openai-core/com.aallam.openai.api.finetune/FineTuneRequest

FineTuneRequest

@Serializable

data class FineTuneRequest(val trainingFile: FileId, val validationFile: FileId? = null, val model: ModelId? = null, val nEpochs: Int? = null, val batchSize: Int? = null, val learningRateMultiplier: Double? = null, val promptLossWeight: Double? = null, val computeClassificationMetrics: Boolean? = null, val classificationNClasses: Int? = null, val classificationPositiveClass: String? = null, val classificationBetas: List<Double>? = null, val suffix: String? = null)

Create a Fine-Tune request.

Constructors

FineTuneRequest

constructor(trainingFile: FileId, validationFile: FileId? = null, model: ModelId? = null, nEpochs: Int? = null, batchSize: Int? = null, learningRateMultiplier: Double? = null, promptLossWeight: Double? = null, computeClassificationMetrics: Boolean? = null, classificationNClasses: Int? = null, classificationPositiveClass: String? = null, classificationBetas: List<Double>? = null, suffix: String? = null)

Properties

batchSize

@SerialName(value = "batch_size")

val batchSize: Int? = null

The batch size to use for training. The batch size is the number of training examples used to train a single forward and backward pass.

classificationBetas

@SerialName(value = "classification_betas")

val classificationBetas: List<Double>? = null

If this is provided, we calculate F-beta scores at the specified beta values. The F-beta score is a generalization of F-1 score. This is only used for binary classification.

classificationNClasses

@SerialName(value = "classification_n_classes")

val classificationNClasses: Int? = null

The number of classes in a classification task.

classificationPositiveClass

@SerialName(value = "classification_positive_class")

val classificationPositiveClass: String? = null

The positive class in binary classification.

computeClassificationMetrics

@SerialName(value = "compute_classification_metrics")

val computeClassificationMetrics: Boolean? = null

If set, we calculate classification-specific metrics such as accuracy and F-1 score using the validation set at the end of every epoch. These metrics can be viewed in the results file.

learningRateMultiplier

@SerialName(value = "learning_rate_multiplier")

val learningRateMultiplier: Double? = null

The learning rate multiplier to use for training. The fine-tuning learning rate is the original learning rate used for pretraining multiplied by this value.

model

@SerialName(value = "model")

val model: ModelId? = null

The name of the base model to fine-tune.

nEpochs

@SerialName(value = "n_epochs")

val nEpochs: Int? = null

The number of epochs to train the model for. An epoch refers to one full cycle through the training dataset.

promptLossWeight

@SerialName(value = "prompt_loss_weight")

val promptLossWeight: Double? = null

The weight to use for loss on the prompt tokens. This controls how much the model tries to learn to generate the prompt (as compared to the completion which always has a weight of 1.0), and can add a stabilizing effect to training when completions are short.

suffix

@SerialName(value = "suffix")

val suffix: String? = null

A string of up to 40 characters that will be added to your fine-tuned model name.

trainingFile

@SerialName(value = "training_file")

val trainingFile: FileId

The ID of an uploaded file that contains training data.

validationFile

@SerialName(value = "validation_file")

val validationFile: FileId? = null

The ID of an uploaded file that contains validation data.