README.md

---
license: mit
tags:
- robotics
- libero
- oat
- dense-visual-memory
---
# OAT Dense LIBERO-10 — Checkpoint Epoch 300

Hugging Face model repository for a **dense cross-attention OAT policy** trained on
**LIBERO-10 (N500)**. This snapshot was taken at **epoch 300** during a long run
(`oat_dense_with_uid_long_0530_220204`).

## Files

| File | Description |
|------|-------------|
| `ep-0300.ckpt` | PyTorch workspace checkpoint (~729 MB) |
| `training_logs.jsonl` | Full training JSONL (train/val curves) |
| `training_metrics_dashboard.png` | Training loss dashboard |
| `overfit_watcher/` | Counterfactual early-stop reports |
| `sim_eval/` | Phase A screen eval (30 ep/task) |
| `sim_eval_phase_b/` | Phase B confirm eval (50 ep/task, 3 exp) — ep-0700 only |
| `experiment_log_dense_visual_memory.md` | Experiment journal |

## Metrics @ epoch 300

| Train loss | Val loss | Reconst MSE | Sim SR (Phase A) |
|-----------:|---------:|------------:|-----------------:|
| 2.4420902729034424 | 5.05718469619751 | 0.07346421480178833 | **39.7%** |

### Training (offline)

![Training metrics dashboard](training_metrics_dashboard.png)

### Sim eval (LIBERO-10, Phase A screen)

**Mean success rate: 39.7%** — 30 episodes/task, `300` total rollouts, seed 1000.
Details: [`sim_eval/eval_summary.md`](sim_eval/eval_summary.md) · [`sim_eval/eval_log.json`](sim_eval/eval_log.json)

![Sim eval dashboard](sim_eval/sim_eval_dashboard.png)


## Model configuration (summary)

- **Policy:** OAT with `use_dense_visual_memory=true` (spatial visual tokens + cross-attn)
- **State memory:** enabled (`use_state_memory_tokens=true`)
- **Task UID:** enabled in state tokens
- **Dataset:** `libero10_N500.zarr`
- **Embed dim:** 256

## Baseline reference

Paper **OAT8** on LIBERO-10: **~56.3%** mean success rate (external reference).

## Citation

If you use this checkpoint, please cite **OAT: Ordered Action Tokenization** and specify
epoch **300** of the dense LIBERO-10 ladder.