Skip to content

Latest commit

 

History

History
18 lines (13 loc) · 1.36 KB

README.md

File metadata and controls

18 lines (13 loc) · 1.36 KB
Raw

Eval Scripts

We provide some example evaluation scrips that will allow you to run our policies on ALOHA and DROID robots.

  • eval_aloha.py will deploy our policies on an ALOHA robot assuming the default 14-DoF joint-state action space. It provides an implementation for both temporal ensembling and receding horizon control with chunked action predictions.
  • eval_droid.py will deploy our policies on a DROID robot using the default cartesian velocity action space. Predicted actions are directly executed on the robot (no test time smoothing).
  • eval_droid_state.py will deploy our policies on a DROID robot using the cartesian position action space. Note that the rotation actions are predicted using a R6 representation (conversion code here) following Chi et. al.. After prediction the actions are further smoothed with temporal ensembling. This action space is ideal for diffusion policy (U-Net) action heads.

Setup Instructions

Just download the policy folder (produced by finetune.py) and add a file named obs_config.yaml to it. This will tell the eval script how to process the observations for the policy. An example is provided below:

img: '26638268_left'
transform:
  _target_: data4robotics.transforms.get_transform_by_name
  name: preproc