smolvla_busybox_buttons_25k
Fine-tuned SmolVLA on SO101 joint-space data.
| Base model | lerobot/smolvla_base |
| Dataset | pravsels/busybox_buttons_bimanual |
| Checkpoint step | 25,000 |
| Training | Isambard GH200 4-GPU DDP, batch 32 global (8/GPU), 30k steps from lerobot/smolvla_base; published step 025000 |
| W&B project | smolvla_busybox_buttons |
| W&B run | w53mzumw |
Checkpoints
Checkpoint at local step 025000 lives at the repository root for direct loading.
Includes CHECKPOINT_MANIFEST.json (SHA-256).
Usage
from lerobot.policies.smolvla.modeling_smolvla import SmolVLAPolicy
policy = SmolVLAPolicy.from_pretrained("pravsels/smolvla_busybox_buttons_25k")