eval_baseline_v3.py

# /// script
# requires-python = ">=3.10"
# dependencies = [
#     "lighteval[vllm]>=0.6.0",
#     "torch>=2.0.0",
#     "transformers>=4.40.0",
#     "accelerate>=0.30.0",
#     "vllm>=0.4.0",
# ]
# ///

"""Evaluate baseline LiquidAI/LFM2.5-1.2B-Instruct."""

import subprocess
import sys

model_args = "pretrained=LiquidAI/LFM2.5-1.2B-Instruct,trust_remote_code=True,add_bos_token=True"
tasks = "leaderboard|mmlu|5,leaderboard|arc:challenge|25,leaderboard|truthfulqa:mc|0"

cmd = ["lighteval", "vllm", model_args, tasks, "--output-dir", "./eval_baseline"]
print(f"Running: {' '.join(cmd)}")
result = subprocess.run(cmd)
sys.exit(result.returncode)