xvla_panda_pick_v1

X-VLA fine-tuned on Rakeshdharan56/pick_robosuite (revision v1): Franka Panda, "Hand me the {bottle|vial|syringe}", LIBERO-style I/O (2 cams 256², 8-D state, 7-D delta-EEF action at 20 Hz).

Init lerobot/xvla-base, action_mode=auto (7-D actions padded to 20 for training, trimmed at inference), full fine-tune, 7,500 steps, batch 32, bf16, RTX 6000 Ada. Camera keys: agentview → observation.images.image, eye_in_hand → observation.images.image2. Chunk 30.

Load with LeRobot 0.5.2 (commit b06ad4088): get_policy_class(cfg.type).from_pretrained("Rakeshdharan56/xvla_panda_pick_v1") and make_pre_post_processors(policy.config, pretrained_path="Rakeshdharan56/xvla_panda_pick_v1"). Closed-loop robosuite eval: pending.