eval_humaneval.py

# /// script
# dependencies = [
#   "torch",
#   "transformers",
#   "accelerate",
# ]
# ///

import subprocess
import sys
import os

def install_harness():
    print("📦 Installing BigCode Evaluation Harness directly from GitHub...")
    # This command uses pip to "clone" and install the repository as a package
    subprocess.run([
        sys.executable, "-m", "pip", "install", 
        "git+https://github.com/bigcode-project/bigcode-evaluation-harness.git"
    ], check=True)

def run_eval():
    # UPDATE THIS with your merged model's path
    MODEL_ID = "moos124/Qwen2.5-1.5B-SSD-Python-Merged" 
    
    print(f"🚀 Starting HumanEval for {MODEL_ID}...")
    
    # We call the module using 'python -m lm_eval' 
    cmd = [
        "python", "-m", "lm_eval",
        "--model", "hf",
        "--model_args", f"pretrained={MODEL_ID},trust_remote_code=True",
        "--tasks", "humaneval",
        "--batch_size", "1", # Start small to verify it works
        "--precision", "bf16",
        "--allow_code_execution",
    ]
    
    # Run the benchmark
    subprocess.run(cmd)

if __name__ == "__main__":
    # install_harness()
    run_eval()