smolVLA ยท UR7e ยท PickandPlace (50 epoch, tp1)
lerobot/smolvla_base ๋ฅผ CoRL2026-CSI/UR7e_CaP_PickandPlace_100epi_10fps_state_tplus1_action ๋ฐ์ดํฐ์ ์ผ๋ก 50 epoch ํ์ธํ๋ํ SmolVLA ์ ์ฑ ๋ชจ๋ธ.
Model details
- Base model:
lerobot/smolvla_base(SmolVLM2-500M-Video-Instruct + action expert) - Robot: UR7e (7-DOF, gripper ํฌํจ)
- Cameras:
realsense_topview,realsense_wrist(480ร640 โ 256ร256 resize) - Action: 7D joint positions (6 joints + gripper)
- State variant:
state_tplus1_action(state at t+1, action at t)
Training
| Config | Value |
|---|---|
| Dataset | CoRL2026-CSI/UR7e_CaP_PickandPlace_100epi_10fps_state_tplus1_action (35878 frames, 100 episodes) |
| Steps | 7100 (= 50 epoch) |
| Global batch | 256 (BATCH=64 ร NUM_GPUS=4) |
| Optimizer | AdamW (lerobot smolvla preset) |
| Mixed precision | no (bf16 inference) |
| Image augmentation | brightness / contrast / saturation / hue / sharpness / affine, max 3 random |
| Hardware | 4ร H100 80GB |
ํ์ต ์คํฌ๋ฆฝํธ: scripts/ur7e_tplus1/train_smolvla_pickandplace.sh (CoRL2026 lerobot fork).
Camera rename
LeRobot dataset ์ ์นด๋ฉ๋ผ ํค์ SmolVLA ์ ์ฑ ํค ๋งคํ:
| Dataset key | Policy key |
|---|---|
observation.images.realsense_wrist |
observation.images.camera1 |
observation.images.realsense_topview |
observation.images.camera2 |
Usage
from lerobot.policies.smolvla.modeling_smolvla import SmolVLAPolicy
policy = SmolVLAPolicy.from_pretrained("CoRL2026-CSI/smolVLA_UR7e_PickandPlace_50epoch_tp1")
Citation / Acknowledgement
Built on top of LeRobot and the SmolVLA checkpoint. Project: CoRL 2026 CSI submission.
- Downloads last month
- 8
Model tree for Cache-SCA/smolVLA_UR7e_PickandPlace_50epoch_tp1
Base model
HuggingFaceTB/SmolLM2-360M Quantized
HuggingFaceTB/SmolLM2-360M-Instruct Quantized
HuggingFaceTB/SmolVLM-500M-Instruct Finetuned
lerobot/smolvla_base