dyk111zn/GRPO-LIBERO-Spatial-Reproduction
AI model by dyk111zn
Older snapshot: check the original source for current information.
From the publisher
Model card & documentation
Source previewRead the publisher’s intended use, setup instructions, evaluations and limitations. The original model card is the source of truth.
最新完整策略断点为 weights/policy/globalstep60/。第 60 步之后的训练请从这份断点恢复。 它包含完整模型权重、优化器状态及 8 个 DCP 分片；真实 LIBERO-Spatial 10 个任务 × 50 次评估成功率为 77.2%（386/500）。 latestpolicy.json 记录最新断点、评估结果和续训入口。40、45、50 步用于历史对比。 结果来源：77.2% 是原单节点八卡训练在第 60 步执行的真实仿真验证， 使用 runner.onlyeval=false、runner.valcheckinterval=5，并非本发布包的独立 portable.py eval 入口得到。原训练日志记录于 2026-10-25 13:18:07；完整 500 条 结果见 evidence/evaluations/globalstep60.json，入口及验证范围见 evidence/portabilityfix20261027.json。 下载并执行下文的 restore、check 后，完整的下一轮实验使用以下入口： 它依次执行：60 步策略采集 500 条真实双相机轨迹 → 更新 Wan/Cosmos → 打包本轮新模型 → 策略从 60 步继续到 80 步。完整配方见“后续世界模型更新”。 以下单条命令只续训 GRPO，不采集数据或更新世界模型；使用包内 Wan 1000/Cosmos 2000： --target 80 表示训练至总步数 80，即新增 20 步；恢复优化器时使用完整断点目录。 本发布包对应 π0.5 + Wan 主视角 + Cosmos-WoVR 腕部视角，在另一台 Linux 服务器上启动八卡 GRPO。训练使用世界模型生成的观测；成功率是在真实 LIBERO 仿真器中，以固定初始状态、10 个任务 × 50 次试验统计的 success-once。 DiffSynth、LIBERO 源码与仿真资源。 顺序，不使用包含重复版本的 pip freeze 输出来重建环境。 完整策略权重及 DCP 训练状态。globalstep45/ 只有完整策略权重，可用于评估 或模型初始化，不用于恢复优化器状态。 推理权重。它们的世界模型训练优化器状态不在本发布包中。 校验和，并更新本地文件指纹，不重新选择关键帧。 参考检查点的真实 Spatial 成功率：40 步 365/500（73.0%），45 步 370/500（74.0%），50 步 349/500（69.8%），最新 60 步 386/500（77.2%）。40 步由前一轮世界模型训练得到； 40→60 的本轮续训使用新 Wan 1000 步和新 Cosmos 2000 步。 二进制环境适用于 Linux x8664，参考系统为 Ubuntu 22.04 / glibc 2.35， 8 张 A800 80 GiB，驱动 595.45.04。另一种 CPU 架构、较旧 glibc、不同 CUDA 驱动或其他显卡需要额外验证。请预留发布包下载空间和约 45 GiB 的环境解压空间， 另为训练日志、视频与完整检查点预留至少 100 GiB。 系统依赖示例： 先用服务器已有的 Python 安装下载工具，下载公开仓库。下载凭据不属于发布包。 restore 默认验证所有发布文件的 SHA-256。Python 3.10 的完整开发头文件由…
Read the full model card ↗ · Preview checked 2026-10-29T23:36:26.142Z
Inside the original model card — Document outline
Headings are captured from the source. Links open the publisher’s document, not a locally hosted copy.
Links from the model card
References supplied by the publisher, not independently verified endorsements. Check the destination before downloading files or entering credentials.
Documentation belongs to its respective authors. A listing is not a grant of reuse or training rights. Confirm the document’s own terms at the source.
Model overview
dyk111zn/GRPO-LIBERO-Spatial-Reproduction is a task-unspecified model repository published by dyk111zn on Hugging Face. A library was not reported.
This page summarizes Hub metadata. For intended use, training data, evaluation results and limitations, consult the original model card.
Model facts
- Task
- Not reported
- Library
- Not reported
- Recent downloads (30 days)
- 0
- Cumulative likes
- 0
- Hugging Face trending score
- 0
- Reported safetensors parameters
- Not reported
- Architecture
- Not reported
- License
- Not reported — check the model card
- Access
- Not gated by Hugging Face
- Created
- 2026-10-25T06:38:00.000Z
- Last modified
- 2026-10-25T06:38:00.000Z
Compatibility and lineage
Language coverage was not reported.
Base-model lineage was not included.
Parameter count is not a RAM/VRAM requirement. Check precision, quantization, context length and runtime compatibility in the model card; no hardware or API-cost claim is inferred here.
Use and evaluate
Check license terms and access requirements first. Review the model card’s documented loading instructions, evaluations and safety limitations. Benchmark scores are not imported by this directory, and popularity is not an accuracy ranking.
No model weights are downloaded or executed by AltAPIs. Never enable remote model code without reviewing it.
Source and freshness
Source: Hugging Face Hub. Metadata observed 2026-10-25T06:46:38.979Z. Daily imports are snapshots, not real-time monitoring.
Popularity and source listings do not establish security, suitability, licensing rights or benchmark performance.