refactor(tests): remove grpc import checks from test files for cleaner code

src/lerobot/rl/__init__.py

@@ -12,15 +12,12 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+"""Reinforcement learning modules.
+
+Distributed actor / learner entry points (``actor``, ``learner``,
+``learner_service``) require ``pip install 'lerobot[hilserl]'``. Algorithms,
+buffer, data sources and trainer are gRPC-free and usable standalone.
 """
-Reinforcement learning modules.
-
-Requires: ``pip install 'lerobot[hilserl]'``
-"""
-
-from lerobot.utils.import_utils import require_package
-
-require_package("grpcio", extra="hilserl", import_name="grpc")
 
 from .algorithms.base import RLAlgorithm as RLAlgorithm
 from .algorithms.configs import RLAlgorithmConfig as RLAlgorithmConfig, TrainingStats as TrainingStats
@@ -28,7 +25,7 @@ from .algorithms.factory import (
     make_algorithm as make_algorithm,
     make_algorithm_config as make_algorithm_config,
 )
-from .algorithms.sac import SACAlgorithm as SACAlgorithm, SACAlgorithmConfig as SACAlgorithmConfig
+from .algorithms.sac.configuration_sac import SACAlgorithmConfig as SACAlgorithmConfig
 from .buffer import ReplayBuffer as ReplayBuffer
 from .data_sources import DataMixer as DataMixer, OnlineOfflineMixer as OnlineOfflineMixer
 from .trainer import RLTrainer as RLTrainer
@@ -39,7 +36,6 @@ __all__ = [
     "TrainingStats",
     "make_algorithm",
     "make_algorithm_config",
-    "SACAlgorithm",
     "SACAlgorithmConfig",
     "RLTrainer",
     "ReplayBuffer",

tests/policies/test_gaussian_actor_policy.py

@@ -15,9 +15,6 @@
 # limitations under the License.
 
 import pytest
-
-pytest.importorskip("grpc")
-
 import torch
 from torch import Tensor, nn
 
@@ -240,7 +237,7 @@ def test_gaussian_actor_policy_forward(batch_size: int, state_dim: int, action_d
 
 
 @pytest.mark.parametrize("batch_size,state_dim,action_dim", [(2, 6, 6), (1, 10, 10)])
-def test_sac_training_through_algorithm(batch_size: int, state_dim: int, action_dim: int):
+def test_gaussian_actor_training_through_sac(batch_size: int, state_dim: int, action_dim: int):
     config = create_default_config(state_dim=state_dim, continuous_action_dim=action_dim)
     algorithm, policy = _make_algorithm(config)
 
@@ -270,7 +267,7 @@ def test_sac_training_through_algorithm(batch_size: int, state_dim: int, action_
 
 
 @pytest.mark.parametrize("batch_size,state_dim,action_dim", [(2, 6, 6), (1, 10, 10)])
-def test_sac_training_with_visual_input(batch_size: int, state_dim: int, action_dim: int):
+def test_gaussian_actor_training_with_visual_input(batch_size: int, state_dim: int, action_dim: int):
     config = create_config_with_visual_input(state_dim=state_dim, continuous_action_dim=action_dim)
     algorithm, policy = _make_algorithm(config)
 
@@ -331,7 +328,7 @@ def test_gaussian_actor_policy_with_pretrained_encoder(
     assert actor_loss.shape == ()
 
 
-def test_sac_training_with_shared_encoder():
+def test_gaussian_actor_training_with_shared_encoder():
     batch_size = 2
     action_dim = 10
     state_dim = 10
@@ -358,7 +355,7 @@ def test_sac_training_with_shared_encoder():
     algorithm.optimizers["actor"].step()
 
 
-def test_sac_training_with_discrete_critic():
+def test_gaussian_actor_training_with_discrete_critic():
     batch_size = 2
     continuous_action_dim = 9
     full_action_dim = continuous_action_dim + 1
@@ -475,7 +472,7 @@ def test_sac_algorithm_with_critics_number_of_heads(num_critics: int):
 
 def test_gaussian_actor_policy_save_and_load(tmp_path):
     """Test that the policy can be saved and loaded from pretrained."""
-    root = tmp_path / "test_sac_save_and_load"
+    root = tmp_path / "test_gaussian_actor_save_and_load"
 
     state_dim = 10
     action_dim = 10
@@ -506,7 +503,7 @@ def test_gaussian_actor_policy_save_and_load(tmp_path):
 
 def test_gaussian_actor_policy_save_and_load_with_discrete_critic(tmp_path):
     """Discrete critic should be saved/loaded as part of the policy."""
-    root = tmp_path / "test_sac_save_and_load_discrete"
+    root = tmp_path / "test_gaussian_actor_save_and_load_discrete"
 
     state_dim = 10
     action_dim = 6

tests/rl/test_data_mixer.py

@@ -13,10 +13,6 @@
 # limitations under the License.
 """Tests for RL data mixing (DataMixer, OnlineOfflineMixer)."""
 
-import pytest
-
-pytest.importorskip("grpc")
-
 import torch
 
 from lerobot.rl.buffer import ReplayBuffer

tests/rl/test_queue.py

@@ -18,10 +18,6 @@ import threading
 import time
 from queue import Queue
 
-import pytest
-
-pytest.importorskip("grpc")
-
 from torch.multiprocessing import Queue as TorchMPQueue
 
 from lerobot.rl.queue import get_last_item_from_queue

tests/rl/test_sac_algorithm.py

@@ -16,9 +16,6 @@
 """Tests for the RL algorithm abstraction and SACAlgorithm implementation."""
 
 import pytest
-
-pytest.importorskip("grpc")
-
 import torch
 
 from lerobot.configs.types import FeatureType, PolicyFeature
@@ -31,7 +28,7 @@ from lerobot.utils.constants import ACTION, OBS_IMAGE, OBS_STATE
 from lerobot.utils.random_utils import set_seed
 
 # ---------------------------------------------------------------------------
-# Helpers (reuse patterns from tests/policies/test_sac_policy.py)
+# Helpers (reuse patterns from tests/policies/test_gaussian_actor_policy.py)
 # ---------------------------------------------------------------------------
 
 

tests/rl/test_trainer.py

@@ -14,10 +14,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import pytest
-
-pytest.importorskip("grpc")
-
 import torch
 from torch import Tensor
 

tests/utils/test_process.py

@@ -22,8 +22,6 @@ from unittest.mock import patch
 
 import pytest
 
-pytest.importorskip("grpc")
-
 from lerobot.rl.process import ProcessSignalHandler
 
 

tests/utils/test_replay_buffer.py

@@ -19,7 +19,6 @@ from collections.abc import Callable
 
 import pytest
 
-pytest.importorskip("grpc")
 pytest.importorskip("datasets", reason="datasets is required (install lerobot[dataset])")
 
 import torch  # noqa: E402