Create fine-tune

Creates a job that fine-tunes a specified model based on a given dataset. See this page for details.

Usage

create_fine_tune(
  training_file,
  validation_file = NULL,
  model,
  n_epochs = 4,
  batch_size = NULL,
  learning_rate_multiplier = NULL,
  prompt_loss_weight = 0.1,
  compute_classification_metrics = FALSE,
  classification_n_classes = NULL,
  classification_positive_class = NULL,
  classification_betas = NULL,
  suffix = NULL,
  openai_api_key = Sys.getenv("OPENAI_API_KEY"),
  openai_organization = NULL
)

Arguments

training_file

required; a length one character vector.

validation_file

optional; defaults to NULL; a length one character vector.

model

required; a length one character vector.

n_epochs

required; defaults to 4; a length one numeric vector with the integer value greater than 0.

batch_size

optional; defaults to NULL; a length one numeric vector with the integer value greater than 0.

learning_rate_multiplier

optional; defaults to NULL; a length one numeric vector with the value greater than 0.

prompt_loss_weight

required; defaults to 0.1; a length one numeric vector.

compute_classification_metrics

required; defaults to FLASE; a length one logical vector.

classification_n_classes

optional; defaults to NULL; a length one numeric vector with the value greater than 0.

classification_positive_class

optional; defaults to NULL; a length one character vector.

classification_betas

optional; defaults to NULL; a list elements of which are numeric values greater than 0.

suffix

optional; defaults to NULL; a length one character vector.

openai_api_key

required; defaults to Sys.getenv("OPENAI_API_KEY") (i.e., the value is retrieved from the .Renviron file); a length one character vector. Specifies OpenAI API key.

openai_organization

optional; defaults to NULL; a length one character vector. Specifies OpenAI organization.

Value

Returns a list, elements of which contain information about the fine-tune.

Details

For arguments description please refer to the official documentation.

See also

Other fine-tune functions: cancel_fine_tune(), delete_fine_tune_model(), list_fine_tune_events(), list_fine_tunes(), retrieve_fine_tune()

Examples

if (FALSE) { # \dontrun{
training_file <- system.file(
    "extdata", "sport_prepared_train.jsonl", package = "openai"
)
validation_file <- system.file(
    "extdata", "sport_prepared_train.jsonl", package = "openai"
)

training_info <- upload_file(training_file, "fine-tune")
validation_info <- upload_file(validation_file, "fine-tune")

info <- create_fine_tune(
    training_file = training_info$id,
    validation_file = validation_info$id,
    model = "ada",
    compute_classification_metrics = TRUE,
    classification_positive_class = " baseball" # Mind space in front
)
} # }