Skip to content
KortexyaKortexyaReasoning Layer SDK

Interface: RlTrainResponse

@kortexya/reasoninglayer

@kortexya/reasoninglayer / RlTrainResponse

Interface: RlTrainResponse

Section titled “Interface: RlTrainResponse”

Defined in: src/types/rl-training.ts:195

Response from an RL training run.

Properties

Section titled “Properties”

avgRewardLast10

Section titled “avgRewardLast10”

avgRewardLast10: number

Defined in: src/types/rl-training.ts:196

cognitiveCyclesRun

Section titled “cognitiveCyclesRun”

cognitiveCyclesRun: number

Defined in: src/types/rl-training.ts:197

collapseEpisodes?

Section titled “collapseEpisodes?”

optional collapseEpisodes?: number[]

Defined in: src/types/rl-training.ts:198

contrastiveLosses?

Section titled “contrastiveLosses?”

optional contrastiveLosses?: number[]

Defined in: src/types/rl-training.ts:199

conversationDecisions?

Section titled “conversationDecisions?”

optional conversationDecisions?: Record<string, number>

Defined in: src/types/rl-training.ts:201

Conversation-specific metrics when training on a conversation environment.

curriculumAdjustments?

Section titled “curriculumAdjustments?”

optional curriculumAdjustments?: number

Defined in: src/types/rl-training.ts:202

curriculumBacktracks?

Section titled “curriculumBacktracks?”

optional curriculumBacktracks?: number

Defined in: src/types/rl-training.ts:203

curriculumStageReached?

Section titled “curriculumStageReached?”

optional curriculumStageReached?: number | null

Defined in: src/types/rl-training.ts:204

deliberativeDecisions?

Section titled “deliberativeDecisions?”

optional deliberativeDecisions?: number

Defined in: src/types/rl-training.ts:205

dynamicsLosses?

Section titled “dynamicsLosses?”

optional dynamicsLosses?: number[]

Defined in: src/types/rl-training.ts:206

episodeRewards

Section titled “episodeRewards”

episodeRewards: number[]

Defined in: src/types/rl-training.ts:207

episodesCompleted

Section titled “episodesCompleted”

episodesCompleted: number

Defined in: src/types/rl-training.ts:209

episodeSteps

Section titled “episodeSteps”

episodeSteps: number[]

Defined in: src/types/rl-training.ts:208

expertAvgEntropy?

Section titled “expertAvgEntropy?”

optional expertAvgEntropy?: number

Defined in: src/types/rl-training.ts:210

expertAvgWeights?

Section titled “expertAvgWeights?”

optional expertAvgWeights?: Record<string, number>

Defined in: src/types/rl-training.ts:211

expertCollapseRecoveries?

Section titled “expertCollapseRecoveries?”

optional expertCollapseRecoveries?: number

Defined in: src/types/rl-training.ts:212

expertFinalWeights?

Section titled “expertFinalWeights?”

optional expertFinalWeights?: Record<string, number>

Defined in: src/types/rl-training.ts:213

finalContrastiveTemperature?

Section titled “finalContrastiveTemperature?”

optional finalContrastiveTemperature?: number

Defined in: src/types/rl-training.ts:214

finalExplorationRate

Section titled “finalExplorationRate”

finalExplorationRate: number

Defined in: src/types/rl-training.ts:215

finalImaginationSteps?

Section titled “finalImaginationSteps?”

optional finalImaginationSteps?: number

Defined in: src/types/rl-training.ts:216

imaginationTransitions?

Section titled “imaginationTransitions?”

optional imaginationTransitions?: number

Defined in: src/types/rl-training.ts:217

modeDistribution?

Section titled “modeDistribution?”

optional modeDistribution?: Record<string, number>

Defined in: src/types/rl-training.ts:218

peerDemonstrationsConsumed?

Section titled “peerDemonstrationsConsumed?”

optional peerDemonstrationsConsumed?: number

Defined in: src/types/rl-training.ts:219

plasticityPhaseTransitions?

Section titled “plasticityPhaseTransitions?”

optional plasticityPhaseTransitions?: unknown[][]

Defined in: src/types/rl-training.ts:220

reactiveDecisions?

Section titled “reactiveDecisions?”

optional reactiveDecisions?: number

Defined in: src/types/rl-training.ts:221

selfPlayAgentFitnesses?

Section titled “selfPlayAgentFitnesses?”

optional selfPlayAgentFitnesses?: number[][]

Defined in: src/types/rl-training.ts:222

selfPlayBestFitness?

Section titled “selfPlayBestFitness?”

optional selfPlayBestFitness?: number | null

Defined in: src/types/rl-training.ts:223

selfPlayMeanFitnessPerRound?

Section titled “selfPlayMeanFitnessPerRound?”

optional selfPlayMeanFitnessPerRound?: number[]

Defined in: src/types/rl-training.ts:224

sortAwareTrainingSteps?

Section titled “sortAwareTrainingSteps?”

optional sortAwareTrainingSteps?: number

Defined in: src/types/rl-training.ts:225

sortsDiscovered?

Section titled “sortsDiscovered?”

optional sortsDiscovered?: number

Defined in: src/types/rl-training.ts:226

totalTransitions

Section titled “totalTransitions”

totalTransitions: number

Defined in: src/types/rl-training.ts:227

trajectoriesExchanged?

Section titled “trajectoriesExchanged?”

optional trajectoriesExchanged?: number

Defined in: src/types/rl-training.ts:228

transcripts?

Section titled “transcripts?”

optional transcripts?: RlTranscriptSummary

Defined in: src/types/rl-training.ts:233

Where the transcripts came from. Present for conversation environments only.

weightsLoaded?

Section titled “weightsLoaded?”

optional weightsLoaded?: boolean

Defined in: src/types/rl-training.ts:239

Whether a previously persisted DQN checkpoint was loaded before training (warm start). Only ever true with persistWeights + enableNeuralDqn and an existing tenant checkpoint.

weightsPersisted?

Section titled “weightsPersisted?”

optional weightsPersisted?: boolean

Defined in: src/types/rl-training.ts:244

Whether the trained DQN weights were persisted to the tenant checkpoint after training.