Options to configure the MediaPipe Universal Embedder Task.
Signature:
export declare interface UniversalEmbedderOptions
Properties
| Property | Type | Description |
|---|---|---|
| activationDataType | ActivationDataType | Optional activation data type for model execution. |
| baseOptions | UniversalEmbedderBaseOptions | Base options specifying the model asset and WebGPU device. |
| l2Normalize | boolean | Whether to L2-normalize the output embedding vector. Defaults to true. |
| maxInputLength | number | Optional maximum sequence length (in tokens) for text encoder signatures. |
| visionTokensPerImage | number | Optional vision tokens per image for vision encoder signatures. |
UniversalEmbedderOptions.activationDataType
Optional activation data type for model execution.
Signature:
activationDataType?: ActivationDataType;
UniversalEmbedderOptions.baseOptions
Base options specifying the model asset and WebGPU device.
Signature:
baseOptions: UniversalEmbedderBaseOptions;
UniversalEmbedderOptions.l2Normalize
Whether to L2-normalize the output embedding vector. Defaults to true.
Signature:
l2Normalize?: boolean;
UniversalEmbedderOptions.maxInputLength
Optional maximum sequence length (in tokens) for text encoder signatures.
Signature:
maxInputLength?: number;
UniversalEmbedderOptions.visionTokensPerImage
Optional vision tokens per image for vision encoder signatures.
Signature:
visionTokensPerImage?: number;