You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

pi05 sample_loading baseline

pi0.5 fine-tuned checkpoint (full-weight) for the RoboSynChallenge sample_loading task.

Train config: pi05_base_robosynchallenge_sample_loading_full (openpi), step 19999
Model name: sample_loading_sim
Training data: official dataset.
Framework: openpi / pi0.5; checkpoint in openpi orbax format (params/ + assets/)

Eval protocol

Action chunk executed per query K=10, flow-matching denoise steps N=10 (model default). CPU physics, eval_reset_sync_steps=1, expert-feasibility check on, fixed v2 seed list (eval_seeds_v2.json), fast-rt renderer, 100 clear + 100 random episodes.

setting	success / 100
clear	4
random	2

Numbers were measured with the fast-rt renderer; the organizers' Hybrid renderer can differ.

Layout

ckpt/
  params/             orbax checkpoint
  assets/             openpi asset handlers (norm stats)
  _CHECKPOINT_METADATA

Run via submission/sample_loading/deploy_policy.yml in the RoboSynChallenge repo (bash submission/eval.sh sample_loading).

Downloads last month: -; Downloads are not tracked for this model. How to track

Video Preview

Robotics