Add model card
Browse files
README.md
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: lerobot
|
| 3 |
+
pipeline_tag: robotics
|
| 4 |
+
tags:
|
| 5 |
+
- smolvla
|
| 6 |
+
- libero
|
| 7 |
+
- robotics
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
# SmolVLA 路 libero_90 路 100k
|
| 11 |
+
|
| 12 |
+
Action-labeled behavior-cloning pretrain of [lerobot/smolvla_base](https://huggingface.co/lerobot/smolvla_base)
|
| 13 |
+
on `nvidia/libero_90` (LeRobot v3), 100k steps.
|
| 14 |
+
|
| 15 |
+
Use as `--policy.path` for few-shot LIBERO-Goal fine-tunes in this project.
|
| 16 |
+
|
| 17 |
+
```bash
|
| 18 |
+
# example
|
| 19 |
+
lerobot-train --policy.path=<this-repo> ...
|
| 20 |
+
```
|
| 21 |
+
|
| 22 |
+
Vision encoder frozen; action expert + state_proj trained. No LIBERO-Goal demos in this checkpoint.
|