Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
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
6 changes: 3 additions & 3 deletions benchmark.py
Original file line number Diff line number Diff line change
Expand Up @@ -560,10 +560,10 @@ def val_emoji(val):
console.print(table)

# Show apples-to-apples status
if comp.get("both_validated"):
if comp.get("both_validated") or (single_valid and multi_valid):
console.print("\n[green]✅ Apples-to-apples comparison: Both outputs validated successfully[/green]")
elif single_val is not None or multi_val is not None:
console.print("\n[yellow]⚠️ Comparison caveat: Outputs differ in quality (see report for details)[/yellow]")
elif (single_val is not None or multi_val is not None) and single_valid != multi_valid:
console.print("\n[yellow]⚠️ Comparison caveat: Output validation differs between runs (see report for details)[/yellow]")


def main():
Expand Down
6 changes: 4 additions & 2 deletions src/report_generator.py
Original file line number Diff line number Diff line change
Expand Up @@ -369,8 +369,8 @@ def _generate_postmortem_analysis(self) -> str:
report += "\n"

# Test count comparison
s_tests = sv.get("test_validation", {}) if sv else {}
m_tests = mv.get("test_validation", {}) if mv else {}
s_tests = (sv.get("test_validation") or {}) if sv else {}
m_tests = (mv.get("test_validation") or {}) if mv else {}
s_passed = s_tests.get("passed", 0)
m_passed = m_tests.get("passed", 0)

Expand Down Expand Up @@ -796,6 +796,8 @@ def fmt_bytes(s):
return report

def save_report(self):
self._calculate_comparison()

json_file = self.output_dir / "benchmark_report.json"
with open(json_file, "w") as f:
json.dump(self.report_data, f, indent=2, default=str)
Expand Down