Documentation for FineTuningIteration and its Method parse

Class Overview

FineTuningIteration represents a specific iteration of the fine-tuning process. It stores the plugin name, run identifier, and batch identifier to distinguish different iterations. Built on pydantic's BaseModel, it offers data validation and easy serialization.

Attributes

• batch_id: Session batch identifier for the iteration.
• run_id: Run identifier of the starting model.
• plugin_name: Name of the tuned embedding.

Purpose

To encapsulate and validate iteration details in the fine-tuning workflow.

Example

Create an iteration instance:

iteration = FineTuningIteration(
    batch_id="001",
    run_id="run123",
    plugin_name="my_embedding"
)

Method: parse

Functionality

The parse method takes an experiment name string and returns a FineTuningIteration object. It supports parsing of both initial and regular experiment formats. In the initial case, only the plugin_name is set. Otherwise, run_id and batch_id are populated.

Parameters

• experiment_name: Experiment name string. Expected format:
• For initial experiments: "plugin_name / initial / ..."
• For regular experiments: "plugin_name / iteration / run_id / batch_id"

Purpose

Parse an experiment name and create a FineTuningIteration instance.

Example

For an initial experiment:

iteration = FineTuningIteration.parse(
    "my_plugin / initial / sample"
)

For a regular experiment:

iteration = FineTuningIteration.parse(
    "my_plugin / iteration / 12345 / 67890"
)