Upload OpenPI checkpoint step 5000 (model card)
Browse files
README.md
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: openpi
|
| 3 |
+
tags: ["openpi", "robotics", "vla"]
|
| 4 |
+
---
|
| 5 |
+
|
| 6 |
+
# KeWangRobotics/soarm-pi05-no-state
|
| 7 |
+
|
| 8 |
+
OpenPI checkpoint uploaded with `scripts/upload_model_to_huggingface.py`.
|
| 9 |
+
|
| 10 |
+
## Checkpoint Info
|
| 11 |
+
|
| 12 |
+
- Format: `orbax`
|
| 13 |
+
- Config name: `soarm_pi05_flow`
|
| 14 |
+
- Experiment name: `so_arm`
|
| 15 |
+
- Step: `5000`
|
| 16 |
+
- Contains assets: `True`
|
| 17 |
+
- Contains train_state: `True`
|
| 18 |
+
|
| 19 |
+
## Expected Layout
|
| 20 |
+
|
| 21 |
+
- `params/`: Orbax/JAX inference weights used by OpenPI.
|
| 22 |
+
- `assets/`: normalization stats and other checkpoint assets.
|
| 23 |
+
|
| 24 |
+
## Loading
|
| 25 |
+
|
| 26 |
+
This repository stores checkpoint artifacts. Load it from OpenPI by pointing policy/config code at the downloaded checkpoint directory.
|