Skip to content

vllm.model_executor.models.smolvlm

SmolVLMForConditionalGeneration

Bases: Idefics3ForConditionalGeneration

Source code in vllm/model_executor/models/smolvlm.py 
43
44
45
46
47
48
49
50
51
52
@MULTIMODAL_REGISTRY.register_processor(SmolVLMMultiModalProcessor,
                                        info=SmolVLMProcessingInfo,
                                        dummy_inputs=SmolVLMDummyInputsBuilder)
class SmolVLMForConditionalGeneration(Idefics3ForConditionalGeneration):

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

__init__ 

__init__(*, vllm_config: VllmConfig, prefix: str = '')
Source code in vllm/model_executor/models/smolvlm.py 
48
49
50
51
52
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
    super().__init__(
        vllm_config=vllm_config,
        prefix=prefix,
    )

SmolVLMProcessingInfo

Bases: Idefics3ProcessingInfo

Source code in vllm/model_executor/models/smolvlm.py 
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
class SmolVLMProcessingInfo(Idefics3ProcessingInfo):

    def get_hf_processor(
        self,
        *,
        max_image_size: Optional[dict[str, int]] = None,
        **kwargs: object,
    ) -> SmolVLMProcessor:
        if max_image_size is not None:
            kwargs["max_image_size"] = max_image_size

        return self.ctx.get_hf_processor(SmolVLMProcessor, **kwargs)

    def _get_image_token(
            self, processor: Optional[SmolVLMProcessor]) -> tuple[str, str]:
        if processor is None:
            processor = self.get_hf_processor()
        image_token = processor.image_token
        fake_image_token = processor.fake_image_token
        global_image_token = processor.global_image_token
        return image_token, fake_image_token, global_image_token

_get_image_token 

_get_image_token(
    processor: Optional[SmolVLMProcessor],
) -> tuple[str, str]
Source code in vllm/model_executor/models/smolvlm.py 
33
34
35
36
37
38
39
40
def _get_image_token(
        self, processor: Optional[SmolVLMProcessor]) -> tuple[str, str]:
    if processor is None:
        processor = self.get_hf_processor()
    image_token = processor.image_token
    fake_image_token = processor.fake_image_token
    global_image_token = processor.global_image_token
    return image_token, fake_image_token, global_image_token

get_hf_processor 

get_hf_processor(
    *,
    max_image_size: Optional[dict[str, int]] = None,
    **kwargs: object,
) -> SmolVLMProcessor
Source code in vllm/model_executor/models/smolvlm.py 
22
23
24
25
26
27
28
29
30
31
def get_hf_processor(
    self,
    *,
    max_image_size: Optional[dict[str, int]] = None,
    **kwargs: object,
) -> SmolVLMProcessor:
    if max_image_size is not None:
        kwargs["max_image_size"] = max_image_size

    return self.ctx.get_hf_processor(SmolVLMProcessor, **kwargs)