ragbench_eval/config.py

import os
from dotenv import load_dotenv

load_dotenv()

HF_TOKEN = os.getenv("HF_TOKEN")
GROQ_API_KEY = os.getenv("GROQ_API_KEY")

RAGBENCH_DATASET = "galileo-ai/ragbench"

#Will be "groq" if env var is missing OR empty
LLM_PROVIDER = (os.getenv("RAGBENCH_LLM_PROVIDER") or "groq").lower()
GEN_MODEL = os.getenv("RAGBENCH_GEN_MODEL", "llama-3.1-8b-instant")
JUDGE_MODEL = os.getenv("RAGBENCH_JUDGE_MODEL", "llama-3.1-70b-versatile")

EMBEDDING_MODEL = os.getenv(
    "RAGBENCH_EMBEDDING_MODEL",
    "sentence-transformers/all-MiniLM-L6-v2",
)

RAGBENCH_DATASET = os.getenv("RAGBENCH_DATASET", "galileo-ai/ragbench")

DOMAIN_TO_SUBSETS = {
    "biomedical": ["pubmedqa", "covidqa"],
    "general_knowledge": ["hotpotqa", "msmarco", "hagrid", "expertqa"],
    "legal": ["cuad"],
    "customer_support": ["delucionqa", "emanual", "techqa"],
    "finance": ["finqa", "tatqa"],
}