Offline checkpoint screening and replanning diagnostics for robot imitation learning.
-
Updated
Oct 29, 2026 - HTML
Offline checkpoint screening and replanning diagnostics for robot imitation learning.
让模型针对本地lora训练集状况自动编写适当的lora训练配置
Evaluation harness that caught a 7M-step PPO checkpoint outperforming the 10M-step final, traced to approx_kl divergence. Multi-seed evaluation and physics perturbation testing.
To associate your repository with the checkpoint-selection topic, visit your repo's landing page and select "manage topics."