arkojit1/pi05_franka_haply_joint_delta

π₀.₅ — joint-delta control on franka_haply_joint_delta

π₀.₅ fine-tuned on Haply-teleoperated Franka demonstrations, predicting joint-space deltas. Checkpoint taken at the step with the lowest eval loss.

Base checkpoint	lerobot/pi05_base
Dataset	Ameyapores/franka_haply_joint_delta — 94 episodes / 50,861 frames @ 20 fps, 1 language task
Step	3,500 of a 5,000-step run (19.6 epochs)
Eval loss	0.1977 (flow-matching objective, 10% held-out episodes)
Trainable	action expert only (--train_expert_only: SigLIP and Gemma-2B frozen, ~300M trainable)

Action space

8-dimensional, from the dataset's action column:

dims	meaning
0–6	per-joint angle deltas (7 DoF)
7	gripper

Observations are three 224×224 RGB cameras (observation.images.base_0_rgb, observation.images.base_1_rgb, observation.images.left_wrist_0_rgb) plus an 8-dim observation.state. empty_cameras=0 — the three cameras exactly fill π₀.₅'s three image slots, with no masked padding.

chunk_size=50 and n_action_steps=50 are pi05_base's own defaults. At 20 fps that is 2.5 s of predicted motion, executed open-loop in full before re-planning. If you want tighter closed-loop control, lower n_action_steps at inference; it does not affect the weights.

Usage

from lerobot.policies.pi05.modeling_pi05 import PI05Policy

policy = PI05Policy.from_pretrained("arkojit1/pi05_franka_haply_joint_delta")

Requires access to the gated google/paligemma-3b-pt-224 repo — the preprocessor loads its tokenizer by id. Accept the license there first and make sure HF_TOKEN is set.

Training

Global batch 256 across 8×MI300X (RCCL DDP, 32 per GPU), learning rate 2.5e-5 with a cosine schedule, bf16 with gradient checkpointing, quantile normalisation of state and action, image augmentation on the train split only.

eval_loss is the training objective evaluated on held-out episodes — not a success rate.