fix: keep async rollouts version-consistent

- serialize shared-model optimizer updates with generation
- reject future or over-lagged rollout results after asynchronous scoring
- close cache publication races
- persist policy versions in online checkpoints
This commit is contained in:
0z5a
2026-09-03 12:01:24 +08:00
parent ce2f9d13b3
commit 587b0ee046
16 changed files with 531 additions and 49 deletions
+3
View File
@@ -164,6 +164,9 @@ class CheckpointCallback(TrainCallback):
**context.config.to_dict(),
"optimizer_step": context.optimizer_step,
}
policy_version = context.strategy.policy_version
if policy_version is not None:
meta["policy_version"] = policy_version
context.checkpoint = Checkpoint(
state_dict=state_dict,
epoch=context.epoch,