Skip to content

llmcompressor.modeling.kimi_k3.kimi_k3_vision_processing

Image processor class for Kimi-K3.

Classes:

KimiK3VisionProcessor 

KimiK3VisionProcessor(media_proc_cfg: dict, **kwargs)

Bases: BaseImageProcessor

Methods:

  • make_image_prompt –

    Build the K3 image placeholder with resolution info.

  • preprocess –

    Preprocess a atom vision input (images) into model-ready tensors.

Source code in src/llmcompressor/modeling/kimi_k3/kimi_k3_vision_processing.py 
27
28
29
30
31
32
33
def __init__(
    self,
    media_proc_cfg: dict,
    **kwargs,
):
    super().__init__(**kwargs)
    self.media_proc_cfg = media_proc_cfg

make_image_prompt classmethod 

make_image_prompt(width: int, height: int) -> str

Build the K3 image placeholder with resolution info.

Source code in src/llmcompressor/modeling/kimi_k3/kimi_k3_vision_processing.py 
57
58
59
60
61
62
63
@classmethod
def make_image_prompt(cls, width: int, height: int) -> str:
    """Build the K3 image placeholder with resolution info."""
    return (
        f"<|media_begin|>image {width}x{height}"
        f"<|media_content|><|media_pad|><|media_end|>"
    )

preprocess 

preprocess(
    medias: list[MediaInput],
    return_tensors: Optional[Union[str, TensorType]] = None,
) -> BatchFeature

Preprocess a atom vision input (images) into model-ready tensors.

Args: medias: List of MediaInput. return_tensors: Desired output format ('pt', 'np', 'tf', or None).

Returns: BatchFeature containing 'pixel_values' and 'grid_thws' tensors.

Source code in src/llmcompressor/modeling/kimi_k3/kimi_k3_vision_processing.py 
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
def preprocess(
    self,
    medias: list[MediaInput],
    return_tensors: Optional[Union[str, TensorType]] = None,
) -> BatchFeature:
    """
    Preprocess a atom vision input (images) into model-ready tensors.

    Args:
        medias: List of MediaInput.
        return_tensors: Desired output format ('pt', 'np', 'tf', or None).

    Returns:
        BatchFeature containing 'pixel_values' and 'grid_thws' tensors.
    """
    if not isinstance(medias, list):
        medias = [medias]
    if medias:
        pixel_values = []
        for item in medias:
            item = ensure_media_type(
                item,
                transparent_bg_config=self._transparent_bg_config,
                transparent_bg_fill_stage=self._transparent_bg_fill_stage,
            )
            resize_config = self.get_resize_config(item)
            new_width, new_height, pad_width, pad_height = (
                resize_config["new_width"],
                resize_config["new_height"],
                resize_config["pad_width"],
                resize_config["pad_height"],
            )
            if item["type"] == "image":
                image = item["image"]
                image_np = self.resize_image(
                    image, new_width, new_height, pad_width, pad_height
                )
                pixel_values.append(np.expand_dims(image_np, axis=0))
            else:
                raise ValueError("Unsupported type: {}".format(item["type"]))
        normalized_pixel_values = []
        image_std_inv = 1.0 / np.array(self.media_proc_cfg["image_std"])
        image_mean = np.array(self.media_proc_cfg["image_mean"])
        for pixels in pixel_values:
            pixels = normalize(pixels, image_mean, image_std_inv)
            pixels_and_thw = navit_patchify(
                pixels,
                self.media_proc_cfg["patch_size"],
            )
            normalized_pixel_values.append(pixels_and_thw)

        pixel_values = torch.cat(
            [
                _to_tensor(pixel_value["pixel_values"])
                for pixel_value in normalized_pixel_values
            ]
        )
        grid_thws = torch.cat(
            [
                _to_tensor(pixel_value["grid_thw"], dtype=torch.int64).unsqueeze(0)
                for pixel_value in normalized_pixel_values
            ]
        )

        data = {
            "pixel_values": pixel_values,
            "grid_thws": grid_thws,
        }

    else:
        data = {}

    return BatchFeature(data=data, tensor_type=return_tensors)