Skip to content

vllm.model_executor.models.skyworkr1v

Classes:

SkyworkR1VImageEmbeddingInputs

Bases: TensorSchema

Dimensions: - ni: Number of images - ifs: Image feature size - hs: Hidden size (must match the hidden size of language model backbone)

Source code in vllm/model_executor/models/skyworkr1v.py 
73
74
75
76
77
78
79
80
81
82
83
84
85
86
class SkyworkR1VImageEmbeddingInputs(TensorSchema):
    """Dimensions:
    - ni: Number of images
    - ifs: Image feature size
    - hs: Hidden size (must match the hidden size of language model
      backbone)
    """

    type: Literal["image_embeds"] = "image_embeds"

    data: Annotated[
        torch.Tensor | list[torch.Tensor],
        TensorShape("ni", "ifs", "hs"),
    ]

SkyworkR1VImagePixelInputs

Bases: TensorSchema

Dimensions: - bnp: Batch size * number of images * (1 + num_patches) - c: Number of channels (3) - h: Height - w: Width - bn: Batch size * number of images

Source code in vllm/model_executor/models/skyworkr1v.py 
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
class SkyworkR1VImagePixelInputs(TensorSchema):
    """Dimensions:
    - bnp: Batch size * number of images * (1 + num_patches)
    - c: Number of channels (3)
    - h: Height
    - w: Width
    - bn: Batch size * number of images
    """

    type: Literal["pixel_values"] = "pixel_values"

    pixel_values_flat: Annotated[
        torch.Tensor,
        TensorShape("bnp", 3, "h", "w"),
    ]

    num_patches: Annotated[
        torch.Tensor,
        TensorShape("bn"),
    ]