dodox/metastable-baselines2
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Fix bug when self.max_grad_norm=None

Browse Source
This commit is contained in:
Dominik Moritz Roth 2024-01-23 13:32:38 +01:00
parent c67f78159b
commit 8bbb01504b
1 changed files with 3 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
metastable_baselines2/trpl/trpl.py
View File

@ -92,7 +92,7 @@ class TRPL(BetterOnPolicyAlgorithm):
normalize_advantage: bool = True,
ent_coef: float = 0.0,
vf_coef: float = 0.5,
max_grad_norm: float = 0.5,
max_grad_norm: Union[float, None] = None,
use_sde: bool = False,
sde_sample_freq: int = -1,
use_pca: bool = False,
@ -306,7 +306,8 @@ class TRPL(BetterOnPolicyAlgorithm):
self.policy.optimizer.zero_grad()
loss.backward()
# Clip grad norm
th.nn.utils.clip_grad_norm_(self.policy.parameters(), self.max_grad_norm)
if self.max_grad_norm is not None:
th.nn.utils.clip_grad_norm_(self.policy.parameters(), self.max_grad_norm)
self.policy.optimizer.step()
self._n_updates += 1