Sign inGet started
Dataset · Hugging Face

ar-mine/e5-c3-evaluation-deployment

E5+C3 RoboTwin evaluation 部署与逐步验收手册 本包供另一台计算机上的 Agent 从干净工作目录部署评估环境,并通过真实渲染、专家规划、固定输入推理和两回合完整闭环验证。默认平台为 Linux x86_64 + 单张 NVIDIA L40S:同一台机器、同一张卡运行仿真渲染和策略推理,两套Python环境及独立进程通过 localhost 通信。不需要云端主机、SSH隧道或分布式部署;其他NVIDIA GPU按实际计算能力编译。包内是2026年10月21日已使用的评估代码快照及迁移脚本。接收机器尚未实测;本包的交付验证不代表目标机运行验收。 默认验收模型是 E5+C3 低epoch版 25723372.pbs101 / step17180,FP32参数/主路径,动作chunk50、去噪10、三路RGB及14维关节接口。此模型用于检查部署一致性,不代表它在低桌任务上最优,也不是正在评估的三组step13000后训练模型。 先将 AGENT_TASK.md 交给目标机器Agent。所有命令从本包根目录或下文明确指定的… See the full description on the dataset page: https://huggingface.co/datasets/ar-mine/e5-c3-evaluation-deployment.

Author / organizationar-mine
Downloads0 (as reported by Hugging Face)
Likes0
Created21 October 2026
Last updated at source21 October 2026
Embed this entry
<iframe src="https://silytics.si/embed/item/845c2271-1886-48fd-9422-04fa999cedab" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Free, no tracking. More about feeds and embeds.

More from ar-mine

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 21 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.