skills.rest
Devsoul2026
Devsoul2026COMMUNITY
Explore all skills by this author →
GitHub

slime-rl-training

Orchestrate end-to-end RL post-training for LLMs with Megatron-LM and SGLang.

150|25|Updated Apr 20, 2026
Search Tags:#training#rollout#reinforcement-learning#slime#rl#sglang#megatron-lm
One-click install
npx skills add https://github.com/Devsoul2026/Hermes-One-Click --skill slime-rl-training-devsoul2026
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: slime-rl-training
Source: https://github.com/Devsoul2026/Hermes-One-Click/tree/main/hermes-agent/optional-skills/mlops/slime
Command: npx skills add https://github.com/Devsoul2026/Hermes-One-Click --skill slime-rl-training-devsoul2026

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

slime enables end-to-end reinforcement learning post-training for large language models by integrating Megatron-LM training with SGLang rollout generation.

Core Features & Use Cases

  • End-to-end RL training for large language models using Megatron-LM and SGLang
  • High-throughput rollout generation and data buffering for efficient training
  • Supports multiple RL algorithms and model families (GLM, Qwen, Llama)

Quick Start

Pull the slime Docker image and start a GPU-enabled container, then install in editable mode to begin RL training.

Frequently Asked Questions about slime-rl-training

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

FAQPage Schema
How do I run end-to-end RL post-training for large language models using Megatron-LM and SGLang?▼

End-to-end RL post-training integrates Megatron-LM training with SGLang rollout generation to orchestrate scalable, high-throughput workflows for large language models. It targets researchers needing production-ready RL pipelines.

What RL algorithms can I configure for LLM post-training with slime?▼

Configurable RL algorithms for LLM post-training include GRPO, GPO, and PPO. These algorithms operate within a production-ready workflow that utilizes a reusable data buffer for efficient rollout generation and training.

Does slime RL training support model families like GLM, Qwen, and Llama?▼

RL training supports multiple model families including GLM, Qwen3, and Llama. It provides scalable, high-throughput reinforcement learning workflows tailored for these architectures on Linux or macOS environments.

What is the best way to generate high-throughput rollouts during reinforcement learning?▼

High-throughput rollout generation is achieved by integrating SGLang with Megatron-LM training. This setup uses a reusable data buffer to efficiently manage and feed rollout data back into the training loop.

Can I use slime for RL post-training on macOS, or do I need a Linux GPU container?▼

RL post-training targets both Linux and macOS environments. The quick start process involves pulling a Docker image to start a GPU-enabled container, then installing the framework in editable mode to begin training.

Related Skills

Explore More Software Engineering Skills→
ovachieverovachiever
52

training-llms-megatron

Train large language models with Megatron-Core using configurable 3D and 4D parallelism.

Community
Advanced
dawsonblockdawsonblock

slime-rl-training

Implement RL post-training for GLM models with slime, Megatron-LM, and SGLang.

Community
Advanced
tadod12tadod12

slime-rl-training

Configure Megatron-LM and SGLang arguments for post-training RL workflows.

Community
Advanced
informatico-madridinformatico-madrid
1

verl-rl-training

Train large language models with PPO and GRPO using verl.

Community
Advanced