skills.rest
zangjeicy
zangjeicyCOMMUNITY
Explore all skills by this author →
GitHub

slime-rl-training

Train LLMs with GRPO and PPO using Megatron-LM and SGLang.

Updated Jul 13, 2026
Search Tags:#llm#reinforcement learning#distributed training#sglang#grpo#post-training#megatron-lm
One-click install
npx skills add https://github.com/zangjeicy/Hermes --skill slime-rl-training-zangjeicy
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: slime-rl-training
Source: https://github.com/zangjeicy/Hermes/tree/main/optional-skills/mlops/slime
Command: npx skills add https://github.com/zangjeicy/Hermes --skill slime-rl-training-zangjeicy

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires sglang-router, ray, torch, transformers, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill addresses the complexity of scaling Reinforcement Learning (RL) for large language models by providing a unified framework that bridges Megatron-LM training with high-throughput SGLang rollout generation.

Core Features & Use Cases

  • Megatron-LM Integration: Enables native training with full parallelism (TP, PP, DP, SP) for massive models.
  • High-Throughput Rollout: Utilizes SGLang for efficient response generation and reward verification.
  • Use Case: Use this framework to train reasoning models like GLM-4 or Llama 3 using GRPO or PPO algorithms, specifically when you need to manage complex multi-turn agentic workflows or custom reward functions.

Quick Start

Execute the slime-rl-training skill to launch a GRPO training job using the qwen3-4B model configuration script.

Frequently Asked Questions about slime-rl-training

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I scale LLM post-training with Reinforcement Learning using Megatron-LM?▼

To scale LLM post-training with Reinforcement Learning using Megatron-LM, you can use a framework that bridges Megatron-LM distributed training with high-throughput SGLang rollout generation. This enables native training with full parallelism for massive models.

Can I train large language models with GRPO and PPO algorithms for multi-turn agentic workflows?▼

Yes, you can train large language models using GRPO and PPO algorithms for multi-turn agentic workflows. The framework supports managing complex agentic workflows and custom reward functions during large-scale model optimization.

What dependencies do I need to run distributed RL training with SGLang and Ray?▼

To run distributed RL training with SGLang and Ray, you need to integrate sglang-router, Ray, PyTorch, and Transformers. These dependencies manage rollout generation, distributed execution, and model configurations.

Does SGLang handle high-throughput rollout generation for LLM reinforcement learning?▼

Yes, SGLang handles high-throughput rollout generation for LLM reinforcement learning. It is utilized specifically for efficient response generation and reward verification during the training synchronization process.

How do I launch a GRPO training job for a qwen3-4B model?▼

To launch a GRPO training job for a qwen3-4B model, execute the training skill to initiate the process using the specific qwen3-4B model configuration script provided within the framework.

What parallelism strategies are supported during Megatron-LM reinforcement learning training?▼

Megatron-LM reinforcement learning training supports native training with full parallelism strategies, including Tensor Parallelism (TP), Pipeline Parallelism (PP), Data Parallelism (DP), and Sequence Parallelism (SP) for massive models.

Related Skills

Explore More Software Engineering Skills→
heysuhasheysuhas
2

slime-rl-training

Integrate Megatron-LM and SGLang for large-scale LLM reinforcement learning training.

Community
Advanced
matthew-johnsonmatthew-johnson

slime-rl-training

Automate RL post-training workflows for large language models with slime.

Community
Advanced
blueskies1818blueskies1818
1

slime-rl-training

Guide LLM post-training with RL using Megatron-LM and SGLang.

Community
Advanced
InverterNetworkInverterNetwork

slime-rl-training

Orchestrate Megatron-LM training and SGLang rollout generation for large-scale RL post-training.

Official
Advanced