metadata
library_name: lerobot
tags:
- molmoact2
- robotics
- lerobot
- vla
base_model: allenai/MolmoAct2-SO100_101
molmoact2_block_stack_20k
Fine-tuned MolmoAct2 (action-expert-only) for block_stack on SO101 data.
| Policy | MolmoAct2 (policy.type=molmoact2) |
| Init checkpoint | allenai/MolmoAct2-SO100_101 |
| Dataset | villekuosmanen/armnetbench_block_stack |
| Task | block_stack |
| Action dim | 6 (single-arm) |
| Cameras | top, wrist, front |
| Training | 20k steps, batch 32 (8/GPU × 4 DDP), 4× RTX 5090 on Vast.ai |
| Prior HF repo | pravsels/molmoact2_block_stack |
| W&B project | molmoact2_block_stack |
| W&B run | tb8n7j94 |
Checkpoints
The final checkpoint (local step 020000, 20k total steps) lives at the repository root for direct loading.
Usage
from lerobot.policies.molmoact2.modeling_molmoact2 import MolmoAct2Policy
policy = MolmoAct2Policy.from_pretrained("pravsels/molmoact2_block_stack_20k")