Skip to content

inputs

Contain classes used to define the inputs of a model.

Audio dataclass

Contains an audio that can be passed to a multimodal model.

Provide one or several instances of this class along with a text prompt in a list as the model_input argument to a model that supports audio processing.

Parameters:

Name Type Description Default
audio Any

The audio to use in the text generation.

 required
Source code in outlines/inputs.py 
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
@dataclass
class Audio:
    """Contains an audio that can be passed to a multimodal model.

    Provide one or several instances of this class along with a text prompt
    in a list as the `model_input` argument to a model that supports audio
    processing.

    Parameters
    ----------
    audio
        The audio to use in the text generation.

    """
    audio: Any

Image dataclass

Contains an image that can be passed to a multimodal model.

Provide one or several instances of this class along with a text prompt in a list as the model_input argument to a model that supports vision.

Parameters:

Name Type Description Default
image Image

The image to use in the text generation.

 required
Source code in outlines/inputs.py 
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
@dataclass
class Image:
    """Contains an image that can be passed to a multimodal model.

    Provide one or several instances of this class along with a text prompt
    in a list as the `model_input` argument to a model that supports vision.

    Parameters
    ----------
    image
        The image to use in the text generation.

    """
    image: PILImage.Image

    def __post_init__(self):
        image = self.image

        if not image.format:
            raise TypeError(
                "Could not read the format of the image passed to the model."
            )

        buffer = BytesIO()
        image.save(buffer, format=image.format)
        self.image_str = base64.b64encode(buffer.getvalue()).decode("utf-8")
        self.image_format = f"image/{image.format.lower()}"

Video dataclass

Contains a video that can be passed to a multimodal model.

Provide one or several instances of this class along with a text prompt in a list as the model_input argument to a model that supports video processing.

Parameters:

Name Type Description Default
video Any

The video to use in the text generation.

 required
Source code in outlines/inputs.py 
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
@dataclass
class Video:
    """Contains a video that can be passed to a multimodal model.

    Provide one or several instances of this class along with a text prompt
    in a list as the `model_input` argument to a model that supports video
    processing.

    Parameters
    ----------
    video
        The video to use in the text generation.

    """
    video: Any