[HIL-SERL] Migrate threading to multiprocessing (#759)

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

lerobot/configs/env/maniskill_example.yaml

@@ -22,3 +22,9 @@ env:
   wrapper:
     joint_masking_action_space: null
     delta_action: null
+
+  video_record:
+    enabled: false
+    record_dir: maniskill_videos
+    trajectory_name: trajectory
+    fps: ${fps}

lerobot/configs/env/so100_real.yaml

@@ -28,4 +28,3 @@ env:
   reward_classifier:
     pretrained_path:  outputs/classifier/13-02-random-sample-resnet10-frozen/checkpoints/best/pretrained_model
     config_path: lerobot/configs/policy/hilserl_classifier.yaml
-

lerobot/configs/policy/sac_maniskill.yaml

@@ -8,14 +8,12 @@
 #   env.gym.obs_type=environment_state_agent_pos \
 
 seed: 1
-# dataset_repo_id: null
 dataset_repo_id: "AdilZtn/Maniskill-Pushcube-demonstration-medium"
 
 training:
   # Offline training dataloader
   num_workers: 4
 
-  # batch_size: 256
   batch_size: 512
   grad_clip_norm: 10.0
   lr: 3e-4
@@ -113,4 +111,7 @@ policy:
 actor_learner_config:
   learner_host: "127.0.0.1"
   learner_port: 50051
-  policy_parameters_push_frequency: 15
+  policy_parameters_push_frequency: 1
+  concurrency:
+    actor: 'processes'
+    learner: 'processes'