Skill rating

How prime-rl vendors, builds, and ships CUDA kernels (the `deps/prime-kernels` submodule and the `prime-kernels` wheel). Use when adding a kernel, building it locally, calling one from training code,

Find, start, use, and stop the local run dashboard for metrics, configs, traces, logs, and reports. Use when asked for its URL, to watch or inspect a run, to control the open dashboard, or to create a

Launch and monitor prime-rl evals — the `uv run eval` entrypoint, its config and CLI shorthands, run directory, resume, logs and metrics. Use when asked to evaluate a model or checkpoint on an environ

How the prime-rl config system works — TOML files, CLI overrides, composition, and special patterns. Use when creating configs, debugging config errors, or overriding values via CLI.

How to prepare and publish GitHub releases for prime-rl. Use when drafting release notes, tagging versions, or publishing releases.

Launch local or SLURM prime-rl training runs with the rl, sft, and inference entrypoints, including dry runs. Model Factory uses model-factory; standalone evals use eval.

Launch and monitor prime-rl runs on Model Factory with prime train, and retrieve their traces. Use for requests such as "run this on model factory", including custom code revisions, private datasets,

Launch and monitor prime-rl training runs. Use when starting, supervising, or debugging an RL/SFT run. Routes Model Factory to model-factory, and local or SLURM runs to start-run and monitor-run.

Monitor a local or SLURM prime-rl training run — find the output directory, tail logs, check key metrics, inspect jobs, and restart safely. Model Factory monitoring uses model-factory.

How to install prime-rl and its optional dependencies. Use when setting up the project, installing extras like DeepEP for multi-node expert parallelism, or troubleshooting dependency issues.