- Added lerobot/scripts/server/gym_manipulator.py that contains all the necessary wrappers to run a gym-style env around the real robot.

- Added `lerobot/scripts/server/find_joint_limits.py` to test the min and max angles of the motion you wish the robot to explore during RL training.
- Added logic in `manipulator.py` to limit the maximum possible joint angles to allow motion within a predefined joint position range. The limits are specified in the yaml config for each robot. Checkout the so100.yaml.

Co-authored-by: Adil Zouitine <adilzouitinegm@gmail.com>

lerobot/configs/policy/hilserl_classifier.yaml

@@ -4,7 +4,7 @@ defaults:
   - _self_
 
 seed: 13
-dataset_repo_id: aractingi/pick_place_lego_cube_1
+dataset_repo_id: aractingi/push_green_cube_hf_cropped_resized
 train_split_proportion: 0.8
 
 # Required by logger
@@ -24,7 +24,8 @@ training:
   eval_freq: 1  # How often to run validation (in epochs)
   save_freq: 1  # How often to save checkpoints (in epochs)
   save_checkpoint: true
-  image_keys: ["observation.images.top", "observation.images.wrist"]
+  # image_keys: ["observation.images.top", "observation.images.wrist"]
+  image_keys: ["observation.images.laptop", "observation.images.phone"]
   label_key: "next.reward"
 
 eval:
@@ -32,7 +33,7 @@ eval:
   num_samples_to_log: 30  # Number of validation samples to log in the table
 
 policy:
-  name: "hilserl/classifier/pick_place_lego_cube_1"
+  name: "hilserl/classifier/push_green_cube_hf_cropped_resized" #"hilserl/classifier/pick_place_lego_cube_1"
   model_name: "facebook/convnext-base-224"
   model_type: "cnn"
   num_cameras: 2 # Has to be len(training.image_keys)