Struct TogetherFormat 

pub struct TogetherFormat {
    pub use_chat_format: bool,
}

Together AI fine-tuning format.

Uses OpenAI-compatible chat format but with text wrapper: {"text": "<s>[INST] ... [/INST] ..."}

For chat format (preferred), same as OpenAI: {"messages": [...]}

Fields

use_chat_format: bool

If true, use chat messages format (OpenAI-compatible). If false, use text template.

Implementations

impl TogetherFormat

pub fn chat() -> Self

Create a Together format using chat messages (OpenAI-compatible).

pub fn text() -> Self

Create a Together format using text template wrapping.

Trait Implementations

impl Default for TogetherFormat

fn default() -> Self

Returns the “default value” for a type.

impl FormatConverter for TogetherFormat

fn name(&self) -> &str

Name of this format (e.g., “openai”, “alpaca”).

fn to_json(&self, example: &TrainingExample) -> DatasetResult<Value>

Convert a TrainingExample to this format’s JSON representation.

fn parse_json(&self, value: &Value) -> DatasetResult<TrainingExample>

Parse this format’s JSON back into a TrainingExample.

fn to_json_batch( &self, examples: &[TrainingExample], ) -> DatasetResult<Vec<Value>>

Convert a batch of examples to this format.

fn parse_json_batch( &self, values: &[Value], ) -> DatasetResult<Vec<TrainingExample>>

Parse a batch of JSON values into training examples.

impl PreferenceConverter for TogetherFormat

fn name(&self) -> &str

Name of this format.

fn preference_to_json(&self, pair: &PreferencePair) -> DatasetResult<Value>

Convert a PreferencePair to this format’s JSON representation.

fn parse_preference_json(&self, value: &Value) -> DatasetResult<PreferencePair>

Parse this format’s JSON back into a PreferencePair.

fn preference_to_json_batch( &self, pairs: &[PreferencePair], ) -> DatasetResult<Vec<Value>>

Convert a batch of preference pairs to this format.

fn parse_preference_json_batch( &self, values: &[Value], ) -> DatasetResult<Vec<PreferencePair>>

Parse a batch of JSON values into preference pairs.