stable-baselines3 2.3.0a5__tar.gz → 2.3.2__tar.gz

{stable_baselines3-2.3.0a5/stable_baselines3.egg-info → stable_baselines3-2.3.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: stable_baselines3
-Version: 2.3.0a5
+Version: 2.3.2
 Summary: Pytorch version of Stable Baselines, implementations of reinforcement learning algorithms.
 Home-page: https://github.com/DLR-RM/stable-baselines3
 Author: Antonin Raffin

{stable_baselines3-2.3.0a5 → stable_baselines3-2.3.2}/README.md

@@ -85,7 +85,7 @@ Documentation is available online: [https://sb3-contrib.readthedocs.io/](https:/
 
 ## Stable-Baselines Jax (SBX)
 
-[Stable Baselines Jax (SBX)](https://github.com/araffin/sbx) is a proof of concept version of Stable-Baselines3 in Jax.
+[Stable Baselines Jax (SBX)](https://github.com/araffin/sbx) is a proof of concept version of Stable-Baselines3 in Jax, with recent algorithms like DroQ or CrossQ.
 
 It provides a minimal number of features compared to SB3 but can be much faster (up to 20x times!): https://twitter.com/araffin2/status/1590714558628253698
 

{stable_baselines3-2.3.0a5 → stable_baselines3-2.3.2}/stable_baselines3/common/save_util.py

@@ -447,7 +447,8 @@ def load_from_zip_file(
                     file_content.seek(0)
                     # Load the parameters with the right ``map_location``.
                     # Remove ".pth" ending with splitext
-                    th_object = th.load(file_content, map_location=device, weights_only=True)
+                    # Note(antonin): we cannot use weights_only=True, as it breaks with PyTorch 1.13, see GH#1911
+                    th_object = th.load(file_content, map_location=device, weights_only=False)
                     # "tensors.pth" was renamed "pytorch_variables.pth" in v0.9.0, see PR #138
                     if file_path == "pytorch_variables.pth" or file_path == "tensors.pth":
                         # PyTorch variables (not state_dicts)

{stable_baselines3-2.3.0a5 → stable_baselines3-2.3.2}/stable_baselines3/common/utils.py

@@ -92,7 +92,9 @@ def get_schedule_fn(value_schedule: Union[Schedule, float]) -> Schedule:
         value_schedule = constant_fn(float(value_schedule))
     else:
         assert callable(value_schedule)
-    return value_schedule
+    # Cast to float to avoid unpickling errors to enable weights_only=True, see GH#1900
+    # Some types are have odd behaviors when part of a Schedule, like numpy floats
+    return lambda progress_remaining: float(value_schedule(progress_remaining))
 
 
 def get_linear_fn(start: float, end: float, end_fraction: float) -> Schedule:

stable_baselines3-2.3.2/stable_baselines3/version.txt

@@ -0,0 +1 @@
+2.3.2

{stable_baselines3-2.3.0a5 → stable_baselines3-2.3.2/stable_baselines3.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: stable_baselines3
-Version: 2.3.0a5
+Version: 2.3.2
 Summary: Pytorch version of Stable Baselines, implementations of reinforcement learning algorithms.
 Home-page: https://github.com/DLR-RM/stable-baselines3
 Author: Antonin Raffin

{stable_baselines3-2.3.0a5 → stable_baselines3-2.3.2}/tests/test_save_load.py

@@ -783,3 +783,17 @@ def test_no_resource_warning(tmp_path):
         fp.seek(0)
         model.load_replay_buffer(fp)
         assert not fp.closed
+
+
+def test_cast_lr_schedule(tmp_path):
+    # See GH#1900
+    model = PPO("MlpPolicy", "Pendulum-v1", learning_rate=lambda t: t * np.sin(1.0))
+    # Note: for recent version of numpy, np.float64 is a subclass of float
+    # so we need to use type here
+    # assert isinstance(model.lr_schedule(1.0), float)
+    assert type(model.lr_schedule(1.0)) is float  # noqa: E721
+    assert np.allclose(model.lr_schedule(0.5), 0.5 * np.sin(1.0))
+    model.save(tmp_path / "ppo.zip")
+    model = PPO.load(tmp_path / "ppo.zip")
+    assert type(model.lr_schedule(1.0)) is float  # noqa: E721
+    assert np.allclose(model.lr_schedule(0.5), 0.5 * np.sin(1.0))

stable_baselines3-2.3.0a5/stable_baselines3/version.txt

@@ -1 +0,0 @@
-2.3.0a5