remove some deprecated code and pin pytest <8

Pytest 8 breaks import order of pytorch and isaacgym, which has to happen a certain due to some singleton issues I don't understand.

gym/utils/task_registry.py

@@ -37,7 +37,6 @@
 from datetime import datetime
 from typing import Tuple
 
-from learning.env import VecEnv
 from learning.runners import *  # noqa: F403
 from learning.utils import set_discount_from_horizon
 
@@ -67,15 +66,15 @@ def __init__(self):
     def register(
         self,
         name: str,
-        task_class: VecEnv,
+        task_class,
         env_cfg: BaseConfig,
         train_cfg: LeggedRobotRunnerCfg,
     ):
         self.task_classes[name] = task_class
         self.env_cfgs[name] = env_cfg
         self.train_cfgs[name] = train_cfg
 
-    def get_task_class(self, name: str) -> VecEnv:
+    def get_task_class(self, name: str):
         return self.task_classes[name]
 
     def get_cfgs(self, name) -> Tuple[LeggedRobotCfg, LeggedRobotRunnerCfg]:
@@ -217,7 +216,7 @@ def make_sim(self):
             self.sim["params"],
         )
 
-    def make_env(self, name, env_cfg) -> VecEnv:
+    def make_env(self, name, env_cfg):
         if name in self.task_classes:
             task_class = self.get_task_class(name)
         else:

learning/runners/BaseRunner.py

@@ -1,12 +1,11 @@
 import torch
 from learning.algorithms import PPO
 from learning.modules import ActorCritic
-from learning.env import VecEnv
 from learning.utils import remove_zero_weighted_rewards
 
 
 class BaseRunner:
-    def __init__(self, env: VecEnv, train_cfg, device="cpu"):
+    def __init__(self, env, train_cfg, device="cpu"):
         self.device = device
         self.env = env
         self.parse_train_cfg(train_cfg)

learning/runners/my_runner.py

@@ -1,5 +1,4 @@
 import torch
-from learning.env import VecEnv
 
 from learning.utils import Logger
 from learning.utils import PotentialBasedRewardShaping
@@ -11,7 +10,7 @@
 
 
 class MyRunner(OnPolicyRunner):
-    def __init__(self, env: VecEnv, train_cfg, device="cpu"):
+    def __init__(self, env, train_cfg, device="cpu"):
         super().__init__(env, train_cfg, device)
         logger.initialize(
             self.env.num_envs,

learning/runners/on_policy_runner.py

@@ -1,6 +1,5 @@
 import os
 import torch
-from learning.env import VecEnv
 
 from learning.utils import Logger
 
@@ -10,7 +9,7 @@
 
 
 class OnPolicyRunner(BaseRunner):
-    def __init__(self, env: VecEnv, train_cfg, device="cpu"):
+    def __init__(self, env, train_cfg, device="cpu"):
         super().__init__(env, train_cfg, device)
         logger.initialize(
             self.env.num_envs,

requirements.txt

@@ -7,7 +7,7 @@ torch>=2.0
 torchvision
 torchaudio
 pygame
-pytest
+pytest<8.0
 mss
 ruff
 pre-commit