186186
SupportsPP):
187187

‎vllm/model_executor/models/llava_next_video.py

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -21,7 +21,7 @@
2121
VideoEmbeddingItems, VideoProcessorItems)
2222
from vllm.multimodal.processing import BaseProcessingInfo, PromptReplacement
2323
from vllm.multimodal.processor import BaseMultiModalProcessor
24-
from vllm.multimodal.profiling import BaseDummyDataBuilder, ProcessorInputs
24+
from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
2525
from vllm.sequence import IntermediateTensors
2626
from vllm.utils import is_list_of
2727

@@ -134,8 +134,8 @@ def get_max_num_frames(self, seq_len: int) -> int:
134134
return max(max_total_frames // max(max_videos, 1), 1)
135135

136136

137-
class LlavaNextVideoDummyDataBuilder(
138-
BaseDummyDataBuilder[LlavaNextVideoProcessingInfo]):
137+
class LlavaNextVideoDummyInputsBuilder(
138+
BaseDummyInputsBuilder[LlavaNextVideoProcessingInfo]):
139139

140140
def get_dummy_processor_inputs(
141141
self,
@@ -269,7 +269,7 @@ def forward(self, image_features: torch.Tensor) -> torch.Tensor:
269269
@MULTIMODAL_REGISTRY.register_processor(
270270
LlavaNextVideoMultiModalProcessor,
271271
info=LlavaNextVideoProcessingInfo,
272-
dummy_data=LlavaNextVideoDummyDataBuilder,
272+
dummy=LlavaNextVideoDummyInputsBuilder,
273273
)
274274
class LlavaNextVideoForConditionalGeneration(nn.Module, SupportsMultiModal,
275275
SupportsPP):

‎vllm/model_executor/models/llava_onevision.py

Lines changed: 6 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -28,7 +28,7 @@
2828

2929
from .clip import CLIPVisionModel
3030
from .interfaces import SupportsMultiModal, SupportsPP
31-
from .llava import LlavaDummyDataBuilder, init_vision_tower_for_llava
31+
from .llava import LlavaDummyInputsBuilder, init_vision_tower_for_llava
3232
from .llava_next import (BaseLlavaNextMultiModalProcessor, LlavaNextLikeConfig,
3333
LlavaNextProcessingInfo)
3434
from .siglip import SiglipVisionModel
@@ -233,8 +233,8 @@ def get_max_video_tokens(self, seq_len: int) -> int:
233233
)
234234

235235

236-
class LlavaOnevisionDummyDataBuilder(
237-
LlavaDummyDataBuilder[LlavaOnevisionProcessingInfo]):
236+
class LlavaOnevisionDummyInputsBuilder(
237+
LlavaDummyInputsBuilder[LlavaOnevisionProcessingInfo]):
238238

239239
def get_dummy_processor_inputs(
240240
self,
@@ -392,10 +392,9 @@ def forward(self, image_features: torch.Tensor) -> torch.Tensor:
392392
return hidden_states
393393

394394

395-
@MULTIMODAL_REGISTRY.register_processor(
396-
LlavaOnevisionMultiModalProcessor,
397-
info=LlavaOnevisionProcessingInfo,
398-
dummy_data=LlavaOnevisionDummyDataBuilder)
395+
@MULTIMODAL_REGISTRY.register_processor(LlavaOnevisionMultiModalProcessor,
396+
info=LlavaOnevisionProcessingInfo,
397+
dummy=LlavaOnevisionDummyInputsBuilder)
399398
class LlavaOnevisionForConditionalGeneration(nn.Module, SupportsMultiModal,
400399
SupportsPP):
401400

‎vllm/model_executor/models/phi3v.py

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -39,7 +39,7 @@
3939
BoundPromptReplacement,
4040
PlaceholderInfo, PromptReplacement)
4141
from vllm.multimodal.processor import BaseMultiModalProcessor
42-
from vllm.multimodal.profiling import BaseDummyDataBuilder, ProcessorInputs
42+
from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
4343
from vllm.sequence import IntermediateTensors
4444
from vllm.utils import is_list_of
4545

@@ -344,7 +344,7 @@ def get_image_size_with_most_features(self) -> ImageSize:
344344
return ImageSize(height=8000, width=50)
345345

346346

347-
class Phi3VDummyDataBuilder(BaseDummyDataBuilder[Phi3VProcessingInfo]):
347+
class Phi3VDummyInputsBuilder(BaseDummyInputsBuilder[Phi3VProcessingInfo]):
348348

349349
def get_dummy_processor_inputs(
350350
self,
@@ -498,7 +498,7 @@ def apply(
498498

499499
@MULTIMODAL_REGISTRY.register_processor(Phi3VMultiModalProcessor,
500500
info=Phi3VProcessingInfo,
501-
dummy_data=Phi3VDummyDataBuilder)
501+
dummy=Phi3VDummyInputsBuilder)
502502
class Phi3VForCausalLM(nn.Module, SupportsMultiModal, SupportsPP):
503503
hf_to_vllm_mapper = WeightsMapper(
504504
orig_to_new_prefix={

‎vllm/model_executor/models/qwen2_audio.py

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -42,7 +42,7 @@
4242
MultiModalDataParser)
4343
from vllm.multimodal.processing import BaseProcessingInfo, PromptReplacement
4444
from vllm.multimodal.processor import BaseMultiModalProcessor
45-
from vllm.multimodal.profiling import BaseDummyDataBuilder, ProcessorInputs
45+
from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
4646
from vllm.sequence import IntermediateTensors
4747

4848
from .interfaces import SupportsMultiModal, SupportsPP
@@ -115,8 +115,8 @@ def get_mm_max_tokens_per_item(self, seq_len: int) -> Mapping[str, int]:
115115
return {"audio": max_output_lengths}
116116

117117

118-
class Qwen2AudioDummyDataBuilder(BaseDummyDataBuilder[Qwen2AudioProcessingInfo]
119-
):
118+
class Qwen2AudioDummyInputsBuilder(
119+
BaseDummyInputsBuilder[Qwen2AudioProcessingInfo]):
120120

121121
def get_dummy_processor_inputs(
122122
self,
@@ -237,7 +237,7 @@ def _always_apply_prompt_replacements(self) -> bool:
237237

238238
@MULTIMODAL_REGISTRY.register_processor(Qwen2AudioMultiModalProcessor,
239239
info=Qwen2AudioProcessingInfo,
240-
dummy_data=Qwen2AudioDummyDataBuilder)
240+
dummy=Qwen2AudioDummyInputsBuilder)
241241
class Qwen2AudioForConditionalGeneration(nn.Module, SupportsMultiModal,
242242
SupportsPP):
243243

‎vllm/model_executor/models/qwen2_vl.py

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -60,7 +60,7 @@
6060
MultiModalDataItems, MultiModalDataParser)
6161
from vllm.multimodal.processing import BaseProcessingInfo, PromptReplacement
6262
from vllm.multimodal.processor import BaseMultiModalProcessor
63-
from vllm.multimodal.profiling import BaseDummyDataBuilder, ProcessorInputs
63+
from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
6464
from vllm.platforms import _Backend
6565
from vllm.sequence import IntermediateTensors
6666
from vllm.transformers_utils.config import uses_mrope
@@ -883,7 +883,7 @@ def get_max_video_tokens(self, seq_len: int) -> int:
883883
)
884884

885885

886-
class Qwen2VLDummyDataBuilder(BaseDummyDataBuilder[Qwen2VLProcessingInfo]):
886+
class Qwen2VLDummyInputsBuilder(BaseDummyInputsBuilder[Qwen2VLProcessingInfo]):
887887

888888
def get_dummy_processor_inputs(
889889
self,
@@ -991,7 +991,7 @@ def _get_mm_fields_config(
991991

992992
@MULTIMODAL_REGISTRY.register_processor(Qwen2VLMultiModalProcessor,
993993
info=Qwen2VLProcessingInfo,
994-
dummy_data=Qwen2VLDummyDataBuilder)
994+
dummy=Qwen2VLDummyInputsBuilder)
995995
class Qwen2VLForConditionalGeneration(nn.Module, SupportsMultiModal,
996996
SupportsLoRA, SupportsPP):
997997
packed_modules_mapping = {

‎vllm/model_executor/models/ultravox.py

Lines changed: 4 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -27,7 +27,7 @@
2727
from vllm.multimodal.parse import MultiModalDataItems, MultiModalDataParser
2828
from vllm.multimodal.processing import BaseProcessingInfo, PromptReplacement
2929
from vllm.multimodal.processor import BaseMultiModalProcessor
30-
from vllm.multimodal.profiling import BaseDummyDataBuilder, ProcessorInputs
30+
from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
3131
from vllm.sequence import IntermediateTensors
3232
from vllm.transformers_utils.configs.ultravox import UltravoxConfig
3333

@@ -98,7 +98,8 @@ def get_mm_max_tokens_per_item(self, seq_len: int) -> Mapping[str, int]:
9898
return {"audio": max_audio_tokens}
9999

100100

101-
class UltravoxDummyDataBuilder(BaseDummyDataBuilder[UltravoxProcessingInfo]):
101+
class UltravoxDummyInputsBuilder(BaseDummyInputsBuilder[UltravoxProcessingInfo]
102+
):
102103

103104
def get_dummy_processor_inputs(
104105
self,
@@ -340,7 +341,7 @@ def forward(
340341

341342
@MULTIMODAL_REGISTRY.register_processor(UltravoxMultiModalProcessor,
342343
info=UltravoxProcessingInfo,
343-
dummy_data=UltravoxDummyDataBuilder)
344+
dummy=UltravoxDummyInputsBuilder)
344345
class UltravoxModel(nn.Module, SupportsMultiModal, SupportsPP):
345346

346347
hf_to_vllm_mapper = WeightsMapper(

0 commit comments

Comments
 (0)