fix: wrong accelerator name when node_rank is not set (RLinf#370)

Signed-off-by: Hao Lin <linhaomails@gmail.com>

Author: andylin-hao

rlinf/scheduler/cluster/node.py

@@ -428,7 +428,7 @@ def _sort_nodes(self, cluster_num_nodes: int):
             # NODE_RANK not set, sort first by accelerator type, then by IP
             nodes_group_by_accel: dict[str, list[NodeInfo]] = {}
             for node in self._nodes:
-                accel_name = f"{node.accelerator_type.value}_{node.accelerator_model}"
+                accel_name = node.accelerator_type
                 nodes_group_by_accel.setdefault(accel_name, [])
                 nodes_group_by_accel[accel_name].append(node)
             for accel_name in nodes_group_by_accel.keys():