Downloads · 30 days
4
5% of all-time downloads
rangers-intrinsic/hf_act_SC_only_connector_3cam_delta6_state16
hf_act_SC_only_connector_3cam_delta6_state16 is a robotics model from rangers-intrinsic. Use it for the robotics task on the model card, and read the license before you ship it in a product. It is set up for lerobot. The card lists the license as apache-2.0.
This policy is an ACT model trained for SC connector insertion using 3 RGB cameras and a compact robot/task state.
Downloads · 30 days
4
5% of all-time downloads
All-time downloads
86
Public
Parameters
51.6M
207 MB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors207 MB · 100%
From the Hugging Face model README
This policy is an ACT model trained for SC connector insertion using 3 RGB cameras and a compact robot/task state.
At each control step, the model receives:
observation.images.left_cameraobservation.images.center_cameraobservation.images.right_cameraEach image has shape:
3 x 1024 x 1152These views provide visual information about:
observation.state has 16 dimensions:
tcp_pose.position.xtcp_pose.position.ytcp_pose.position.ztcp_pose.orientation.xtcp_pose.orientation.ytcp_pose.orientation.ztcp_pose.orientation.wtask.target_validtask.cable_type_idtask.plug_type_idtask.port_type_idtask.target_module_idtask.target_port_idtask.target_module_indextask.target_port_indextask.time_limitThis state provides:
The model predicts an action vector of 6 dimensions:
delta.position.xdelta.position.ydelta.position.zdelta.rotation.rxdelta.rotation.rydelta.rotation.rzThis is a delta pose policy, not an absolute pose policy.
delta.position.* are Cartesian translation incrementsdelta.rotation.* is a rotation-vector incrementAt inference time, the predicted action is applied relative to the previous commanded pose (or current pose on the first step) to produce the next Cartesian target.
In other words, the model learns to output a correction:
instead of directly predicting a final absolute pose.
Delta actions are often better suited for fine manipulation and insertion tasks because they:
This is especially useful for connector insertion, where small visual errors should produce small corrective motions rather than large jumps to an absolute pose.
Action Chunking with Transformers (ACT) is an imitation-learning method that predicts short action chunks instead of single steps. It learns from teleoperated data and often achieves high success rates.
This policy has been trained and pushed to the Hub using LeRobot. See the full documentation at LeRobot Docs.
For a complete walkthrough, see the training guide. Below is the short version on how to train and run inference/eval:
lerobot-train \
--dataset.repo_id=${HF_USER}/<dataset> \
--policy.type=act \
--output_dir=outputs/train/<desired_policy_repo_id> \
--job_name=lerobot_training \
--policy.device=cuda \
--policy.repo_id=${HF_USER}/<desired_policy_repo_id>
--wandb.enable=true
Writes checkpoints to outputs/train/<desired_policy_repo_id>/checkpoints/.
lerobot-record \
--robot.type=so100_follower \
--dataset.repo_id=<hf_user>/eval_<dataset> \
--policy.path=<hf_user>/<desired_policy_repo_id> \
--episodes=10
Prefix the dataset repo with eval_ and supply --policy.path pointing to a local or hub checkpoint.