Downloads · 30 days
4
19% of all-time downloads
airvla_ftd_15000
hanapasta/airvla_ftd_15000
airvla_ftd_15000 is a robotics model from hanapasta. Use it for the robotics task on the model card, and read the license before you ship it in a product. The card lists the license as mit.
Grounding-scale fine-tune: tests whether scaling paired-command data further improves language grounding. It does not - wrong-object approaches increase despite lower validation error.
All-time downloads
21
Public
Parameters
4B
8.9 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors8.9 GB · 100%
Parameter types
How the weights are stored.
BF163.6B · 90%
Model card
From the Hugging Face model README
AirVLA FT-D (pi0, step 15000)
Grounding-scale fine-tune: tests whether scaling paired-command data further improves language grounding. It does not - wrong-object approaches increase despite lower validation error.
Part of the AirVLA MSc dissertation campaign: adapting a pretrained pi0 vision-language-action policy to a quadcopter with a 2-DoF arm and parallel gripper, evaluated in MuJoCo 3.3.4.
| Property | Value |
|---|---|
| Base model | lerobot/pi0 |
| Training data | hanapasta/airvla_v22 (v2 + F1 + F2 paired-command scaling) |
| Initialised from | FT-C |
| Training steps | 15,000 |
| Checkpoint selection | lowest pinned-noise validation MSE over the saved ladder |
Frozen-protocol result (n = 60 pick + 20 navigation, paired scenes)
1/60 grasped - 40/60 correct-target - 204.9 mm median (110.7 mm target-true) - 10/20 navigation.
Usage
Evaluated with the frozen harness eval_v2.py from the code repository:
python eval_v2.py <this_checkpoint> 60 20 --torchseed 1000 --tag run --video
Code, reproduction guide and the full experimental ledger: https://github.com/robotics-hana/drone-version2
Evaluation uses MuJoCo 3.3.4; a different simulator version changes contact behaviour enough to invalidate comparison with the banked results.
- Task
- Robotics
- License
- mit
- Created
- Sep 18, 2026
- Updated
- Sep 18, 2026