Upload 2 files

lerobot yaml files

lerobot_configs/act_koch_real.yaml

@@ -0,0 +1,102 @@
+# @package _global_
+
+# Use `act_koch_real.yaml` to train on real-world datasets collected on Alexander Koch's robots.
+# Compared to `act.yaml`, it contains 2 cameras (i.e. laptop, phone) instead of 1 camera (i.e. top).
+# Also, `training.eval_freq` is set to -1. This config is used to evaluate checkpoints at a certain frequency of training steps.
+# When it is set to -1, it deactivates evaluation. This is because real-world evaluation is done through our `control_robot.py` script.
+# Look at the documentation in header of `control_robot.py` for more information on how to collect data , train and evaluate a policy.
+#
+# Example of usage for training:
+# ```bash
+# python lerobot/scripts/train.py \
+#   policy=act_koch_real \
+#   env=koch_real
+# ```
+
+seed: 1000
+dataset_repo_id: lerobot/koch_pick_place_lego
+
+override_dataset_stats:
+  observation.images.laptop:
+    # stats from imagenet, since we use a pretrained vision model
+    mean: [[[0.485]], [[0.456]], [[0.406]]]  # (c,1,1)
+    std: [[[0.229]], [[0.224]], [[0.225]]]  # (c,1,1)
+  observation.images.logitech:
+    # stats from imagenet, since we use a pretrained vision model
+    mean: [[[0.485]], [[0.456]], [[0.406]]]  # (c,1,1)
+    std: [[[0.229]], [[0.224]], [[0.225]]]  # (c,1,1)
+
+training:
+  offline_steps: 80000
+  online_steps: 0
+  eval_freq: -1
+  save_freq: 10000
+  log_freq: 100
+  save_checkpoint: true
+
+  batch_size: 8
+  lr: 1e-5
+  lr_backbone: 1e-5
+  weight_decay: 1e-4
+  grad_clip_norm: 10
+  online_steps_between_rollouts: 1
+
+  delta_timestamps:
+    action: "[i / ${fps} for i in range(${policy.chunk_size})]"
+
+eval:
+  n_episodes: 50
+  batch_size: 50
+
+# See `configuration_act.py` for more details.
+policy:
+  name: act
+
+  # Input / output structure.
+  n_obs_steps: 1
+  chunk_size: 100
+  n_action_steps: 100
+
+  input_shapes:
+    # TODO(rcadene, alexander-soare): add variables for height and width from the dataset/env?
+    observation.images.laptop: [3, 480, 640]
+    observation.images.logitech: [3, 480, 640]
+    observation.state: ["${env.state_dim}"]
+  output_shapes:
+    action: ["${env.action_dim}"]
+
+  # Normalization / Unnormalization
+  input_normalization_modes:
+    observation.images.laptop: mean_std
+    observation.images.logitech: mean_std
+    observation.state: mean_std
+  output_normalization_modes:
+    action: mean_std
+
+  # Architecture.
+  # Vision backbone.
+  vision_backbone: resnet18
+  pretrained_backbone_weights: ResNet18_Weights.IMAGENET1K_V1
+  replace_final_stride_with_dilation: false
+  # Transformer layers.
+  pre_norm: false
+  dim_model: 512
+  n_heads: 8
+  dim_feedforward: 3200
+  feedforward_activation: relu
+  n_encoder_layers: 4
+  # Note: Although the original ACT implementation has 7 for `n_decoder_layers`, there is a bug in the code
+  # that means only the first layer is used. Here we match the original implementation by setting this to 1.
+  # See this issue https://github.com/tonyzhaozh/act/issues/25#issue-2258740521.
+  n_decoder_layers: 1
+  # VAE.
+  use_vae: true
+  latent_dim: 32
+  n_vae_encoder_layers: 4
+
+  # Inference.
+  temporal_ensemble_momentum: null
+
+  # Training and loss computation.
+  dropout: 0.1
+  kl_weight: 10.0

lerobot_configs/koch.yaml

@@ -0,0 +1,46 @@
+_target_: lerobot.common.robot_devices.robots.koch.KochRobot
+calibration_path: .cache/calibration/koch.pkl
+leader_arms:
+  main:
+    _target_: lerobot.common.robot_devices.motors.dynamixel.DynamixelMotorsBus
+    port: /dev/tty.usbmodem585A0085151
+    motors:
+      # name: (index, model)
+      shoulder_pan: [1, "xl330-m077"]
+      shoulder_lift: [2, "xl330-m077"]
+      elbow_flex: [3, "xl330-m077"]
+      wrist_flex: [4, "xl330-m077"]
+      wrist_roll: [5, "xl330-m077"]
+      gripper: [6, "xl330-m077"]
+follower_arms:
+  main:
+    _target_: lerobot.common.robot_devices.motors.dynamixel.DynamixelMotorsBus
+    port: /dev/tty.usbmodem585A0081771
+    motors:
+      # name: (index, model)
+      shoulder_pan: [1, "xl430-w250"]
+      shoulder_lift: [2, "xl430-w250"]
+      elbow_flex: [3, "xl330-m288"]
+      wrist_flex: [4, "xl330-m288"]
+      wrist_roll: [5, "xl330-m288"]
+      gripper: [6, "xl330-m288"]
+cameras:
+  logitech:
+    _target_: lerobot.common.robot_devices.cameras.opencv.OpenCVCamera
+    camera_index: 0
+    fps: 30
+    width: 640
+    height: 480
+  laptop:
+    _target_: lerobot.common.robot_devices.cameras.opencv.OpenCVCamera
+    camera_index: 3
+    fps: 30
+    width: 640
+    height: 480
+# `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
+# Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
+# the number of motors in your follower arms.
+max_relative_target: null
+# Sets the leader arm in torque mode with the gripper motor set to this angle. This makes it possible
+# to squeeze the gripper and have it spring back to an open position on its own.
+gripper_open_degree: 35.156