makepluscode
/

ch08-01-train-smolvla-smolvla-2000-step

Model card Files Files and versions

ch08-01-train-smolvla-smolvla-2000-step / README.md

makepluscode's picture

Upload smolvla-2000-step (2000 steps)

d48dfb8 verified about 20 hours ago

|

history blame contribute delete

1.05 kB

	---
	license: apache-2.0
	tags:
	- lerobot
	- smolvla
	- vla-robotics
	- so-arm101
	library_name: lerobot
	pipeline_tag: robotics
	---

	# smolvla-2000-step

	VLA 로보틱스 입문서 예제용 SmolVLA 파인튜닝 체크포인트입니다.

	## 경로 (GitHub `code/` 와 대응)

	\| 항목 \| 값 \|
	\|------\|-----\|
	\| 논리 경로 \| `vla-robotics/ch08/01-train-smolvla/smolvla-2000-step` \|
	\| GitHub \| [vla-robotics-examples](https://github.com/makepluscode/vla-robotics-examples) \|
	\| 예제 폴더 \| `code/ch08/01-train-smolvla/` \|
	\| Hub 모델 ID \| `makepluscode/ch08-01-train-smolvla-smolvla-2000-step` \|
	\| 학습 step \| 2000 \|
	\| 데이터셋 \| `makepluscode/so_arm101_block_picking_main` \|
	\| 베이스 \| `lerobot/smolvla_base` \|

	## 로드

	```python
	from lerobot.policies.smolvla.modeling_smolvla import SmolVLAPolicy

	policy = SmolVLAPolicy.from_pretrained("makepluscode/ch08-01-train-smolvla-smolvla-2000-step")
	```

	전·후처리는 동일 폴더의 `policy_preprocessor.json` /
	`policy_postprocessor.json` 을 `from_pretrained` 로 함께 로드합니다.