Downloads · 30 days
3
21% of all-time downloads
grasp-anything-10501
charlesH777/grasp-anything-10501
grasp-anything-10501 is a image-text-to-text model from charlesH777. Use it for the image-text-to-text task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
基于 NVIDIA LocateAnything-3B 的语言引导二维抓取模型,multigt + Vision LoRA 阶段产物。
All-time downloads
14
Public
Parameters
3.6B
7.8 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors7.8 GB · 100%
At a glance
- Task
- Image-Text-to-Text
- License
- apache-2.0
- Model type
- locateanything
- Access
- Public
- Created
- Aug 1, 2026
- Updated
- Aug 1, 2026
- SHA
- 4764c801
Try a prompt
Model card
From the Hugging Face model README
Grasp Anything 2D — Checkpoint 10501 (grasp_contact 专用)
基于 NVIDIA LocateAnything-3B 的语言引导二维抓取模型,multigt + Vision LoRA 阶段产物。
任务
grasp_contact — 输出平行夹爪两个接触点 (x1, y1, x2, y2),四坐标均为 [0, 1000] 离散 token。
评测结果(RealVLG 官方 mini633)
| Split | n | gAcc_corrected_strict | mIoU_strict | center_err_median (px) |
|---|---|---|---|---|
| seen | 243 | 66.67% | 57.52% | 10.33 |
| similar | 226 | 50.00% | 49.03% | 16.55 |
| novel | 164 | 28.05% | 26.86% | 56.87 |
- 评测协议:
evaluate_realvlg_contact.py,fast 模式 - contact 专用 checkpoint 中性能最强
使用方法
from locate_anything_service.model import LocateAnythingRuntime
from locate_anything_service.config import Settings
settings = Settings(model_id="charlesH777/grasp-anything-10501")
runtime = LocateAnythingRuntime(settings)
result = runtime.predict("image.jpg", "抓取红色杯子", mode="grasp_contact")
训练配置
- 阶段: multigt + Vision LoRA (MoonViT LoRA rank 8, last 4 layers)
- LLM LoRA: rank 32
- 训练量: seen_contact_blocks = 168,016
- Task
- Image-Text-to-Text
- Type
- locateanything
- License
- apache-2.0
- Languages
- en, zh
- Created
- Aug 1, 2026
- Updated
- Aug 1, 2026