mani-skill-nightly 2025.10.22.143__py3-none-any.whl → 2025.10.22.157__py3-none-any.whl

mani_skill/utils/structs/base.py

@@ -2,7 +2,7 @@ from __future__ import annotations
 
 from dataclasses import dataclass
 from functools import cached_property
-from typing import TYPE_CHECKING, Generic, List, TypeVar
+from typing import TYPE_CHECKING, Generic, TypeVar
 
 import numpy as np
 import sapien.physx as physx
@@ -24,7 +24,7 @@ class BaseStruct(Generic[T]):
     Base class of all structs that manage sapien objects on CPU/GPU
     """
 
-    _objs: List[T]
+    _objs: list[T]
     """list of objects of type T managed by this dataclass. This should not be modified after initialization. The struct hash is dependent on the hash of this list."""
     _scene_idxs: torch.Tensor
     """a list of indexes parallel to `self._objs` indicating which sub-scene each of those objects are actually in by index"""
@@ -67,7 +67,7 @@ class BaseStruct(Generic[T]):
 
 @dataclass
 class PhysxRigidBaseComponentStruct(BaseStruct[T], Generic[T]):
-    _bodies: List[physx.PhysxRigidBaseComponent]
+    _bodies: list[physx.PhysxRigidBaseComponent]
 
     # ---------------------------------------------------------------------------- #
     # API from physx.PhysxRigidBaseComponent
@@ -91,7 +91,7 @@ class PhysxRigidBaseComponentStruct(BaseStruct[T], Generic[T]):
 
 @dataclass
 class PhysxRigidBodyComponentStruct(PhysxRigidBaseComponentStruct[T], Generic[T]):
-    _bodies: List[physx.PhysxRigidBodyComponent]
+    _bodies: list[physx.PhysxRigidBodyComponent]
     _body_data_name: str
     _body_data_index_internal: slice = None
 
@@ -210,7 +210,7 @@ class PhysxRigidBodyComponentStruct(PhysxRigidBaseComponentStruct[T], Generic[T]
         else:
             return torch.tensor(
                 np.array([body.angular_velocity for body in self._bodies]),
-                device=self.device
+                device=self.device,
             )
 
     @property
@@ -265,7 +265,9 @@ class PhysxRigidBodyComponentStruct(PhysxRigidBaseComponentStruct[T], Generic[T]
             # for link entities, namely 7:10 was angular velocity and 10:13 was linear velocity. SAPIEN 3.0.0 and above fixes this
             return self._body_data[self._body_data_index, 7:10]
         else:
-            return torch.from_numpy(self._bodies[0].linear_velocity[None, :]).to(self.device)
+            return torch.from_numpy(self._bodies[0].linear_velocity[None, :]).to(
+                self.device
+            )
 
     @property
     def mass(self) -> torch.Tensor:
@@ -298,7 +300,7 @@ class PhysxRigidBodyComponentStruct(PhysxRigidBaseComponentStruct[T], Generic[T]
 
 @dataclass
 class PhysxRigidDynamicComponentStruct(PhysxRigidBodyComponentStruct[T], Generic[T]):
-    _bodies: List[physx.PhysxRigidDynamicComponent]
+    _bodies: list[physx.PhysxRigidDynamicComponent]
 
     def get_angular_velocity(self) -> torch.Tensor:
         return self.angular_velocity
@@ -357,7 +359,9 @@ class PhysxRigidDynamicComponentStruct(PhysxRigidBodyComponentStruct[T], Generic
         if self.scene.gpu_sim_enabled:
             return self._body_data[self._body_data_index, 10:13]
         else:
-            return torch.from_numpy(self._bodies[0].angular_velocity[None, :]).to(self.device)
+            return torch.from_numpy(self._bodies[0].angular_velocity[None, :]).to(
+                self.device
+            )
 
     @angular_velocity.setter
     def angular_velocity(self, arg1: Array):
@@ -426,7 +430,8 @@ class PhysxRigidDynamicComponentStruct(PhysxRigidBodyComponentStruct[T], Generic
             return self._body_data[self._body_data_index, 7:10]
         else:
             return torch.tensor(
-                np.array([body.linear_velocity for body in self._bodies]), device=self.device
+                np.array([body.linear_velocity for body in self._bodies]),
+                device=self.device,
             )
 
     @linear_velocity.setter

mani_skill/utils/structs/drive.py

@@ -2,7 +2,7 @@ from __future__ import annotations
 
 import typing
 from dataclasses import dataclass
-from typing import TYPE_CHECKING, List, Sequence, Union
+from typing import TYPE_CHECKING, Sequence, Union
 
 import sapien
 import sapien.physx as physx
@@ -38,7 +38,7 @@ class Drive(PhysxJointComponentStruct[physx.PhysxDriveComponent]):
         pose1: Union[sapien.Pose, Pose] = None,
         scene_idxs: torch.Tensor = None,
     ):
-        physx_drives: List[physx.PhysxDriveComponent] = []
+        physx_drives: list[physx.PhysxDriveComponent] = []
         assert bodies1 is not None
         if bodies0 is None:
             bodies0 = [None] * len(bodies1)

mani_skill/utils/structs/link.py

@@ -2,7 +2,7 @@ from __future__ import annotations
 
 from dataclasses import dataclass, field
 from functools import cached_property
-from typing import TYPE_CHECKING, Callable, Dict, List, Union
+from typing import TYPE_CHECKING, Callable, Union
 
 import sapien
 import sapien.physx as physx
@@ -38,7 +38,7 @@ class Link(PhysxRigidBodyComponentStruct[physx.PhysxArticulationLinkComponent]):
     joint: ArticulationJoint = None
     """the joint of which this link is a child of. If this is a view/merged link then this joint is also a view/merged joint"""
 
-    meshes: Dict[str, List[trimesh.Trimesh]] = field(default_factory=dict)
+    meshes: dict[str, list[trimesh.Trimesh]] = field(default_factory=dict)
     """
     map from user-defined mesh groups (e.g. "handle" meshes for cabinets) to a list of trimesh.Trimesh objects corresponding to each physx link object managed here
     """
@@ -58,7 +58,7 @@ class Link(PhysxRigidBodyComponentStruct[physx.PhysxArticulationLinkComponent]):
     @classmethod
     def create(
         cls,
-        physx_links: List[physx.PhysxArticulationLinkComponent],
+        physx_links: list[physx.PhysxArticulationLinkComponent],
         scene: ManiSkillScene,
         scene_idxs: torch.Tensor,
     ):
@@ -75,7 +75,7 @@ class Link(PhysxRigidBodyComponentStruct[physx.PhysxArticulationLinkComponent]):
         )
 
     @classmethod
-    def merge(cls, links: List["Link"], name: str = None):
+    def merge(cls, links: list["Link"], name: str = None):
         objs = []
         joint_objs = []
         merged_joint_indexes = []
@@ -126,7 +126,7 @@ class Link(PhysxRigidBodyComponentStruct[physx.PhysxArticulationLinkComponent]):
         """
         Returns each managed link objects render shape list (a list of lists)
         """
-        all_render_shapes: List[List[sapien.render.RenderShape]] = []
+        all_render_shapes: list[list[sapien.render.RenderShape]] = []
         for obj in self._objs:
             rb_comp = obj.entity.find_component_by_type(
                 sapien.render.RenderBodyComponent
@@ -137,7 +137,7 @@ class Link(PhysxRigidBodyComponentStruct[physx.PhysxArticulationLinkComponent]):
 
     def get_visual_meshes(
         self, to_world_frame: bool = True, first_only: bool = False
-    ) -> List[trimesh.Trimesh]:
+    ) -> list[trimesh.Trimesh]:
         """
         Returns the visual mesh of each managed link object. Note results of this are not cached or optimized at the moment
         so this function can be slow if called too often
@@ -183,7 +183,7 @@ class Link(PhysxRigidBodyComponentStruct[physx.PhysxArticulationLinkComponent]):
         filter: Callable[
             [physx.PhysxArticulationLinkComponent, sapien.render.RenderShape], bool
         ],
-    ) -> List[trimesh.primitives.Box]:
+    ) -> list[trimesh.primitives.Box]:
         # First we need to pre-compute the bounding box of the link at 0. This will be slow the first time
         bboxes = []
         for link, link_render_shapes in zip(self._objs, self.render_shapes):
@@ -244,7 +244,9 @@ class Link(PhysxRigidBodyComponentStruct[physx.PhysxArticulationLinkComponent]):
                 raw_pose = new_pose
             return Pose.create(raw_pose)
         else:
-            return Pose.create([obj.entity_pose for obj in self._objs], device=self.device)
+            return Pose.create(
+                [obj.entity_pose for obj in self._objs], device=self.device
+            )
 
     @pose.setter
     def pose(self, arg1: Union[Pose, sapien.Pose, Array]) -> None:

mani_skill/utils/structs/pose.py

@@ -1,5 +1,5 @@
 from dataclasses import dataclass
-from typing import List, Optional, Union
+from typing import Optional, Union
 
 import numpy as np
 import sapien
@@ -21,7 +21,7 @@ def add_batch_dim(x):
     return x
 
 
-def to_batched_tensor(x: Union[List, Array], device: Optional[Device] = None):
+def to_batched_tensor(x: Union[list, Array], device: Optional[Device] = None):
     if x is None:
         return None
     return add_batch_dim(common.to_tensor(x, device=device))
@@ -121,7 +121,7 @@ class Pose:
     @classmethod
     def create(
         cls,
-        pose: Union[torch.Tensor, sapien.Pose, List[sapien.Pose], "Pose"],
+        pose: Union[torch.Tensor, sapien.Pose, list[sapien.Pose], "Pose"],
         device: Optional[Device] = None,
     ) -> "Pose":
         """Creates a Pose object from a given ``pose``, which can be a torch tensor, sapien.Pose, list of sapien.Pose, or Pose"""

mani_skill/utils/structs/render_camera.py

@@ -1,5 +1,5 @@
 from dataclasses import dataclass
-from typing import TYPE_CHECKING, Any, List, Union
+from typing import TYPE_CHECKING, Any, Union
 
 from mani_skill.utils.geometry.rotation_conversions import quaternion_to_matrix
 
@@ -30,7 +30,7 @@ class RenderCamera:
     Wrapper around sapien.render.RenderCameraComponent
     """
 
-    _render_cameras: List[sapien.render.RenderCameraComponent]
+    _render_cameras: list[sapien.render.RenderCameraComponent]
     name: str
     # NOTE (stao): I cannot seem to use ManiSkillScene as a type here, it complains it is undefined despite using TYPE_CHECKING variable. Without typchecking there is a ciruclar import error
     scene: Any
@@ -47,7 +47,7 @@ class RenderCamera:
     @classmethod
     def create(
         cls,
-        render_cameras: List[sapien.render.RenderCameraComponent],
+        render_cameras: list[sapien.render.RenderCameraComponent],
         scene: Any,
         mount: Union[Actor, Link] = None,
     ):
@@ -157,7 +157,7 @@ class RenderCamera:
     def get_near(self) -> float:
         return self._render_cameras[0].get_near()
 
-    def get_picture(self, names: Union[str, List[str]]) -> List[torch.Tensor]:
+    def get_picture(self, names: Union[str, list[str]]) -> list[torch.Tensor]:
         if isinstance(names, str):
             names = [names]
         if self.scene.gpu_sim_enabled and not self.scene.parallel_in_single_scene:

mani_skill/utils/visualization/jupyter_utils.py

@@ -1,12 +1,10 @@
-from typing import List
-
 import matplotlib.pyplot as plt
 import numpy as np
 from IPython.display import HTML, display
 from matplotlib import animation
 
 
-def display_images(images: List[np.ndarray], dpi=100.0, format="html5_video", **kwargs):
+def display_images(images: list[np.ndarray], dpi=100.0, format="html5_video", **kwargs):
     """Display images as an animation in jupyter notebook.
 
     Args:

mani_skill/utils/visualization/misc.py

@@ -1,5 +1,5 @@
 import os
-from typing import Dict, List, Optional
+from typing import Optional
 
 import imageio
 import numpy as np
@@ -11,7 +11,7 @@ from mani_skill.utils.structs.types import Array
 
 
 def images_to_video(
-    images: List[Array],
+    images: list[Array],
     output_dir: str,
     video_name: str,
     fps: int = 10,
@@ -51,7 +51,7 @@ def images_to_video(
     writer.close()
 
 
-def tile_images(images: List[Array], nrows=1) -> Array:
+def tile_images(images: list[Array], nrows=1) -> Array:
     """
     Tile multiple images to a single image comprised of nrows and an appropriate number of columns to fit all the images.
     The images can also be batched (e.g. of shape (B, H, W, C)), but give images must all have the same batch size.
@@ -118,7 +118,7 @@ def tile_images(images: List[Array], nrows=1) -> Array:
 TEXT_FONT = None
 
 
-def put_text_on_image(image: np.ndarray, lines: List[str]):
+def put_text_on_image(image: np.ndarray, lines: list[str]):
     global TEXT_FONT
     assert image.dtype == np.uint8, image.dtype
     image = image.copy()
@@ -139,7 +139,7 @@ def put_text_on_image(image: np.ndarray, lines: List[str]):
     return np.array(image)
 
 
-def put_info_on_image(image, info: Dict[str, float], extras=None, overlay=True):
+def put_info_on_image(image, info: dict[str, float], extras=None, overlay=True):
     lines = [
         f"{k}: {v:.3f}" if isinstance(v, float) else f"{k}: {v}"
         for k, v in info.items()

mani_skill/utils/wrappers/cached_reset.py

@@ -1,5 +1,5 @@
 from dataclasses import asdict, dataclass
-from typing import List, Optional, Union
+from typing import Optional, Union
 
 import dacite
 import gymnasium as gym
@@ -66,7 +66,9 @@ class CachedResetWrapper(gym.Wrapper):
             self._cached_resets_obs_buffer = []
             while self._num_cached_resets < self.cached_resets_config.num_resets:
                 obs, _ = self.env.reset(
-                    seed=self.cached_resets_config.seed if self._num_cached_resets == 0 else None,
+                    seed=self.cached_resets_config.seed
+                    if self._num_cached_resets == 0
+                    else None,
                     options=dict(
                         env_idx=torch.arange(
                             0,
@@ -125,7 +127,7 @@ class CachedResetWrapper(gym.Wrapper):
     def reset(
         self,
         *args,
-        seed: Optional[Union[int, List[int]]] = None,
+        seed: Optional[Union[int, list[int]]] = None,
         options: Optional[dict] = None,
         **kwargs
     ):

mani_skill/utils/wrappers/flatten.py

@@ -1,5 +1,4 @@
 import copy
-from typing import Dict
 
 import gymnasium as gym
 import gymnasium.spaces.utils
@@ -42,7 +41,7 @@ class FlattenRGBDObservationWrapper(gym.ObservationWrapper):
         new_obs = self.observation(self.base_env._init_raw_obs)
         self.base_env.update_obs_space(new_obs)
 
-    def observation(self, observation: Dict):
+    def observation(self, observation: dict):
         sensor_data = observation.pop("sensor_data")
         del observation["sensor_param"]
         rgb_images = []

mani_skill/utils/wrappers/record.py

@@ -2,7 +2,7 @@ import copy
 import time
 from dataclasses import dataclass
 from pathlib import Path
-from typing import Callable, List, Optional, Union
+from typing import Callable, Optional, Union
 
 import gymnasium as gym
 import h5py
@@ -119,21 +119,21 @@ class RecordEpisode(gym.Wrapper):
 
     Each JSON file contains:
 
-    - `env_info` (Dict): task (also known as environment) information, which can be used to initialize the task
+    - `env_info` (dict): task (also known as environment) information, which can be used to initialize the task
     - `env_id` (str): task id
     - `max_episode_steps` (int)
-    - `env_kwargs` (Dict): keyword arguments to initialize the task. **Essential to recreate the environment.**
-    - `episodes` (List[Dict]): episode information
+    - `env_kwargs` (dict): keyword arguments to initialize the task. **Essential to recreate the environment.**
+    - `episodes` (list[dict]): episode information
     - `source_type` (Optional[str]): a simple category string describing what process generated the trajectory data. ManiSkill official datasets will usually write one of "human", "motionplanning", or "rl" at the moment.
     - `source_desc` (Optional[str]): a longer explanation of how the data was generated.
 
     The episode information (the element of `episodes`) includes:
 
     - `episode_id` (int): a unique id to index the episode
-    - `reset_kwargs` (Dict): keyword arguments to reset the task. **Essential to reproduce the trajectory.**
+    - `reset_kwargs` (dict): keyword arguments to reset the task. **Essential to reproduce the trajectory.**
     - `control_mode` (str): control mode used for the episode.
     - `elapsed_steps` (int): trajectory length
-    - `info` (Dict): information at the end of the episode.
+    - `info` (dict): information at the end of the episode.
 
     With just the meta data, you can reproduce the task the same way it was created when the trajectories were collected as so:
 
@@ -356,7 +356,7 @@ class RecordEpisode(gym.Wrapper):
     def reset(
         self,
         *args,
-        seed: Optional[Union[int, List[int]]] = None,
+        seed: Optional[Union[int, list[int]]] = None,
         options: Optional[dict] = None,
         save=True,
         **kwargs,
@@ -367,7 +367,9 @@ class RecordEpisode(gym.Wrapper):
             # if doing a full reset then we flush all trajectories including incompleted ones
             if self._trajectory_buffer is not None:
                 if options is None or "env_idx" not in options:
-                    self.flush_trajectory(env_idxs_to_flush=np.arange(self.num_envs), save=save)
+                    self.flush_trajectory(
+                        env_idxs_to_flush=np.arange(self.num_envs), save=save
+                    )
                 else:
                     self.flush_trajectory(
                         env_idxs_to_flush=common.to_numpy(options["env_idx"], save=save)

mani_skill/utils/wrappers/visual_encoders.py

@@ -1,4 +1,4 @@
-from typing import Dict, Literal
+from typing import Literal
 
 import gymnasium as gym
 import torch
@@ -38,7 +38,7 @@ class VisualEncoderWrapper(gym.ObservationWrapper):
         super().__init__(env)
 
     @torch.no_grad()
-    def observation(self, obs: Dict):
+    def observation(self, obs: dict):
         vec_img_embeddings_list = []
         image_obs = obs.pop("sensor_data")
         del obs["sensor_param"]

mani_skill/vector/wrappers/gymnasium.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-from typing import TYPE_CHECKING, Dict, List, Optional, Tuple, Union
+from typing import TYPE_CHECKING, Optional, Tuple, Union
 
 import gymnasium as gym
 import torch
@@ -36,7 +36,7 @@ class ManiSkillVectorEnv(VectorEnv):
     def __init__(
         self,
         env: Union[Env, str],
-        num_envs: int = None,
+        num_envs: int = 1,
         auto_reset: bool = True,
         ignore_terminations: bool = False,
         record_metrics: bool = False,
@@ -78,7 +78,7 @@ class ManiSkillVectorEnv(VectorEnv):
 
     @property
     def base_env(self) -> BaseEnv:
-        return self._env.unwrapped
+        return self._env.unwrapped  # type: ignore
 
     @property
     def unwrapped(self):
@@ -87,13 +87,15 @@ class ManiSkillVectorEnv(VectorEnv):
     def reset(
         self,
         *,
-        seed: Optional[Union[int, List[int]]] = None,
+        seed: Optional[Union[int, list[int]]] = None,
         options: Optional[dict] = None,
     ):
-        obs, info = self._env.reset(seed=seed, options=options)
+        obs, info = self._env.reset(seed=seed, options=options)  # type: ignore
         if options is not None and "env_idx" in options:
             env_idx = options["env_idx"]
-            mask = torch.zeros(self.num_envs, dtype=bool, device=self.base_env.device)
+            mask = torch.zeros(
+                self.num_envs, dtype=torch.bool, device=self.base_env.device
+            )
             mask[env_idx] = True
             if self.record_metrics:
                 self.success_once[mask] = False
@@ -106,11 +108,11 @@ class ManiSkillVectorEnv(VectorEnv):
                 self.returns[:] = 0
         return obs, info
 
-    def step(
-        self, actions: Union[Array, Dict]
-    ) -> Tuple[Array, Array, Array, Array, Dict]:
+    def step(  # pyright: ignore[reportIncompatibleMethodOverride]
+        self, actions: Union[Array, dict]
+    ) -> Tuple[Array, Array, Array, Array, dict]:
         obs, rew, terminations, truncations, infos = self._env.step(actions)
-
+        episode_info: Optional[dict] = None
         if self.record_metrics:
             episode_info = dict()
             self.returns += rew
@@ -131,7 +133,7 @@ class ManiSkillVectorEnv(VectorEnv):
 
         if self.ignore_terminations:
             terminations[:] = False
-            if self.record_metrics:
+            if episode_info:
                 if "success" in infos:
                     episode_info["success_at_end"] = infos["success"].clone()
                 if "fail" in infos:
@@ -139,7 +141,9 @@ class ManiSkillVectorEnv(VectorEnv):
         if self.record_metrics:
             infos["episode"] = episode_info
 
-        dones = torch.logical_or(terminations, truncations)
+        dones = torch.logical_or(
+            terminations, truncations  # pyright: ignore[reportArgumentType]
+        )
 
         if dones.any() and self.auto_reset:
             final_obs = torch_clone_dict(obs)
@@ -155,7 +159,13 @@ class ManiSkillVectorEnv(VectorEnv):
             infos["_final_observation"] = dones
             infos["_elapsed_steps"] = dones
             # NOTE (stao): Unlike gymnasium, the code here does not add masks for every key in the info object.
-        return obs, rew, terminations, truncations, infos
+        return (
+            obs,
+            rew,
+            terminations,
+            truncations,
+            infos,
+        )  # pyright: ignore[reportReturnType]
 
     def close_extras(self, **kwargs):
         self._env.close()

mani_skill/vector/wrappers/sb3.py

@@ -1,5 +1,5 @@
 import time
-from typing import Any, List, Optional, Type, Union
+from typing import Any, Optional, Type, Union
 
 import gymnasium as gym
 import numpy as np
@@ -52,7 +52,7 @@ class ManiSkillSB3VectorEnv(SB3VecEnv):
     def base_env(self) -> BaseEnv:
         return self._env.unwrapped
 
-    def seed(self, seed: Optional[int] = None) -> List[Union[None, int]]:
+    def seed(self, seed: Optional[int] = None) -> list[Union[None, int]]:
         self._last_seed = seed
 
     def reset(self) -> VecEnvObs:
@@ -113,7 +113,7 @@ class ManiSkillSB3VectorEnv(SB3VecEnv):
     def close(self) -> None:
         return self._env.close()
 
-    def get_attr(self, attr_name: str, indices: VecEnvIndices = None) -> List[Any]:
+    def get_attr(self, attr_name: str, indices: VecEnvIndices = None) -> list[Any]:
         return self._env.get_attr(attr_name, indices)
 
     def set_attr(
@@ -127,12 +127,12 @@ class ManiSkillSB3VectorEnv(SB3VecEnv):
         *method_args,
         indices: VecEnvIndices = None,
         **method_kwargs
-    ) -> List[Any]:
+    ) -> list[Any]:
         return self._env.env_method(
             method_name, *method_args, indices=indices, **method_kwargs
         )
 
     def env_is_wrapped(
         self, wrapper_class: Type[gym.Wrapper], indices: VecEnvIndices = None
-    ) -> List[bool]:
+    ) -> list[bool]:
         return [False] * self.num_envs

{mani_skill_nightly-2025.10.22.143.dist-info → mani_skill_nightly-2025.10.22.157.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mani_skill-nightly
-Version: 2025.10.22.143
+Version: 2025.10.22.157
 Summary: ManiSkill3: A Unified Benchmark for Generalizable Manipulation Skills
 Home-page: https://github.com/haosulab/ManiSkill
 Author: ManiSkill contributors