Add pipeline tag, links to paper and code
#1
by nielsr HF Staff - opened
README.md
CHANGED
|
@@ -1,19 +1,23 @@
|
|
| 1 |
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
library_name: pytorch
|
|
|
|
|
|
|
| 4 |
tags:
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
---
|
| 10 |
|
| 11 |
# Anchored Path Matching weights
|
| 12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
Publication status: **uploading / incomplete**.
|
| 14 |
|
| 15 |
-
This repository stages target/EMA inference weights for *Distilling World
|
| 16 |
-
Action Models for Efficient Bimanual Manipulation via Anchored Path Matching*.
|
| 17 |
Formal APM matches the frozen teacher and student **instantaneous direction
|
| 18 |
fields** at the same detached student-reached state and time. It also uses a
|
| 19 |
restart endpoint anchor and anchor-priority gradient projection.
|
|
@@ -64,5 +68,4 @@ python tools/materialize.py \
|
|
| 64 |
|
| 65 |
Only target/EMA inference tensors are included. Online weights, optimizer and
|
| 66 |
scheduler states, trainer recovery state, datasets, simulator assets, and the
|
| 67 |
-
LingBot-VA teacher are not redistributed.
|
| 68 |
-
|
|
|
|
| 1 |
---
|
|
|
|
| 2 |
library_name: pytorch
|
| 3 |
+
license: apache-2.0
|
| 4 |
+
pipeline_tag: robotics
|
| 5 |
tags:
|
| 6 |
+
- robotics
|
| 7 |
+
- world-action-model
|
| 8 |
+
- model-distillation
|
| 9 |
+
- anchored-path-matching
|
| 10 |
---
|
| 11 |
|
| 12 |
# Anchored Path Matching weights
|
| 13 |
|
| 14 |
+
This repository stages target/EMA inference weights for [*Distilling World Action Models for Efficient Bimanual Manipulation via Anchored Path Matching*](https://huggingface.co/papers/2611.22592).
|
| 15 |
+
|
| 16 |
+
Project page: https://jialonglong.github.io/Anchored-Path-Matching/
|
| 17 |
+
Code: https://github.com/JIAlonglong/Anchored-Path-Matching
|
| 18 |
+
|
| 19 |
Publication status: **uploading / incomplete**.
|
| 20 |
|
|
|
|
|
|
|
| 21 |
Formal APM matches the frozen teacher and student **instantaneous direction
|
| 22 |
fields** at the same detached student-reached state and time. It also uses a
|
| 23 |
restart endpoint anchor and anchor-priority gradient projection.
|
|
|
|
| 68 |
|
| 69 |
Only target/EMA inference tensors are included. Online weights, optimizer and
|
| 70 |
scheduler states, trainer recovery state, datasets, simulator assets, and the
|
| 71 |
+
LingBot-VA teacher are not redistributed.
|
|
|