Downloads · 30 days
17
49% of all-time downloads
RyanL22/pi05-openarm-rh56f1-ablation-speedd-realsynth-20k
pi05-openarm-rh56f1-ablation-speedd-realsynth-20k is a robotics model from RyanL22. 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.
LeRobot-native pi05 (v0.6.1) fine-tuned on real teleop v4 (261 ep, 20 Hz) + 324 synthetic RH56F1 episodes from the EE-speed ablation: Dnorm2x — the Anorm time map played 2x faster. 8 synthetic tasks (airfryer2, ball,…
Downloads · 30 days
17
49% of all-time downloads
All-time downloads
35
Public
Parameters
4.1B
9.4 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors9.4 GB · 100%
How the weights are stored.
BF163.6B · 87%
From the Hugging Face model README
LeRobot-native pi05 (v0.6.1) fine-tuned on real teleop v4 (261 ep, 20 Hz) + 324 synthetic RH56F1 episodes from the EE-speed ablation: D_norm2x — the A_norm time map played 2x faster.
8 synthetic tasks (airfryer_2, ball, bottle, bottle_over_shelf, bottle_pour_2, box, coffee_pot, doll); 4 real + 8 synthetic = 12 cells, shares ∝ sqrt(frames).
Synthetic actions: depth IDM + HaWoR wrist IK on the speed-warped videos.
Comparison points (same recipe): RyanL22/pi05-openarm-rh56f1-ablation-noskel-realsynth-20k, ...-firstkf-realsynth-20k, and the full pipeline RyanL22/pi05-anyh2r-rh56f1-0916-wristik-30k.
Checkpoint: step 20,000 (final), NAVER MLXP 4x H200, batch 16 x 4 = 64.
| value | |
|---|---|
| vision encoder (SigLIP) | fine-tuned (not frozen) |
| image augmentation | photometric + affine, one draw replayed across the stereo pair |
| mirror augmentation | off |
| optimizer | AdamW, peak lr 2.5e-5, cosine decay, warmup 1000 |
| chunk | 50 actions @ 20 fps (2.5 s) |
| normalization | quantiles; constant dims (head) widened to mean ± 0.1 rad |
Inputs: base_0_rgb <- left ZED 288x512, left_wrist_0_rgb <- right ZED; state/action 28 dims neck(2) | left_arm(7) | right_arm(7) | left_hand(6) | right_hand(6).
Rollout: feed the state neck as 0.889 / 0.001.
from lerobot.policies.pi05.modeling_pi05 import PI05Policy
policy = PI05Policy.from_pretrained("RyanL22/pi05-openarm-rh56f1-ablation-speedd-realsynth-20k")