Update ManiSkill configuration and replay buffer to support truncation and dataset handling

- Reduced image size in ManiSkill environment configuration from 128 to 64
- Added support for truncation in replay buffer and actor server
- Updated SAC policy configuration to use a specific dataset and modify vision encoder settings
- Improved dataset conversion process with progress tracking and task naming
- Added flexibility for joint action space masking in learner server

lerobot/configs/env/maniskill_example.yaml

@@ -5,16 +5,20 @@ fps: 20
 env:
   name: maniskill/pushcube
   task:  PushCube-v1
-  image_size: 128
+  image_size: 64
   control_mode: pd_ee_delta_pose
   state_dim: 25
   action_dim: 7
   fps: ${fps}
   obs: rgb
   render_mode: rgb_array
-  render_size: 128
+  render_size: 64
   device: cuda
 
   reward_classifier:
     pretrained_path: null
-    config_path: null
+    config_path: null
+
+  wrapper:
+    joint_masking_action_space: null
+    delta_action: null

lerobot/configs/policy/sac_maniskill.yaml

@@ -8,7 +8,8 @@
 #   env.gym.obs_type=environment_state_agent_pos \
 
 seed: 1
-dataset_repo_id: null
+# dataset_repo_id: null
+dataset_repo_id: "AdilZtn/Maniskill-Pushcube-demonstration-medium"
 
 training:
   # Offline training dataloader
@@ -52,12 +53,14 @@ policy:
   n_action_steps: 1
 
   shared_encoder: true
-  vision_encoder_name: "helper2424/resnet10"
-  freeze_vision_encoder: true
+  # vision_encoder_name: "helper2424/resnet10"
+  vision_encoder_name: null
+  # freeze_vision_encoder: true
+  freeze_vision_encoder: false
   input_shapes:
     # # TODO(rcadene, alexander-soare): add variables for height and width from the dataset/env?
     observation.state: ["${env.state_dim}"]
-    observation.image: [3, 128, 128]
+    observation.image: [3, 64, 64]
   output_shapes:
     action: [7]