---
title: hands-on-modern-rl-experiment12-ai2thor-embodied
canonical_url: "https://www.modelscope.cn/studios/walkinglab/hands-on-modern-rl-experiment12-ai2thor-embodied"
md_url: "https://www.modelscope.cn/studios/walkinglab/hands-on-modern-rl-experiment12-ai2thor-embodied.md"
repository: walkinglab/hands-on-modern-rl-experiment12-ai2thor-embodied
chinese_name: "Experiment 12 · AI2-THOR xGPU Embodied Home"
last_updated: 2026-08-20
sdk_type: gradio
sdk_version: 6.17.3
downloads: 0
stars: 0
---

# hands-on-modern-rl-experiment12-ai2thor-embodied

> hands-on-modern-rl-experiment12-ai2thor-embodied - walkinglab 在 ModelScope 创建的在线 Demo。WalkingLab × Hands-On Modern RL: visual ObjectNav training in interactive AI2-THOR rooms with learned-policy replay.

walkinglab/hands-on-modern-rl-experiment12-ai2thor-embodied 是 ModelScope 魔搭社区上的在线可交互 Demo（创空间），基于 gradio 6.17.3 构建，中文名为「Experiment 12 · AI2-THOR xGPU Embodied Home」。

- **Repository**: walkinglab/hands-on-modern-rl-experiment12-ai2thor-embodied
- **SDK**: gradio
- **SDK version**: 6.17.3
- **Downloads**: 0
- **Stars**: 0
- **Last updated**: 2026-08-20

Source: https://www.modelscope.cn/studios/walkinglab/hands-on-modern-rl-experiment12-ai2thor-embodied

---

# WalkingLab × Hands-On Modern RL · AI2-THOR xGPU Embodied Home

**WalkingLab** 与开源课程 **Hands-On Modern RL（《动手学现代强化学习》）** 的实验 12。它在 AI2-THOR 的真实可交互房间中训练视觉 ObjectNav PPO，并从学习后的策略生成第一人称 GIF 回放。

- Project: <https://github.com/walkinglabs/hands-on-modern-rl>
- WalkingLab: <https://modelscope.cn/organization/walkinglab>
- Companion chapter: <https://walkinglabs.github.io/hands-on-modern-rl>
- Live Studio: <https://modelscope.cn/studios/walkinglab/hands-on-modern-rl-experiment12-ai2thor-embodied>
- Training source: <https://modelscope.cn/studios/walkinglab/hands-on-modern-rl-experiment12-ai2thor-embodied/file/view/master/space_runtime.py>
- Companion Notebook: <https://modelscope.cn/notebook/share/github/walkinglabs/hands-on-modern-rl/blob/main/code/online-experiments/hands-on-modern-rl-experiment12-ai2thor-embodied.ipynb> (requires a scheduled xGPU Notebook)