Add model card and reproduction pointer
Browse files
README.md
CHANGED
|
@@ -1,3 +1,28 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
+
tags:
|
| 4 |
+
- robotics
|
| 5 |
+
- world-model
|
| 6 |
+
- recursive-search
|
| 7 |
+
- libero
|
| 8 |
---
|
| 9 |
+
|
| 10 |
+
# Recursive-VAM Task-3 models
|
| 11 |
+
|
| 12 |
+
Frozen model bundle for continuing the sparse-milestone and dynamic recursive
|
| 13 |
+
depth experiments in
|
| 14 |
+
[JunhaoWu7/Recursive-VAM](https://github.com/JunhaoWu7/Recursive-VAM).
|
| 15 |
+
|
| 16 |
+
Contents:
|
| 17 |
+
|
| 18 |
+
- `libero_goal_step5000_model/`: resolved transformer, text encoder, tokenizer,
|
| 19 |
+
and Wan VAE used by the frozen same-root D1/D2/D3 experiment;
|
| 20 |
+
- `action_semantic_goal_object_v3_seed42/`: frozen sparse cumulative milestone
|
| 21 |
+
head and its training summary.
|
| 22 |
+
|
| 23 |
+
The authoritative protocol, hashes, data paths, and interpretation rules are
|
| 24 |
+
in `docs/TASK3_HANDOFF_20260727.md` and
|
| 25 |
+
`experiments/task3_artifact_manifest.json` in the GitHub repository.
|
| 26 |
+
|
| 27 |
+
This repository contains model artifacts only. Matching preencoded data is in
|
| 28 |
+
the private dataset repository `Wjjjh/Recursive-VAM-data`.
|