config.py

import os
from dotenv import load_dotenv
from datasets import load_dataset

load_dotenv()

# --- CONFIGURATION ---
MODAL_BASE_URL = "https://mohdfanis--unsloth-model-arena-backend" # Base URL
MY_AUTH_TOKEN = os.environ.get("ARENA_AUTH_TOKEN")

# --- DATASETS ---
print("Loading Hugging Face datasets...")
try:
    math_dataset = load_dataset("microsoft/orca-math-word-problems-200k", split="train")
    bio_dataset = load_dataset("bio-nlp-umass/bioinstruct", split="train")
    print("✅ Datasets loaded successfully.")
except Exception as e:
    print(f"❌ Failed to load datasets: {e}")
    math_dataset, bio_dataset = [], []

DATASET_CONFIG = {
    "Math": {
        "dataset": math_dataset,
        "question_col": "question",
        "answer_col": "answer"
    },
    "Bio": {
        "dataset": bio_dataset,
        "instruction_col": "instruction",
        "input_col": "input",
        "answer_col": "output"
    }
}

# --- MODEL DEFINITIONS ---
BASE_MODELS = {
    "Base Llama-3.1 8B Instruct": "unsloth/llama-3.1-8b-instruct-bnb-4bit",
    "Base Llama-3 8B Instruct": "unsloth/llama-3-8b-instruct-bnb-4bit",
    "Base Llama-2 7B Chat": "unsloth/llama-2-7b-chat-bnb-4bit",
    "Base Mistral 7B Instruct": "unsloth/mistral-7b-v0.3-instruct-bnb-4bit",
    "Base Qwen-2 7B Instruct": "unsloth/qwen2-7B-instruct-bnb-4bit",
    "Base Gemma-2 9B Instruct": "unsloth/gemma-2-9b-it-bnb-4bit",
    "Base Gemma 7B Instruct": "unsloth/gemma-7b-it-bnb-4bit",
}

FINETUNED_MATH = {
    "Finetuned Llama-3.1 8B (e3) - MATH": "farhananis005/lora-llama-3.1-8b-Math-e3",
    "Finetuned Llama-3.1 8B (e1) - MATH": "farhananis005/lora-llama-3.1-8b-Math-e1",
    "Finetuned Llama-3 8B (e3) - MATH": "farhananis005/lora-llama-3-8b-Math-e3",
    "Finetuned Llama-3 8B (e1) - MATH": "farhananis005/lora-llama-3-8b-Math-e1",
    "Finetuned Llama-2 7B (e3) - MATH": "farhananis005/lora-llama-2-7b-Math-e3",
    "Finetuned Llama-2 7B (e1) - MATH": "farhananis005/lora-llama-2-7b-Math-e1",
    "Finetuned Mistral 7B (e3) - MATH": "farhananis005/lora-mistral-7b-v0.3-Math-e3",
    "Finetuned Mistral 7B (e1) - MATH": "farhananis005/lora-mistral-7b-v0.3-Math-e1",
    "Finetuned Qwen-2 7B (e3) - MATH": "farhananis005/lora-qwen-2-7b-Math-e3",
    "Finetuned Qwen-2 7B (e1) - MATH": "farhananis005/lora-qwen-2-7b-Math-e1",
    "Finetuned Gemma-2 9B (e3) - MATH": "farhananis005/lora-gemma-2-9b-Math-e3",
    "Finetuned Gemma-2 9B (e1) - MATH": "farhananis005/lora-gemma-2-9b-Math-e1",
    "Finetuned Gemma 7B (e3) - MATH": "farhananis005/lora-gemma-7b-Math-e3",
    "Finetuned Gemma 7B (e1) - MATH": "farhananis005/lora-gemma-7b-Math-e1",
}

FINETUNED_BIO = {
    "Finetuned Llama-3.1 8B (e3) - BIO": "farhananis005/lora-llama-3.1-8b-Bio-e3",
    "Finetuned Llama-3.1 8B (e1) - BIO": "farhananis005/lora-llama-3.1-8b-Bio-e1",
    "Finetuned Llama-3 8B (e3) - BIO": "farhananis005/lora-llama-3-8b-Bio-e3",
    "Finetuned Llama-3 8B (e1) - BIO": "farhananis005/lora-llama-3-8b-Bio-e1",
    "Finetuned Llama-2 7B (e3) - BIO": "farhananis005/lora-llama-2-7b-Bio-e3",
    "Finetuned Llama-2 7B (e1) - BIO": "farhananis005/lora-llama-2-7b-Bio-e1",
    "Finetuned Mistral 7B (e3) - BIO": "farhananis005/lora-mistral-7b-v0.3-Bio-e3",
    "Finetuned Mistral 7B (e1) - BIO": "farhananis005/lora-mistral-7b-v0.3-Bio-e1",
    "Finetuned Qwen-2 7B (e3) - BIO": "farhananis005/lora-qwen-2-7b-Bio-e3",
    "Finetuned Qwen-2 7B (e1) - BIO": "farhananis005/lora-qwen-2-7b-Bio-e1",
    "Finetuned Gemma-2 9B (e3) - BIO": "farhananis005/lora-gemma-2-9b-Bio-e3",
    "Finetuned Gemma-2 9B (e1) - BIO": "farhananis005/lora-gemma-2-9b-Bio-e1",
    "Finetuned Gemma 7B (e3) - BIO": "farhananis005/lora-gemma-7b-Bio-e3",
    "Finetuned Gemma 7B (e1) - BIO": "farhananis005/lora-gemma-7b-Bio-e1",
}

ALL_MODELS = {
    "Math": {"-- Select Math Model --": None, **BASE_MODELS, **FINETUNED_MATH},
    "Bio": {"-- Select Bio Model --": None, **BASE_MODELS, **FINETUNED_BIO}
}