skills.rest
cxnaive
cxnaiveCOMMUNITY
Explore all skills by this author →
GitHub

slime-rl-training

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

Updated Jun 17, 2026
Search Tags:#llm#reinforcement learning#sglang#grpo#post-training#megatron-lm
One-click install
npx skills add https://github.com/cxnaive/hermes-agent-llbot --skill slime-rl-training-cxnaive
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: slime-rl-training
Source: https://github.com/cxnaive/hermes-agent-llbot/tree/main/optional-skills/mlops/slime
Command: npx skills add https://github.com/cxnaive/hermes-agent-llbot --skill slime-rl-training-cxnaive

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires sglang-router, ray, torch, transformers, and includes references (resource) components.

What problem does it solve?

This Skill addresses the complexity of scaling Reinforcement Learning (RL) for Large Language Models by providing a unified framework that bridges Megatron-LM training with SGLang inference.

Core Features & Use Cases

  • Integrated RL Pipeline: Supports GRPO, PPO, and other advantage estimators with native Megatron-LM parallelism.
  • High-Throughput Rollout: Utilizes SGLang for efficient response generation, enabling faster training cycles.
  • Multi-Turn Agentic Training: Facilitates training for complex reasoning models and agents that require tool-use or multi-step interaction.

Quick Start

Execute the slime training script by providing the path to your prompt data and selecting a pre-configured model script from the models directory.

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 post-training for LLMs with Megatron-LM and SGLang?▼

Scale reinforcement learning post-training for LLMs by bridging Megatron-LM distributed training with SGLang inference, providing a unified framework that simplifies scaling RL pipelines and accelerates rollout generation.

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

Generate high-throughput rollouts during LLM reinforcement learning by utilizing SGLang for efficient response generation, which enables faster training cycles and maximizes GPU resource utilization.

Can I use GRPO and PPO advantage estimators for multi-turn agentic training?▼

Yes, you can use GRPO and PPO advantage estimators for multi-turn agentic training. The framework facilitates training for complex reasoning models and agents that require tool-use or multi-step interaction.

Does this LLM reinforcement learning framework support multi-turn agentic reasoning?▼

Yes, this LLM reinforcement learning framework supports multi-turn agentic reasoning. It facilitates training for complex reasoning models and agents that require tool-use or multi-step interaction.

How do I start RL post-training using a pre-configured model script and prompt data?▼

Start RL post-training by executing the training script with the path to your prompt data and selecting a pre-configured model script from the models directory to initiate the pipeline.

Do I need Ray and Torch to run distributed LLM reinforcement learning with Megatron-LM?▼

Yes, you need Ray and Torch as dependencies to run distributed LLM reinforcement learning with Megatron-LM, alongside sglang-router and transformers for pipeline orchestration and inference.

Related Skills

Explore More Software Engineering Skills→
ovachieverovachiever
52

grpo-rl-training

Apply GRPO/RL training with TRL to optimize language models.

Community
Advanced
matthew-johnsonmatthew-johnson

slime-rl-training

Automate RL post-training workflows for large language models with slime.

Community
Advanced
qcmuuqcmuu

ray-train

Orchestrate distributed ML training on multi-node Ray clusters.

Community
Intermediate
Ced3-hanCed3-han

openrlhf-training

Run distributed RLHF training for large language models with Ray and vLLM.

Community
Advanced