- Added JointMaskingActionSpace wrapper in gym_manipulator in order to select which joints will be controlled. For example, we can disable the gripper actions for some tasks.

- Added Nan detection mechanisms in the actor, learner and gym_manipulator for the case where we encounter nans in the loop.
- changed the non-blocking in the `.to(device)` functions to only work for the case of cuda because they were causing nans when running the policy on mps
- Added some joint clipping and limits in the env, robot and policy configs. TODO clean this part and make the limits in one config file only.

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

lerobot/configs/env/so100_real.yaml

@@ -18,10 +18,12 @@ env:
     control_time_s: 20
     reset_follower_pos: true
     use_relative_joint_positions: true
-    reset_time_s: 10
+    reset_time_s: 5
     display_cameras: false
     delta_action: 0.1
+    joint_masking_action_space: [1, 1, 1, 1, 0, 0] # disable wrist and gripper
 
   reward_classifier:
     pretrained_path: outputs/classifier/checkpoints/best/pretrained_model
     config_path: lerobot/configs/policy/hilserl_classifier.yaml
+    

lerobot/configs/policy/sac_real.yaml

@@ -20,7 +20,7 @@ training:
   lr: 3e-4
 
   eval_freq: 2500
-  log_freq: 500
+  log_freq: 1
   save_freq: 2000000
 
   online_steps: 1000000
@@ -31,7 +31,7 @@ training:
   online_env_seed: 10000
   online_buffer_capacity: 1000000
   online_buffer_seed_size: 0
-  online_step_before_learning: 100 #5000
+  online_step_before_learning: 1000 #5000
   do_online_rollout_async: false
   policy_update_freq: 1
 
@@ -76,8 +76,10 @@ policy:
       mean: [0.485, 0.456, 0.406]
       std: [0.229, 0.224, 0.225]
     observation.state:
-      min: [-88.50586,  23.81836, 0.87890625, -32.16797, 78.66211,   0.53691274]
-      max: [84.55078, 187.11914, 145.98633, 101.60156, 146.60156,  88.18792]
+      min: [-87.09961,     62.402344,    67.23633,     36.035156,    77.34375,0.53691274] 
+      max: [58.183594,   131.83594,    145.98633,     82.08984,     78.22266, 0.60402685]
+      # min: [-88.50586,  23.81836, 0.87890625, -32.16797, 78.66211,   0.53691274]
+      # max: [84.55078, 187.11914, 145.98633, 101.60156, 146.60156,  88.18792]
 
   output_normalization_modes:
     action: min_max

lerobot/configs/robot/so100.yaml

@@ -15,8 +15,13 @@ calibration_dir: .cache/calibration/so100
 # the number of motors in your follower arms.
 max_relative_target: null
 joint_position_relative_bounds: 
-  min: [-88.50586,  23.81836, 0.87890625, -32.16797, 78.66211,   0.53691274]
-  max: [84.55078, 187.11914, 145.98633, 101.60156, 146.60156,  88.18792]
+   min: [-87.09961,     62.402344,    67.23633,     36.035156,    77.34375,
+   0.53691274] 
+   max: [58.183594,   131.83594,    145.98633,     82.08984,     78.22266,
+   0.60402685]
+   
+  # min: [-88.50586,  23.81836, 0.87890625, -32.16797, 78.66211,   0.53691274]
+  # max: [84.55078, 187.11914, 145.98633, 101.60156, 146.60156,  88.18792]
 
 leader_arms:
   main: