Orchestrate multi-stage curriculum training for robotic navigation tasks.
npx skills add https://github.com/mzqef/MotrixLab --skill curriculum-learning
Please help me install this Agent Skill. Skill: curriculum-learning Source: https://github.com/mzqef/MotrixLab/tree/main/.github/skills/curriculum-learning Command: npx skills add https://github.com/mzqef/MotrixLab --skill curriculum-learning
This Skill orchestrates multi-stage curriculum-based training for robotic navigation tasks, enabling progressive skill acquisition and efficient transfer of knowledge across stages.
Review the Curriculum Plan Schema and Stage Configuration Reference, then run the automation script to generate initial schedules or start training with the first stage.
High-intent search queries and answers about installing and using this skill.
Multi-stage curriculum learning for robot navigation is configured by defining stages with specific environments, rewards, promotion criteria, and warm-start transfers to guide progressive skill acquisition from simple to complex challenges.
Checkpoint transfer in curriculum training moves the best checkpoints between stages to accelerate learning and warm-start subsequent phases, ensuring knowledge acquired in simpler environments transfers efficiently to harder tasks.
Yes, curriculum learning supports per-stage configuration of promotion thresholds, reward overrides, and learning rate multipliers, allowing structured experimentation and customized progression through increasingly complex robotic navigation tasks.
Curriculum learning supports environment flexibility by mapping between environments and starter kits, enabling structured experimentation across progressively complex robotic navigation scenarios and seamless stage transitions.
The best way to start progressive skill acquisition is to review the Curriculum Plan Schema and Stage Configuration Reference, then run the automation script to generate initial schedules or begin training with the first stage.
castform-aiValidate Castform training environments by interpreting scorecards and reward variance.
niznik-devOrchestrate SLURM fine-tuning jobs with torchtune and evaluate with inspect-ai.
tachyon-beepDesign reward functions that preserve optimal policies in reinforcement learning.
InverterNetworkFine-tune large language models with TRL for SFT, DPO, and GRPO workflows.
