diff --git a/swanlab/converter/wb/wb_converter.py b/swanlab/converter/wb/wb_converter.py
index 5f0dfae4..673459d1 100644
--- a/swanlab/converter/wb/wb_converter.py
+++ b/swanlab/converter/wb/wb_converter.py
@@ -31,6 +31,13 @@ def parse_wandb_logs(self, wb_project: str, wb_entity: str, wb_run_id: str = Non
             raise TypeError(
                 "Wandb Converter requires wandb. Install with 'pip install wandb'."
             )
+            
+        try:
+            import pandas as pd
+        except ImportError as e:
+            raise TypeError(
+                "Wandb Converter requires pandas when process wandb logs. Install with 'pip install pandas'."
+            )
 
         client = wandb.Api()
 
@@ -51,11 +58,16 @@ def parse_wandb_logs(self, wb_project: str, wb_entity: str, wb_run_id: str = Non
                     workspace=self.workspace,
                     experiment_name=wb_run.name,
                     description=wb_run.notes,
-                    mode="cloud" if self.cloud else "local",
+                    cloud=self.cloud,
                     logdir=self.logdir,
                 )
             else:
                 swanlab_run = swanlab.get_run()
+                
+            try:
+                wb_run_metadata = {"wandb_metadata": wb_run_metadata}
+            except:
+                wb_run_metadata = None
 
             wb_config = {
                 "wandb_run_id": wb_run.id,
@@ -64,8 +76,8 @@ def parse_wandb_logs(self, wb_project: str, wb_entity: str, wb_run_id: str = Non
                 "wandb_user": wb_run.user,
                 "wandb_tags": wb_run.tags,
                 "wandb_url": wb_run.url,
-                "wandb_metadata": wb_run.metadata,
             }
+            wb_config.update(wb_run_metadata)
 
             swanlab_run.config.update(wb_config)
             swanlab_run.config.update(wb_run.config)
@@ -73,7 +85,13 @@ def parse_wandb_logs(self, wb_project: str, wb_entity: str, wb_run_id: str = Non
             # Get the first history record to extract available keys
             history = wb_run.history(stream="default")
             if len(history) > 0:
-                keys = [key for key in history[0].keys() if not key.startswith("_")]
+                # 检查 history 是否为 DataFrame 类型
+                if isinstance(history, pd.DataFrame):
+                    # 如果是 DataFrame，直接获取列名
+                    keys = [key for key in history.columns if not key.startswith("_")]
+                else:
+                    # 原来的逻辑，假设是字典列表
+                    keys = [key for key in history[0].keys() if not key.startswith("_")]
             else:
                 keys = []