Agent Skills › PrimeIntellect-ai/prime-rl › training

training

GitHub

用于启动、监控和调试 RL/SFT 训练任务。涵盖单节点/SLURM 启动及日志指标监控，支持模型评估与在线验证流程管理。

skills/training/SKILL.md PrimeIntellect-ai/prime-rl

Trigger Scenarios

启动强化学习或监督微调训练监控训练运行状态和指标调试训练或评估任务

Install

npx skills add PrimeIntellect-ai/prime-rl --skill training -g -y

More Options

Use without installing

npx skills use PrimeIntellect-ai/prime-rl@training

指定 Agent (Claude Code)

npx skills add PrimeIntellect-ai/prime-rl --skill training -a claude-code -g -y

安装 repo 全部 skill

npx skills add PrimeIntellect-ai/prime-rl --all -g -y

预览 repo 内 skill

npx skills add PrimeIntellect-ai/prime-rl --list

SKILL.md

Frontmatter

{
    "name": "training",
    "description": "Launch and monitor prime-rl training runs. Use when starting, supervising, or debugging an RL\/SFT run. Routes to `start-run` (entrypoints + how to launch) and `monitor-run` (logs, metrics, check-ins)."
}

Training

Two phases — start the run, then watch it.

Start a run — see start-run/SKILL.md for the rl, sft, and inference entrypoints and how to launch them (single-node, SLURM, dry-run).
Monitor a run — see monitor-run/SKILL.md for the runbook: how to find the output dir, what to tail, which metrics to watch, and how to restart safely.

Both subskills assume the configs skill for config-loading mechanics. Standalone evals (uv run eval) are the eval skill; online evals of a training run are configured in start-run and read in monitor-run.