Skip to content

Add --trend flag to benchmark.py #266

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
olearycrew merged 4 commits into from
Apr 9, 2026
+20 −4
Merged

Add --trend flag to benchmark.py #266

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
7221f33
Add --trend flag for post-run trend analysis
olearycrew Apr 9, 2026
33b55d0
Address review feedback
olearycrew Apr 9, 2026
a3ae51b
Remove unused top-level import (now lazy-loaded inside try block)
olearycrew Apr 9, 2026
78e82cd
Escape % in argparse help string
olearycrew Apr 9, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 16 additions & 4 deletions scripts/benchmark.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -251,17 +251,28 @@ def _parse_args() -> argparse.Namespace:
action="store_true",
help="Continue running all tasks even if sanity check scores 0%%",
)
parser.add_argument(
"--trend",
action="store_true",
help="Run trend analysis after benchmark completes (requires ≥2 runs in output dir)",
)
parser.add_argument(
"--trend-window",
type=int,
default=None,
default=10,
metavar="N",
help="Analyze score trends over the last N runs after benchmarking (requires ≥2 runs)",
help="Number of recent runs to include in trend analysis (default: 10)",
)
Comment on lines +259 to +265
Copy link

Copilot AI Apr 9, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

--trend-window accepts any int, including 0/negative values; with the current slicing (pts[-self.window:]) a value of 0 analyzes all runs and negative values produce surprising slices. Consider validating --trend-window as >= 2 (or >= 1, but then run() will always produce no output) and failing fast via parser.error(...) to avoid confusing behavior.

Copilot uses AI. Check for mistakes.
parser.add_argument(
"--trend-threshold",
type=float,
default=-0.5,
help="Slope (%%/run) below which regression is flagged (default: -0.5)",
)
args = parser.parse_args()

# Validate --trend-window
if args.trend_window is not None and args.trend_window < 2:
if args.trend_window < 2:
parser.error("--trend-window must be >= 2")

return args
Expand Down Expand Up @@ -846,13 +857,14 @@ def _build_and_write_results():
_log_category_summary(task_entries, tasks_by_id)
_log_efficiency_summary(efficiency, grades_by_task_id)
# Run trend analysis if requested
if args.trend_window is not None:
if args.trend:
try:
from lib_trend import RunTrendAnalyzer

analyzer = RunTrendAnalyzer(
results_dir=output_dir,
window=args.trend_window,
regression_threshold=args.trend_threshold,
)
analyzer.run(model=args.model)
except Exception as exc:
Expand Down
4 changes: 4 additions & 0 deletions scripts/lib_trend.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -88,6 +88,10 @@ def load_points(self, model: Optional[str] = None) -> Dict[str, List[RunPoint]]:
except (json.JSONDecodeError, OSError):
continue

# Skip incomplete runs (written incrementally during benchmark)
if data.get("in_progress"):
continue

m = data.get("model", "")
ts = data.get("timestamp", 0.0)
run_id = data.get("run_id", path.stem)
Expand Down
Loading