Instructions to use RoboColosseum/DROID-models with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- LeRobot
How to use RoboColosseum/DROID-models with LeRobot:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
Upload lingbot-vla-v2/closelaptop/README.md with huggingface_hub
Browse files
lingbot-vla-v2/closelaptop/README.md
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# LingBot-VLA v2 6B — DROID CloseLaptop
|
| 2 |
+
|
| 3 |
+
- **Task:** "close laptop" — 100 episodes, 27,649 frames @ 30 fps (`aarontung/droid-close-laptop`)
|
| 4 |
+
- **Inputs/outputs:** cameras left exterior / right exterior / wrist + instruction; 8-D absolute joint state and action (7 Franka joints + gripper)
|
| 5 |
+
- **Base:** `robbyant/lingbot-vla-v2-6b`
|
| 6 |
+
- **Trained:** Full post-train (recipe `configs/vla/real_robot/real_robot.yaml`)
|
| 7 |
+
- **Optim:** Muon, constant LR 1.25e-5 (same as BimanualYAM); GBS 32 (16/GPU x 2); 5 epochs; fp32 master weights, bf16 compute; 2x H200 NVL; wall time 2h20m
|
| 8 |
+
- **Files:** Final checkpoint in HF format plus `norm_stats.json` (meanstd).
|
| 9 |
+
- **Load:** use as `--model_path` with the lingbot-vla-v2 inference code, together with `--norm_stats_file norm_stats.json`
|
| 10 |
+
- **Code:** RoboColosseum harness (`scripts/train/lingbot.sh`); logs in W&B project `RoboColosseum`
|