skills.rest
jacardl
jacardlCOMMUNITY
Explore all skills by this author →
GitHub

slime-rl-training

Run RL post-training for large language models with Megatron-LM and SGLang.

Updated Apr 16, 2026
Search Tags:#llm#reinforcement-learning#slime#rl-training#sglang#megatron-lm
One-click install
npx skills add https://github.com/jacardl/New-Radar --skill slime-rl-training-jacardl
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: slime-rl-training
Source: https://github.com/jacardl/New-Radar/tree/main/backend/frameworks/hermes-agent/optional-skills/mlops/slime
Command: npx skills add https://github.com/jacardl/New-Radar --skill slime-rl-training-jacardl

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires sglang-router>=0.2.3, ray, torch>=2.0.0, transformers>=4.40.0, and includes references (resource) components.

What problem does it solve?

slime accelerates and stabilizes reinforcement learning-based post-training for large language models by integrating Megatron-LM with SGLang for high-throughput rollout and data-buffer workflows.

Core Features & Use Cases

  • End-to-end RL post-training for large language models using a Megatron+SGLang stack to improve alignment and throughput.
  • Supports multiple RL algorithms (GRPO, GPO, PPO, reinforce_plus_plus) and multi-turn tool-use workflows with data buffers.
  • Practical scenarios include training GLM-4.x, Qwen3, or Llama3-scale models with custom data generation and rollout pipelines.

Quick Start

Start RL post-training by running the slime training workflow with your model script and dataset.

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 RL post-training for large language models using slime?▼

RL post-training for large language models using slime is performed by executing the slime training workflow with your configured model script and dataset to optimize alignment and policy.

Does slime RL training support PPO and GRPO algorithms?▼

Yes, slime RL training supports PPO and GRPO algorithms, alongside GPO and reinforce_plus_plus configurations, to optimize large language model alignment and policy.

Can I use slime for RL training with Megatron-LM and SGLang?▼

Yes, you can use slime for RL training with Megatron-LM and SGLang, as it integrates Megatron-LM with SGLang rollout and data-buffer workflows to accelerate and stabilize throughput.

What dependencies do I need to run slime for reinforcement learning?▼

To run slime for reinforcement learning, you need the slime runtime and its dependencies, which include sglang-router, ray, torch, and transformers, along with configured rollout parameters.

How does slime handle multi-turn tool-using pipelines during RL training?▼

slime handles multi-turn tool-using pipelines during RL training by utilizing data buffers within the Megatron-LM and SGLang stack to maintain high-throughput rollout workflows.

Related Skills

Explore More Software Engineering Skills→
nelohenriqnelohenriq
1

slime-rl-training

Orchestrates post-training reinforcement learning workflows for LLMs using slime, Megatron-LM integration, and SGLang rollout.

Community
Advanced
huggingfacehuggingface
10.9k

hugging-face-model-trainer

Train and fine-tune language models on Hugging Face Jobs using TRL methods.

Official
Advanced
KappTech88KappTech88

slime-rl-training

Deploy RLHF-style post-training pipelines for LLMs with Megatron-LM and SGLang.

Community
Advanced
ldzhhxxldzhhxx
1

slime-rl-training

Run reinforcement learning post-training for LLMs with slime, Megatron-LM, and SGLang.

Community
Advanced