Update inference.py

inference.py

@@ -98,14 +98,7 @@ def main():
         difficulty = DIFFICULTIES[task_id]
 
         # ── [START] log ──────────────────────────────────────────
-        print(json.dumps({
-            "type": "[START]",
-            "task_id": task_id,
-            "task_name": task_name,
-            "difficulty": difficulty,
-            "model": MODEL_NAME,
-            "n_episodes": N_EPISODES
-        }), flush=True)
+        print(f"[START] task={task_name}", flush=True)
 
         episode_scores = []
         for seed in range(N_EPISODES):
@@ -113,26 +106,13 @@ def main():
             episode_scores.append(reward)
 
             # ── [STEP] log ────────────────────────────────────────
-            print(json.dumps({
-                "type": "[STEP]",
-                "task_id": task_id,
-                "episode": seed,
-                "reward": reward
-            }), flush=True)
+            print(f"[STEP] step={seed+1} reward={reward}", flush=True)
 
         avg_score = round(sum(episode_scores) / len(episode_scores), 4)
         all_results[task_id] = avg_score
 
         # ── [END] log ─────────────────────────────────────────────
-        print(json.dumps({
-            "type": "[END]",
-            "task_id": task_id,
-            "task_name": task_name,
-            "difficulty": difficulty,
-            "avg_score": avg_score,
-            "scores": episode_scores,
-            "done": True
-        }), flush=True)
+        print(f"[END] task={task_name} score={avg_score} steps={len(episode_scores)}", flush=True)
 
     overall = round(sum(all_results.values()) / 3, 4)
 
@@ -140,11 +120,7 @@ def main():
     with open("scores.json", "w") as f:
         json.dump({"tasks": all_results, "overall": overall}, f, indent=2)
 
-    print(json.dumps({
-        "type": "[SUMMARY]",
-        "task_scores": all_results,
-        "overall": overall
-    }), flush=True)
+    print(f"[SUMMARY] overall_score={overall} task_scores={all_results}", flush=True)
 
 
 if __name__ == "__main__":