Dataset Viewer

The dataset could not be loaded because the splits use different data file formats, which is not supported. Read more about the splits configuration. Click for more details.

Couldn't infer the same data file format for all splits. Got {NamedSplit('train'): ('text', {}), NamedSplit('validation'): (None, {}), NamedSplit('test'): ('harbor', {})}

Error code:   FileFormatMismatchBetweenSplitsError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

so101-tray-sort offline evaluation — assets index (2026-10-03)

This repo holds everything needed to reproduce / finish the tray_sort cross-policy CI-MSE study without any of the original machines.

Assets (Hugging Face)

what	repo
datasets (trimmed): merged 73 / train 64 / val 7	`Kavin60606/so101-tray-sort-{merged,train,val}` and `FidelityDynamics/so101-tray-sort-{merged,train,val}`
raw recordings	`FidelityDynamics/so101-dataset` (kavin, sanskar), `angkul07/so101-tray-sort`
MolmoAct2 checkpoints (2 seeds × 500/1000/1500/2000, postprocessor patched)	`Kavin60606/so101-tray-sort-ma2-ckpts`
π0.5 LoRA adapters (2 seeds × 4) on `lerobot/pi05_base`	`Kavin60606/so101-tray-sort-pi05-ckpts`
offline results: CI-MSE variants, val loss, prediction dumps, VLM intervals	this repo: `ma2/`, `pi05/`
docs + scripts	this repo: `docs/`, `scripts/`

Status

Scored: MA2 seed 1000 (4), π0.5 seed 1000 (4), π0.5 seed 2000 (4). To redo: MA2 seed 2000 (4 ckpts) — see docs/REDO_offline_eval.md (≈45 min on a 16 GB GPU).

Key docs

docs/REDO_offline_eval.md — exact commands to finish scoring
docs/CHECKLIST_tray_sort.md — what was done right / wrong during implementation
docs/PLAN_tray_sort_crosspolicy.md, docs/PLAN_isaac05_finetune.md — plans
docs/ci_vlm.json (+raw, prompt TraySort.json, annotate_or.py) — VLM critical intervals for the 7 val episodes
docs/trim_bounds.csv — idle-trim bounds used to build the datasets

Scripts

scripts/offline/ — score_one.sh, dump_score.sh, dump_predictions_v2.py, val_loss.py, patch_ma2_post.py, box setup + train scripts
scripts/vast/ — dataset build/resplit, server (serve_multi.py), health/post chains
scripts/mac/ — robot rollout client (rollout.py, clamps, grading), analysis, record scripts

Downloads last month: 371

Total file size:

170 MB