eval_baseline_v6.py

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

"""Baseline eval: LiquidAI/LFM2.5-1.2B-Instruct on standard benchmarks."""

import subprocess, sys

model_args = "model_name=LiquidAI/LFM2.5-1.2B-Instruct,trust_remote_code=True,dtype=bfloat16"
tasks = "leaderboard|mmlu|5,leaderboard|arc:challenge|25,leaderboard|truthfulqa:mc|0"

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