Handle empty EVMbench summaries

project/evmbench/evmbench/nano/eval.py

@@ -181,7 +181,6 @@ async def get_full_summary(
     ) -> dict[str, Any]:
         ctx_logger = logger.bind(run_group_id=self.run_group_id, runs_dir=self.runs_dir)
 
-        tasks = [t for t, _ in results]
         final_results = [
             r.grade.evmbench_result
             for _, r in results
@@ -240,7 +239,7 @@ async def get_full_summary(
             "params": params,
             "run_health": run_health,
             "metrics": metrics,
-            "run_group_id": getattr(tasks[0], "run_group_id", "undefined"),
+            "run_group_id": self.run_group_id,
         }
 
         ctx_logger.info(

project/evmbench/tests/test_empty_summary.py

@@ -0,0 +1,15 @@
+import pytest
+
+from evmbench.nano.eval import EVMbench
+
+
+@pytest.mark.asyncio
+async def test_evmbench_full_summary_handles_empty_results() -> None:
+    eval = EVMbench()
+
+    summary = await eval.get_full_summary([])
+
+    assert summary["params"]["n_samples"] == 0
+    assert summary["metrics"]["score"] == 0
+    assert summary["metrics"]["max_score"] == 0
+    assert summary["run_group_id"] == eval.run_group_id