skills.rest
ChimeraFoundationa
ChimeraFoundationaCOMMUNITY
Explore all skills by this author →
GitHub

slime-rl-training

Orchestrate RL-based post-training for large language models with Megatron-LM and SGLang rollout.

Updated Mar 22, 2026
Search Tags:#llm#reinforcement-learning#slime#rl-training#sglang#megatron-lm
One-click install
npx skills add https://github.com/ChimeraFoundationa/Agentx --skill slime-rl-training-chimerafoundationa
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: slime-rl-training
Source: https://github.com/ChimeraFoundationa/Agentx/tree/main/skills/mlops/training/slime
Command: npx skills add https://github.com/ChimeraFoundationa/Agentx --skill slime-rl-training-chimerafoundationa

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

slime provides an end-to-end RL-based post-training framework for large language models by integrating Megatron-LM training with SGLang rollout, enabling scalable experimentation.

Core Features & Use Cases

  • Training with Megatron-LM parallelism and SGLang-based rollout for data generation.
  • Flexible workflows including custom generate functions, reward models, and off-policy data sampling for research and production experiments.
  • Use Case: A team wants to fine-tune a GLM-4.x model for improved instruction-following using RLHF-style signals with re-usable data buffers and multi-turn interactions.

Quick Start

Launch slime with your dataset and model script 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 scale reinforcement learning training for large language models using Megatron-LM?▼

You can scale RL training for large language models by using slime to orchestrate Megatron-LM parallel training with SGLang rollout. This integration enables scalable data generation and iterative policy optimization across GLM, Qwen, and Llama architectures.

Can I use SGLang rollout for data generation during RLHF post-training?▼

Yes, SGLang rollout is supported for data generation during RLHF post-training. The framework enables flexible workflows including custom generate functions, reward models, and off-policy data sampling for both research and production experiments.

Does slime RL training work with Qwen and Llama-like model architectures?▼

Slime RL training applies across GLM, Qwen, and Llama-like architectures. It provides pre-configured model scripts accessed via scripts/models to launch end-to-end RL training workflows for these specific frameworks.

What is the best way to set up multi-turn RLHF fine-tuning with data buffering?▼

The best way to set up multi-turn RLHF fine-tuning is using slime's re-usable data buffers and custom generate functions. This allows you to fine-tune models like GLM-4.x for improved instruction-following using RLHF-style signals.

Are custom reward models supported for off-policy data sampling in LLM training?▼

Custom reward models and off-policy data sampling are fully supported for LLM training. The framework orchestrates reward modeling and iterative policy optimization, allowing flexible data buffering and custom generate functions for scalable experimentation.

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
ovachieverovachiever
52

openrlhf-training

Train large language models with RLHF using Ray and vLLM acceleration.

Community
Advanced
hongbietcodehongbietcode
3

power-agent-creator

Create autonomous AI agents with system prompts and tool implementations.

Community
Advanced
KappTech88KappTech88

verl-rl-training

Orchestrate PPO/GRPO RL training for large language models with verl.

Community
Advanced