Skip to content

vllm.models.step5.amd.vl

Inference-only Step5 vision-language model on ROCm.

Classes:

Step5ForConditionalGeneration

Bases: Step3p7ForConditionalGeneration

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

Step5-specific weight names are mapped by the text model's own mapper.

Source code in vllm/models/step5/amd/vl.py 
11
12
13
14
15
16
17
18
19
20
21
22
23
class Step5ForConditionalGeneration(Step3p7ForConditionalGeneration):
    """Step5 VLM using the Step3p7 multimodal frontend and Step5 text model.

    Step5-specific weight names are mapped by the text model's own mapper."""

    packed_modules_mapping = Step5ForCausalLM.packed_modules_mapping

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