drlab 0.2.6__tar.gz → 0.2.7__tar.gz

{drlab-0.2.6/src/drlab.egg-info → drlab-0.2.7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: drlab
-Version: 0.2.6
+Version: 0.2.7
 Summary: Deep Reinforcement Learning kit for research.
 Author: Tomas Osarte
 Keywords: reinforcement-learning,deep-learning,pytorch,gymnasium

{drlab-0.2.6 → drlab-0.2.7}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "drlab"
-version = "0.2.6"
+version = "0.2.7"
 description = "Deep Reinforcement Learning kit for research."
 readme = "README.md"
 requires-python = ">=3.11"

{drlab-0.2.6 → drlab-0.2.7}/src/drlab/__init__.py

@@ -35,7 +35,7 @@ from drlab.learners import (
 from drlab.replay import ReplayBuffer, TransitionBatch
 from drlab.runners import Runner
 
-__version__ = "0.2.6"
+__version__ = "0.2.7"
 
 __all__ = [
     "__version__",

{drlab-0.2.6 → drlab-0.2.7}/src/drlab/learners/off_policy/sac.py

@@ -196,12 +196,18 @@ class SACLearner(OffPolicyLearner):
         ])
 
         loss = actor_loss + critic_loss + alpha_loss
+
+        # A single batched host transfer avoids five separate device
+        # synchronizations (one per .item() call) on every training step.
+        actor_v, critic_v, alpha_v, reg_v, total_v = th.stack(
+            [policy_loss, critic_loss, alpha_loss, reg_loss, loss]
+        ).tolist()
         self.last_losses = {
-            "actor": float(policy_loss.item()),
-            "critic": float(critic_loss.item()),
-            "alpha": float(alpha_loss.item()),
-            "regularization": float(reg_loss.item()),
-            "total": float(loss.item()),
+            "actor": actor_v,
+            "critic": critic_v,
+            "alpha": alpha_v,
+            "regularization": reg_v,
+            "total": total_v,
         }
 
         return self.last_losses["total"]

{drlab-0.2.6 → drlab-0.2.7/src/drlab.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: drlab
-Version: 0.2.6
+Version: 0.2.7
 Summary: Deep Reinforcement Learning kit for research.
 Author: Tomas Osarte
 Keywords: reinforcement-learning,deep-learning,pytorch,gymnasium

{drlab-0.2.6 → drlab-0.2.7}/tests/test_public_api.py

@@ -37,7 +37,7 @@ class PublicAPITest(unittest.TestCase):
             "WarmupController",
         }
 
-        self.assertEqual(drlab.__version__, "0.2.5")
+        self.assertEqual(drlab.__version__, "0.2.7")
         self.assertTrue(expected.issubset(set(drlab.__all__)))
         for name in expected:
             self.assertTrue(hasattr(drlab, name), name)