---
name: primeintellect-ai-training
description: "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."
license: "Apache-2.0"
metadata:
  decimalai-skill-id: "83fa62f3-4762-4afd-aef0-663699e70a71"
  decimalai-version-id: "1611d41b-281d-4f91-930a-a84abf10163d"
  decimalai-version: "3"
  decimalai-body-sha256: "57eb9bd3dda917a0c108b1c9b19ad03d3ed3dee901130182716511b226150db4"
  decimalai-name: "primeintellect-ai/training"
  decimalai-format: "1"
  decimalai-invocation: "model"
  decimalai-category: "ai-ml"
  decimalai-source: "https://app.decimal.ai/s/primeintellect-ai-training@3/SKILL.md"
---
# Training

Choose the deployment workflow first.

- **Model Factory** — see [model-factory](../model-factory/SKILL.md) for `prime train` launches, custom source revisions, credentials, private datasets, and remote monitoring. Requests such as "run this on model factory" use this path unless the user explicitly requests SLURM.
- **Start a local / SLURM 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 local / SLURM 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.

The local start/monitor subskills assume the `configs` skill for config-loading mechanics. Model Factory dispatch takes a TOML filename directly and has its own CLI flags. Standalone evals (`uv run eval`) are the `eval` skill; online evals follow the selected training workflow.