fine_tuning

Fine-tuning Models for RaLEs Tasks

This directory contains code and utilities for fine-tuning models on the RaLEs datasets using pre-trained models from the HuggingFace model zoo.

Key Files & Descriptions:

1. cli.py: The primary command-line interface for fine-tuning models.

   • Usage: python cli.py --config [path_to_config.yaml]

2. constants.py: Contains various constants and configurations used across the fine-tuning scripts.

3. find_best_models.py: Script to identify and select the best performing models after training.

   • Usage: python find_best_models.py [additional_arguments]

4. configs: A directory containing configuration files for different models and tasks. Provided are configurations used to evaluate models for RaLEs. Ensure you select the appropriate configuration for your task and model.

5. utils: A directory with utility functions and scripts to assist in the fine-tuning process.

Step-by-step Guide:

1. Prepare Data: Make sure you've preprocessed the data (see instructions in the datasets directory).
2. Update constants.py file with appropriate path to datasets.
3. Fine-tuning: Use cli.py to initiate the fine-tuning process. Specify the path to a config YAML file.
4. Next, use your model for inference.