Interface PreferenceOptimizationHyperParameters

Hyperparameters for Preference Optimization. This data type is not supported in Gemini API.

interface PreferenceOptimizationHyperParameters {
    adapterSize?: AdapterSize;
    beta?: number;
    epochCount?: string;
    learningRateMultiplier?: number;
}

Properties

adapterSize?: AdapterSize

Optional. Adapter size for preference optimization.

beta?: number

Optional. Weight for KL Divergence regularization.

epochCount?: string

Optional. Number of complete passes the model makes over the entire training dataset during training.

learningRateMultiplier?: number

Optional. Multiplier for adjusting the default learning rate.