Skip to content

SWE-agent takes a GitHub issue and tries to automatically fix it, using your LM of choice. It can also be employed for offensive cybersecurity or competitive coding challenges. [NeurIPS 2024]

Enh(retry): Directly save best attmpt idx

Kilian Lieret54868da067c5
diff --git a/sweagent/agent/agents.py b/sweagent/agent/agents.py
index 333fd256..7b75056d 100644
--- a/sweagent/agent/agents.py
+++ b/sweagent/agent/agents.py
@@ -551,6 +551,7 @@ class Agent:
"attempts": [get_attempt_data(i) for i in range(self._i_attempt + 1)],
**get_attempt_data(best_attempt_idx),
}
+ data["info"]["best_attempt_idx"] = best_attempt_idx
# Total model stats
data["info"]["model_stats"] = self.model.stats.model_dump()
data["extra_info"] = {"comparisons": [(a, b, comp.model_dump()) for a, b, comp in self._rloop.comparisons]}