yimmmmmy/musubi-tuner-wanmove-adapt
musubi-tuner Wan-Move adaptation 给另一台机器 Codex 的交接说明(请先阅读) 1. 这个项目在做什么 本项目是在 musubi-tuner 上实现的 Wan-Move 风格 LIBERO 域适配: 基座只使用官方 Ruihang/Wan-Move-14B-480P; 输入为首帧、文本任务描述,以及夹爪/被操作物体或部件的二维点轨迹; 点轨迹被映射到 VAE latent 网格,并将首帧对应特征沿未来轨迹传播; 只训练 DiT 的 LoRA,目标是让模型更适应 LIBERO 的固定第三视角、 Franka 机械臂、物体几何和交互运动; 数据统一为 16 FPS、81 帧、约 5 秒,训练分辨率为 624x624; 原始训练集 5,843 条;默认按任务分层抽取 40%,在缓存前额外生成 2,337 条 时序一致的增强副本,因此正式训练清单共 8,180 条。650 条验证数据保持 原始、完全不增强,覆盖 130 个任务。 轨迹是点,不是… See the full description on the dataset page: https://huggingface.co/datasets/yimmmmmy/musubi-tuner-wanmove-adapt.
| Author / organization | yimmmmmy |
|---|---|
| License (as declared) | other |
| Downloads | 161 (as reported by Hugging Face) |
| Likes | 2 |
| Created | 21 August 2026 |
| Last updated at source | 22 August 2026 |
Popularity, last 2 days
161 now (+16 since 2026-10-19). Counts as reported by the source, snapshotted daily by Silytics.
More from yimmmmmy
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 20 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.