Bring device_mesh=None to FSDPUlyssesShardingManager when SP is not enable

Author: A1waysBeenHere

verl/workers/engine/veomni/transformer_impl.py

@@ -95,8 +95,11 @@ def __init__(
         self._is_offload_optimizer = self.engine_config.optimizer_offload
         self._is_lora = self.model_config.lora_rank > 0
 
-        self.ulysses_sharding_manager = FSDPUlyssesShardingManager(parallel_state.get_parallel_state().device_mesh)
         self.use_ulysses_sp = parallel_state.get_parallel_state().sp_enabled
+        if self.use_ulysses_sp:
+            self.ulysses_sharding_manager = FSDPUlyssesShardingManager(parallel_state.get_parallel_state().device_mesh)
+        else:
+            self.ulysses_sharding_manager = FSDPUlyssesShardingManager(None)
 
         if self.engine_config.entropy_from_logits_with_chunking:
             entropy_from_logits = verl_F.entropy_from_logits_with_chunking