diff --git a/slime/backends/megatron_utils/actor.py b/slime/backends/megatron_utils/actor.py index fd5027645..8b0424fab 100644 --- a/slime/backends/megatron_utils/actor.py +++ b/slime/backends/megatron_utils/actor.py @@ -183,9 +183,10 @@ def init( # empty cache after initialization clear_memory() - if self.args.offload_train: - # recover to actor in the end. + if self._active_model_tag != "actor": self._switch_model("actor") + + if self.args.offload_train: self.sleep() self.rollout_engines = None