Unverified Commit a0f05da8 by Chi Zhang Committed by GitHub

[ckpt] fix: fix oom when resume from ckpt (#402)

parent 052b0a39
......@@ -576,6 +576,9 @@ class ActorRolloutRefWorker(Worker):
if self._is_offload_param:
offload_fsdp_model_to_cpu(self.actor_module_fsdp)
if self._is_offload_optimizer:
offload_fsdp_optimizer(self.actor_optimizer)
class CriticWorker(Worker):
......@@ -848,6 +851,9 @@ class CriticWorker(Worker):
if self._is_offload_param:
offload_fsdp_model_to_cpu(self.critic_module)
if self._is_offload_optimizer:
offload_fsdp_optimizer(self.critic_optimizer)
# TODO(sgm): we may need to extract it to dp_reward_model.py
class RewardModelWorker(Worker):
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment