Improve Hub upload robustness: return detailed error messages to UI and print exceptions to logs

app.py

@@ -159,16 +159,19 @@ def _append_local_persist_csv(task_name: str, row: Dict[str, object]) -> bool:
         return False
 
 
-def _upload_eval_record_to_dataset(task_name: str, row: Dict[str, object]) -> bool:
+def _upload_eval_record_to_dataset(task_name: str, row: Dict[str, object]) -> tuple[bool, str]:
     """Upload a single-eval JSONL record to a dataset repo.
     Repo is taken from EVAL_REPO_ID env or defaults to 'peiranli0930/VisEval'.
+    Returns (ok, message) for UI feedback and debugging.
     """
     if HfApi is None:
-        return False
+        return False, "huggingface_hub 未安装"
     token = os.environ.get("HF_TOKEN") or os.environ.get("HUGGINGFACEHUB_API_TOKEN")
     repo_id = os.environ.get("EVAL_REPO_ID", "peiranli0930/VisEval")
-    if not token or not repo_id:
-        return False
+    if not token:
+        return False, "未找到写入 Token (HF_TOKEN/HUGGINGFACEHUB_API_TOKEN)"
+    if not repo_id:
+        return False, "未设置 EVAL_REPO_ID"
     try:
         from huggingface_hub import CommitOperationAdd
 
@@ -185,9 +188,14 @@ def _upload_eval_record_to_dataset(task_name: str, row: Dict[str, object]) -> bo
             operations=operations,
             commit_message=f"Add eval {folder} {row.get('test_id')} {uid}",
         )
-        return True
-    except Exception:
-        return False
+        return True, f"上传成功: {repo_id}/{path_in_repo}"
+    except Exception as e:
+        # Print to logs for debugging in Space
+        try:
+            print("[VisArena] Upload to dataset failed:", repr(e))
+        except Exception:
+            pass
+        return False, f"异常: {type(e).__name__}: {e}"
 
 
 def on_task_change(task_name: str, _state_pairs: List[Dict[str, str]]):
@@ -260,12 +268,12 @@ def on_submit(
     }
     row = _build_eval_row(pair, score_map)
     ok_local = _append_local_persist_csv(task_name, row)
-    ok_hub = _upload_eval_record_to_dataset(task_name, row)
+    ok_hub, hub_msg = _upload_eval_record_to_dataset(task_name, row)
 
     next_index = min(index + 1, len(pairs) - 1)
     info = f"已保存 Test ID {pair['test_id']} 的评价结果。"
     info += " 本地持久化" + ("成功" if ok_local else "失败") + "。"
-    info += " 上传Hub" + ("成功" if ok_hub else "失败") + "。"
+    info += " 上传Hub" + ("成功" if ok_hub else "失败") + (f"（{hub_msg}）" if hub_msg else "") + "。"
 
     if next_index != index:
         pair = pairs[next_index]