From cc6a2cac432c2a04de1c46e1f3291c10a698a142 Mon Sep 17 00:00:00 2001 From: Nikodem Bartnik Date: Wed, 20 May 2026 10:55:18 +0200 Subject: [PATCH] update policy deployment instruction with rollout --- src/lerobot/templates/lerobot_modelcard_template.md | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-) diff --git a/src/lerobot/templates/lerobot_modelcard_template.md b/src/lerobot/templates/lerobot_modelcard_template.md index b93e83b6e..4fd442684 100644 --- a/src/lerobot/templates/lerobot_modelcard_template.md +++ b/src/lerobot/templates/lerobot_modelcard_template.md @@ -73,14 +73,16 @@ _Writes checkpoints to `outputs/train//checkpoints/`._ ### Evaluate the policy/run inference ```bash -lerobot-record \ - --robot.type=so100_follower \ - --dataset.repo_id=/eval_ \ +lerobot-rollout \ + --strategy.type=base \ + --robot.type=so101_follower \ + --robot.cameras="{ up: {type: opencv, index_or_path: /dev/video1, width: 640, height: 480, fps: 30}, side: {type: opencv, index_or_path: /dev/video5, width: 640, height: 480, fps: 30}}" \ --policy.path=/ \ - --episodes=10 + --task="Put lego brick into the transparent box" \ + --duration=60 ``` -Prefix the dataset repo with **eval\_** and supply `--policy.path` pointing to a local or hub checkpoint. +If you want to record a dataset while testing the policy use `--dataset.repo_id=/eval_dataset_name` it is important to use the prefix **eval_**. For the policy path use the policy from the Hugging Face Hub or a local one. Skipping duration will make the policy run indefinitely. ---