unsloth-grpo
Unsloth-grpo enables training of reasoning models using Group Relative Policy Optimization (GRPO). This technique replaces traditional PPO Reward and Value models with group statistics, achieving 8x memory savings and allowing long-context RL training on limited VRAM.
npx skills add majiayu000/claude-skill-registry --skill unsloth-grpo-cuba6112-skillfactory-c0f5d1f3 --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
--- name: unsloth-grpo description: Implementation of Group Relative Policy Optimization (GRPO) for training reasoning models, optimized for 8x memory savings (triggers: GRPO, reasoning, DeepSeek-R1, reinforcement learning, RLVR, GRPOTrainer, thinking tokens). --- ## Overview Unsloth-grpo enables training of reasoning models using Group Relative Policy Optimization (GRPO). This technique replaces traditional PPO Reward and Value models with group statistics, achieving 8x memory savings and allowing long-context RL training on limited VRAM. ## When to Use - When building DeepSeek-R1 style reasoning models. - When performing Reinforcement Learning with Verifiable Rewards (RLVR) for math or code. - When training models with long context lengths (e.g., 20K tokens) on single GPUs. ## Decision Tree 1. Is your model size < 1.5B? - Yes: Model may struggle with consistent thinking tokens; consider 1.5B-8B. 2. Is the reward verifiable (e.g., math answer)? - Yes: Use RLVR with regex-based reward functions. 3. Are you training on a single GPU with long context? - Yes: Use `GRPOTrainer` to benefit from the 8x memory reduction. ## Workflows 1. **Converting to Reasoning LLM**: Load a base model w
- Overview
- When to Use
- Decision Tree
- Workflows
- Non-Obvious Insights
- Evidence
- Scripts
- Dependencies
- References
What does the unsloth-grpo skill do?
Unsloth-grpo enables training of reasoning models using Group Relative Policy Optimization (GRPO). This technique replaces traditional PPO Reward and Value models with group statistics, achieving 8x memory savings and allowing long-context RL training on limited VRAM.
How do I install it?
Run `npx skills add majiayu000/claude-skill-registry --skill unsloth-grpo-cuba6112-skillfactory-c0f5d1f3 --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From majiayu000/claude-skill-registry, a repository with 534 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.