gqt/-----
Archived
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

修复PPO评估推理返回None异常

Browse Source
This commit is contained in:
郭启童
2026-04-26 15:35:19 +08:00
parent ba6cf2a797
commit 3d0a8122bb
3 changed files with 51 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
agent_ppo/workflow/train_workflow.py
View File

@@ -132,8 +132,12 @@ class EpisodeRunner:
final_reward = 0.0
if done:
fm = self.agent.preprocessor
env_info = env_obs["observation"]["env_info"]
extra_info = env_obs.get("extra_info", {})
observation = env_obs.get("observation") or {}
observation = observation if isinstance(observation, dict) else {}
env_info = observation.get("env_info") or {}
env_info = env_info if isinstance(env_info, dict) else {}
extra_info = env_obs.get("extra_info") or {}
extra_info = extra_info if isinstance(extra_info, dict) else {}
total_score = env_info.get("total_score", fm.total_score)
remaining_charge = env_info.get("remaining_charge", fm.remaining_charge)
charge_count = env_info.get("charge_count", fm.charge_count)