Microduck RL 4096x6000
Velocity-tracking reproduction pinned to an upstream commit: 4096 parallel environments, 6000 PPO iterations, with intermediate checkpoints, the ONNX export, training config, TensorBoard events and closed-loop replay video all kept, and a matching ball-kick run keeps the same record. Sim-only.
https://huggingface.co/Datawhale/Microduck-RL-4096x600061 OBSERVATIONS 14 ACTIONS 50 HZ COMPATIBILITY NOT INDEPENDENTLY VERIFIED
AWESOME MICRODUCK INDEX
Microduck RL 4096x6000
Velocity-tracking reproduction pinned to an upstream commit: 4096 parallel environments, 6000 PPO iterations, with intermediate checkpoints, the ONNX export, training config, TensorBoard events and closed-loop replay video all kept, and a matching ball-kick run keeps the same record. Sim-only.
Project Overview
--- license: other library name: rsl-rl pipeline tag: reinforcement-learning tags: - robotics - reinforcement-learning - ppo - mujoco - microduck - bipedal-locomotion - embodied-ai ---
这是 Every Embodied 为 Duck Together 共建计划完成的一次 Microduck 速度跟踪策略复现。训练基于 Pollen Robotics 的 microduck rl ，使用 4096 个并行环境完成 6000 次 PPO 迭代，并保留中间 checkpoint、最终 ONNX、训练配置、TensorBoard 事件文件和连续闭环回放视频。
项目 数值 --- ---: 上游提交 29e887ecfbf5d37144759e5a9f8a176dfb83d547 训练任务 Mjlab-Velocity-Flat-MicroDuck 并行环境 4096 PPO 迭代 6000 环境步数 589,824,000 训练耗时 01:27:41 最后一轮 mean reward 119.57 最后一轮 mean episode length 972.52 ONNX 输入/输出 [1, 61] - [1, 14]
Task
| Field | Value |
|---|---|
| Source | HUGGING FACE MODEL SOURCE |
| Runtime contract | 61 observations / 14 actions / 50 Hz |
| Verification | Pending independent Duckio review |
| Artifact | Not indexed |
Source Coverage
The upstream README includes: 关键结果, 文件结构, 下载, 复现与回放, ONNX 说明, 使用边界, 许可与来源.
Intended Use
Use this entry to assess the project and locate its canonical implementation. Read the complete README, inspect the manifest and artifacts, and reproduce the source's evaluation before assigning simulation or hardware verification.