hongshaorou/dreamerv3-torch
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix missing advantage computation when reward_EMA is disabled

Browse Source 
This PR fixes an issue where reward_EMA=False caused adv to be undefined in _compute_actor_loss.

Previously, adv was only computed inside the reward_EMA branch, which resulted in a runtime error when the option was disabled.
This commit is contained in:
Unai Ruiz
2026-03-03 16:34:52 +01:00
committed by GitHub
parent 6253f988fe
commit a8456e95bc
1 changed files with 2 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
models.py
View File

@@ -409,6 +409,8 @@ class ImagBehavior(nn.Module):
metrics.update(tools.tensorstats(normed_target, "normed_target")) metrics.update(tools.tensorstats(normed_target, "normed_target"))
metrics["EMA_005"] = to_np(self.ema_vals[0]) metrics["EMA_005"] = to_np(self.ema_vals[0])
metrics["EMA_095"] = to_np(self.ema_vals[1]) metrics["EMA_095"] = to_np(self.ema_vals[1])
else:
adv = target - base
if self._config.imag_gradient == "dynamics": if self._config.imag_gradient == "dynamics":
actor_target = adv actor_target = adv