variant_a/scripts/print_results.py

#!/usr/bin/env python3
"""Print results summary from eval JSON."""
import json, sys

path = sys.argv[1] if len(sys.argv) > 1 else '/dev/shm/eval/results_a/full_results.json'
with open(path) as f:
    data = json.load(f)

if 'results' not in data:
    print("No results found!")
    sys.exit(1)

scores = []
for task, metrics in sorted(data['results'].items()):
    found = False
    for key in ['acc_norm', 'acc', 'f1', 'exact_match']:
        # Try both plain key and key,none suffix
        for k in [key, f'{key},none']:
            if k in metrics:
                val = metrics[k]
                if isinstance(val, (int, float)):
                    print(f"  {task}: {key} = {val:.4f} ({val*100:.2f}%)")
                    scores.append(val)
                    found = True
                    break
        if found:
            break
    if not found:
        # Print whatever numeric metrics exist
        for k, v in metrics.items():
            if isinstance(v, (int, float)) and not k.startswith('alias'):
                print(f"  {task}: {k} = {v:.4f}")

if scores:
    avg = sum(scores) / len(scores)
    print(f"\n  Average: {avg:.4f} ({avg*100:.2f}%)")
    print(f"  Baseline (IQ2_XXS): 61.34%")
    print(f"  FP16 Instruct:      65.71%")
    if avg * 100 > 61.34:
        print(f"  >>> BEATS BASELINE by {avg*100 - 61.34:.2f}pp <<<")
    else:
        print(f"  >>> Below baseline by {61.34 - avg*100:.2f}pp <<<")