[CI] Resolve DMC and mujoco pinned versions

.github/unittest/linux/scripts/environment.yml

@@ -24,8 +24,8 @@ dependencies:
     - tensorboard
     - imageio==2.26.0
     - wandb
-    - dm_control<1.0.21
-    - mujoco<3.2.1
+    - dm_control
+    - mujoco
     - mlflow
     - av
     - coverage

.github/unittest/linux/scripts/run_all.sh

@@ -91,7 +91,7 @@ echo "installing gymnasium"
 pip3 install "gymnasium"
 pip3 install ale_py
 pip3 install mo-gymnasium[mujoco]  # requires here bc needs mujoco-py
-pip3 install "mujoco<3.2.1" -U
+pip3 install "mujoco" -U
 
 # sanity check: remove?
 python3 -c """

.github/unittest/linux_distributed/scripts/environment.yml

@@ -23,8 +23,8 @@ dependencies:
     - tensorboard
     - imageio==2.26.0
     - wandb
-    - dm_control<1.0.21
-    - mujoco<3.2.1
+    - dm_control
+    - mujoco
     - mlflow
     - av
     - coverage

.github/unittest/linux_examples/scripts/environment.yml

@@ -21,8 +21,8 @@ dependencies:
     - scipy
     - hydra-core
     - imageio==2.26.0
-    - dm_control<1.0.21
-    - mujoco<3.2.1
+    - dm_control
+    - mujoco
     - mlflow
     - av
     - coverage

.github/unittest/linux_libs/scripts_envpool/environment.yml

@@ -18,6 +18,6 @@ dependencies:
     - expecttest
     - pyyaml
     - scipy
-    - dm_control<1.0.21
-    - mujoco<3.2.1
+    - dm_control
+    - mujoco
     - coverage

.github/unittest/linux_olddeps/scripts_gym_0_13/environment.yml

@@ -22,7 +22,7 @@ dependencies:
     - scipy
     - hydra-core
     - dm_control -e git+https://github.com/deepmind/dm_control.git@c053360edea6170acfd9c8f65446703307d9d352#egg={dm_control}
-    - mujoco<3.2.1
+    - mujoco
     - patchelf
     - pyopengl==3.1.4
     - ray

.github/workflows/benchmarks.yml

@@ -35,7 +35,7 @@ jobs:
           python3 setup.py develop
           python3 -m pip install pytest pytest-benchmark
           python3 -m pip install "gym[accept-rom-license,atari]"
-          python3 -m pip install "dm_control<1.0.21" "mujoco<3.2.1"
+          python3 -m pip install "dm_control" "mujoco"
           export TD_GET_DEFAULTS_TO_NONE=1
       - name: Run benchmarks
         run: |
@@ -97,7 +97,7 @@ jobs:
           python3 setup.py develop
           python3 -m pip install pytest pytest-benchmark
           python3 -m pip install "gym[accept-rom-license,atari]"
-          python3 -m pip install "dm_control<1.0.21" "mujoco<3.2.1"
+          python3 -m pip install "dm_control" "mujoco"
           export TD_GET_DEFAULTS_TO_NONE=1
       - name: check GPU presence
         run: |

.github/workflows/benchmarks_pr.yml

@@ -34,7 +34,7 @@ jobs:
           python3 setup.py develop
           python3 -m pip install pytest pytest-benchmark
           python3 -m pip install "gym[accept-rom-license,atari]"
-          python3 -m pip install "dm_control<1.0.21" "mujoco<3.2.1"
+          python3 -m pip install "dm_control" "mujoco"
           export TD_GET_DEFAULTS_TO_NONE=1
       - name: Setup benchmarks
         run: |
@@ -108,7 +108,7 @@ jobs:
           python3 setup.py develop
           python3 -m pip install pytest pytest-benchmark
           python3 -m pip install "gym[accept-rom-license,atari]"
-          python3 -m pip install "dm_control<1.0.21" "mujoco<3.2.1"
+          python3 -m pip install "dm_control" "mujoco"
           export TD_GET_DEFAULTS_TO_NONE=1
       - name: check GPU presence
         run: |

docs/requirements.txt

@@ -14,8 +14,8 @@ docutils
 sphinx_design
 
 torchvision
-dm_control<1.0.21
-mujoco<3.2.1
+dm_control
+mujoco
 atari-py
 ale-py
 gym[classic_control,accept-rom-license]

sota-implementations/cql/cql_offline.py

@@ -58,14 +58,14 @@ def main(cfg: "DictConfig"):  # noqa: F821
             device = "cpu"
     device = torch.device(device)
 
+    # Create replay buffer
+    replay_buffer = make_offline_replay_buffer(cfg.replay_buffer)
+
     # Create env
     train_env, eval_env = make_environment(
         cfg, train_num_envs=1, eval_num_envs=cfg.logger.eval_envs, logger=logger
     )
 
-    # Create replay buffer
-    replay_buffer = make_offline_replay_buffer(cfg.replay_buffer)
-
     # Create agent
     model = make_cql_model(cfg, train_env, eval_env, device)
     del train_env
@@ -107,9 +107,6 @@ def main(cfg: "DictConfig"):  # noqa: F821
 
         q_loss = q_loss + cql_loss
 
-        alpha_loss = loss_vals["loss_alpha"]
-        alpha_prime_loss = loss_vals["loss_alpha_prime"]
-
         # update model
         alpha_loss = loss_vals["loss_alpha"]
         alpha_prime_loss = loss_vals["loss_alpha_prime"]

test/test_transforms.py

@@ -10423,17 +10423,18 @@ def test_transform_no_env(self, batch):
     reason="EndOfLifeTransform can only be tested when Gym is present.",
 )
 class TestEndOfLife(TransformBase):
+    pytest.mark.filterwarnings("ignore:The base_env is not a gym env")
+
     def test_trans_parallel_env_check(self, maybe_fork_ParallelEnv):
         def make():
             with set_gym_backend("gymnasium"):
                 return GymEnv(BREAKOUT_VERSIONED())
 
-        with pytest.warns(UserWarning, match="The base_env is not a gym env"):
-            with pytest.raises(AttributeError):
-                env = TransformedEnv(
-                    maybe_fork_ParallelEnv(2, make), transform=EndOfLifeTransform()
-                )
-                check_env_specs(env)
+        with pytest.raises(AttributeError):
+            env = TransformedEnv(
+                maybe_fork_ParallelEnv(2, make), transform=EndOfLifeTransform()
+            )
+            check_env_specs(env)
 
     def test_trans_serial_env_check(self):
         def make():