skills.rest
choice5346
choice5346COMMUNITY
Explore all skills by this author →
GitHub

slime-rl-training

Train LLMs with RL using Megatron-LM and SGLang.

Updated Aug 27, 2026
Search Tags:#llm#distributed systems#training#reinforcement learning#sglang#megatron-lm
One-click install
npx skills add https://github.com/choice5346/BiSHE --skill slime-rl-training-choice5346
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: slime-rl-training
Source: https://github.com/choice5346/BiSHE/tree/main/.github/skills/slime
Command: npx skills add https://github.com/choice5346/BiSHE --skill slime-rl-training-choice5346

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 streamlines the complex process of post-training Large Language Models (LLMs) using Reinforcement Learning (RL), specifically integrating Megatron-LM for efficient training and SGLang for high-throughput inference.

Core Features & Use Cases

  • RL Training: Facilitates advanced LLM training techniques like GRPO, GSP, PPO, and Reinforce++.
  • Framework Integration: Seamlessly connects Megatron-LM's distributed training capabilities with SGLang's inference engine.
  • Model Support: Works with a wide range of models including GLM, Qwen, DeepSeek, and Llama.
  • Use Case: Train a custom LLM to generate more helpful and harmless responses by fine-tuning it with RL, leveraging distributed training for faster iteration.

Quick Start

Use the slime-rl-training skill to launch a GRPO training job for the qwen3-4B model using data from /path/to/train.jsonl.

Frequently Asked Questions about slime-rl-training

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

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

LLM post-training with Reinforcement Learning is achieved by integrating Megatron-LM for distributed training and SGLang for high-throughput inference. This combination streamlines scalable RL workflows for large language models.

What RL algorithms are supported for distributed LLM training?▼

Supported RL algorithms for distributed LLM training include GRPO, GSP, PPO, and Reinforce++. These algorithms enable fine-tuning models to generate more helpful and harmless responses.

Can I use SGLang with Ray and Megatron-LM for LLM inference and training?▼

Yes, SGLang works with Ray and Megatron-LM for distributed LLM training and inference. The framework leverages SGLang's inference engine alongside Megatron-LM's distributed training capabilities.

Does this RL training framework support Qwen, DeepSeek, and Llama model architectures?▼

Yes, the RL training framework supports multiple model architectures including Qwen, DeepSeek, Llama, and GLM. This allows flexible fine-tuning across various prominent large language models.

How do I launch a GRPO training job for a Qwen model?▼

To launch a GRPO training job, use the slime-rl-training skill and specify your target model like qwen3-4B along with your training data path. This initiates the distributed RL training process.

Related Skills

Explore More Software Engineering Skills→
hhhi21ghhhi21g

verl-rl-training

Train large language models with reinforcement learning on distributed infrastructure.

Community
Advanced
vasilyu1983vasilyu1983
73

ai-llm-inference

Optimize LLM inference performance and cost with vLLM and quantization.

Community
Advanced
zangjeicyzangjeicy

slime-rl-training

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

Community
Advanced
jason660519jason660519

slime-rl-training

Integrate Megatron-LM training with SGLang rollout for RL post-training.

Community
Advanced