Skip to content

Fine-Tuning Method

This is the main part you'll need to implement yourself.

You need to write a script that initiates the fine-tuning of the model. This script inherits from the FineTuningMethod class and should implement methods:

  • upload_initial_model: used for uploading your model to MLflow
  • get_fine_tuning_builder: used for configuring the model's fine-tuning.

An example of such a script can be found in the directory plugins/default_fine_tuning_method.py.

You can have multiple fine-tuning scripts. The choice of which script to use occurs when launching the task:

curl -X POST http://localhost:5000/api/v1/fine-tuning/task \
  -H 'Content-Type: application/json' \
  -d '{
    "fine_tuning_method": "Default Fine Tuning Method"
}'
where fine_tuning_method is the name of your fine-tuning script, taken from the meta.name field.

Note

The meta.name field is a required field in your FineTuningMethod.

The path to the plugins directory is specified in the ES_PLUGINS_PATH environment variable. By default, it points to the plugins directory at the project's root. It's easiest to change this in the .env file.

For more details on plugins, see the Plugins section.