slime-rl-training

Name: slime-rl-training
Availability: InStock
Author: Rawgrowth-Consulting

Automate LLM post-training with reinforcement learning using Megatron-LM and SGLang.

Updated Apr 23, 2026

Search Tags:#llm #reinforcement-learning #slime #rl-training #sglang #megatron-lm #data-buffer

Ready to run

npx skills add https://github.com/Rawgrowth-Consulting/rawclaw-agent --skill slime-rl-training-rawgrowth-consulting

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: slime-rl-training
Source: https://github.com/Rawgrowth-Consulting/rawclaw-agent/tree/main/optional-skills/mlops/slime
Command: npx skills add https://github.com/Rawgrowth-Consulting/rawclaw-agent --skill slime-rl-training-rawgrowth-consulting

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

slime provides guidance for LLM post-training with reinforcement learning using slime, a Megatron+SGLang framework, enabling scalable, repeatable RL training workflows.

Core Features & Use Cases

End-to-end RL post-training: integrates Megatron-LM training with SGLang rollout for high-throughput data generation.
Data management: includes a data buffer system and tooling for custom data generation workflows.
Flexible model support: supports GLM variants and related LLMs with pre-configured scripting paths.

Quick Start

Set up and start an RL post-training workflow for an LLM using slime with Megatron-LM and SGLang.

Frequently Asked Questions about slime-rl-training

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

FAQPage Schema

How do I set up reinforcement learning post-training for LLMs using Megatron-LM and SGLang?▼

Reinforcement learning post-training for LLMs using Megatron-LM and SGLang is automated through a modular training loop, rollout generation, and pre-configured model scripts to enable scalable training workflows. It integrates Megatron-LM training with SGLang rollout for high-throughput data generation.

What is a data buffer system and how does it work in RL training pipelines?▼

A data buffer system in RL training pipelines manages and stores generated rollout data for model updates. It is included as part of the end-to-end workflow to handle custom data generation and ensure repeatable reinforcement learning cycles for large language models.

Can I use custom generate functions for rollout generation in LLM RL training?▼

Yes, you can use custom generate functions for rollout generation in LLM RL training. The framework provides tooling for custom data generation workflows alongside the integrated SGLang rollout to support flexible research and production scaling requirements.

Does this RL training pipeline support GLM variants and related large language model architectures?▼

Yes, this RL training pipeline supports GLM variants and related large language model architectures. It provides pre-configured scripting paths specifically designed to handle RL scaling and post-training for these models within the Megatron-LM and SGLang integration.

What is the best way to scale LLM reinforcement learning workflows for production?▼

The best way to scale LLM reinforcement learning workflows for production is using an integrated Megatron-LM and SGLang framework. This approach combines high-throughput rollout data generation with a modular training loop and data buffer system to ensure scalable, repeatable training.