vllm.models.step5.nvidia.vl

Inference-only Step5 vision-language model.

Classes:

• Step5ForConditionalGeneration –

  Step5 VLM using the Step3p7 multimodal frontend and Step5 text model.

Step5ForConditionalGeneration

Bases: Step3p7ForConditionalGeneration, SupportsLoRA

Step5 VLM using the Step3p7 multimodal frontend and Step5 text model.

Source code in vllm/models/step5/nvidia/vl.py

class Step5ForConditionalGeneration(Step3p7ForConditionalGeneration, SupportsLoRA):
    """Step5 VLM using the Step3p7 multimodal frontend and Step5 text model."""

    packed_modules_mapping = STEP5_PACKED_MODULES_MAPPING
    hf_to_vllm_mapper = (
        Step3p7ForConditionalGeneration.hf_to_vllm_mapper | STEP5_SHARE_EXPERT_MAPPER
    )

    def get_mm_mapping(self) -> MultiModelKeys:
        return MultiModelKeys.from_string_field(
            language_model="language_model",
            connector="vit_large_projector",
            tower_model="vision_model",
        )