google-genai-kotlin/com.google.genai.kotlin.types/ReinforcementTuningHyperParameters

ReinforcementTuningHyperParameters

@Serializable

data class ReinforcementTuningHyperParameters(val epochCount: Long? = null, val learningRateMultiplier: Double? = null, val adapterSize: AdapterSize? = null, val samplesPerPrompt: Int? = null, val batchSize: Int? = null, val evaluateInterval: Int? = null, val checkpointInterval: Int? = null, val maxOutputTokens: Int? = null, val thinkingLevel: ReinforcementTuningThinkingLevel? = null, val thinkingBudget: Int? = null)

Hyperparameters for Reinforcement Tuning.

Constructors

constructor(epochCount: Long? = null, learningRateMultiplier: Double? = null, adapterSize: AdapterSize? = null, samplesPerPrompt: Int? = null, batchSize: Int? = null, evaluateInterval: Int? = null, checkpointInterval: Int? = null, maxOutputTokens: Int? = null, thinkingLevel: ReinforcementTuningThinkingLevel? = null, thinkingBudget: Int? = null)

Properties

adapterSize

val adapterSize: AdapterSize? = null

Optional. Adapter size for Reinforcement Tuning.

batchSize

val batchSize: Int? = null

Optional. Batch size for the tuning job. How many prompts to process at a train step. If not set, the batch size will be determined automatically.

checkpointInterval

val checkpointInterval: Int? = null

Optional. How often at steps to save checkpoints during training. If not set, one checkpoint per epoch will be set. ``total_steps = epoch_count * samples_per_prompt / total_prompts_in_dataset``

epochCount

val epochCount: Long? = null

Optional. Number of training epoches for the tuning job.

evaluateInterval

val evaluateInterval: Int? = null

Optional. How often at steps to evaluate the tuning job during training. If not set, evel will be run per epoch. total_steps = epoch_count * samples_per_prompt / total_prompts_in_dataset

learningRateMultiplier

val learningRateMultiplier: Double? = null

Learning rate multiplier for Reinforcement Learning.

maxOutputTokens

val maxOutputTokens: Int? = null

Optional. The maximum number of tokens to generate per prompt. Default to 32768.

samplesPerPrompt

val samplesPerPrompt: Int? = null

Optional. Number of different responses to generate per prompt during tuning.

thinkingBudget

val thinkingBudget: Int? = null

Optional. The thinking budget for the tuning job to optimize for (Gemini 2.5 only). * -1 means dynamic thinking * 0 means no thinking * 0 means thinking budget in tokens If not set, default to -1 (dynamic thinking).

thinkingLevel

val thinkingLevel: ReinforcementTuningThinkingLevel? = null

Indicates the maximum thinking depth during tuning. Starting from Gemini 3.5 models, the old thinking_budget will no longer be supported and will result in a user error if set. Instead, users should use the thinking_level parameter to control the maximum thinking depth.