PokéEvaluator: Unleashing the Power of Model Fine-tuning on Pokémon Data

Dive deep into the realm of Pokémon with PokéEvaluator! This project seeks to compare the prowess of various models, both fine-tuned and vanilla, as they tackle the vibrant and intricate world of Pokémon data.

Requirements

Ensure you have the required packages installed:

pip install -r requirements.txt

Getting Started

Notebooks

1. Fine-tuning: Use the gpt_finetune.ipynb notebook to generate training data and format it for fine-tuning and fine-tune away.
2. Evaluation: Use the gpt_evaluate.ipynb notebook to evaluate the models. This leverages the PokemonGroundTruth class to produce and score the data related to Pokémon.

Model Interfacing

• LLM: Use the llm.py for the base language model structures and implementations.
• OpenAI: The gpt_openai.py for OpenAI-powered models within the project.
• Llama2: The llama2_exllama.py for Llama2-powered models within the project.

Other Scripts

• prompt.py: Provides utilities for generating and managing prompts for the language models.
• ground_truth_aiopoke.py: Manages the generation and scoring of Pokémon data.