allow eval to fail

app.py

@@ -144,7 +144,12 @@ def _backend_routine():
     for model_id, sha in pending_models:
         logger.info(f"Running evaluation on {model_id}")
         report = {"config": {"model_id": model_id, "model_sha": sha}}
-        evaluations = evaluate(model_id, revision=sha)
+        try:
+            evaluations = evaluate(model_id, revision=sha)
+        except Exception as e:
+            logger.error(f"Error evaluating {model_id}: {e}")
+            evaluations = None
+
         if evaluations is not None:
             report["results"] = evaluations
             report["status"] = "DONE"