skills.rest
projectedanx
projectedanxCOMMUNITY
Explore all skills by this author →
𝕏@ALifeInArtifyAIGitHub

slime-rl-training

Train reasoning models with Megatron-LM and SGLang for RL post-training.

1|Updated May 12, 2026
Search Tags:#llm#reinforcement learning#sglang#grpo#post-training#megatron-lm
One-click install
npx skills add https://github.com/projectedanx/hermes-agent --skill slime-rl-training-projectedanx
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: slime-rl-training
Source: https://github.com/projectedanx/hermes-agent/tree/main/optional-skills/mlops/slime
Command: npx skills add https://github.com/projectedanx/hermes-agent --skill slime-rl-training-projectedanx

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve?

This Skill addresses the complexity of scaling Reinforcement Learning (RL) for large language models by integrating Megatron-LM for training with SGLang for high-throughput rollout generation.

Core Features & Use Cases

  • RL Scaling: Supports GRPO, PPO, and other advantage estimators for reasoning model training.
  • Hybrid Architecture: Combines Megatron-LM's training efficiency with SGLang's high-throughput inference.
  • Use Case: Use this framework to train reasoning-heavy models like GLM or DeepSeek by managing complex rollout buffers and multi-turn agentic interactions.

Quick Start

Execute the slime-rl-training skill to launch a GRPO training job using the qwen3-4B model configuration and your provided training data.

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 for LLM post-training with Megatron-LM?▼

Scale LLM post-training with reinforcement learning by integrating Megatron-LM for distributed training and SGLang for high-throughput rollout generation. This framework manages complex rollout buffers to facilitate training reasoning models.

Can I use GRPO and PPO advantage estimators for reasoning model training?▼

GRPO and PPO advantage estimators are supported for reasoning model training. The framework provides custom data buffer management to handle these estimators during the reinforcement learning post-training process.

Does SGLang work with Megatron-LM for high-throughput rollout generation?▼

SGLang works with Megatron-LM to provide a hybrid architecture for high-throughput rollout generation. This combination leverages Megatron-LM's training efficiency alongside SGLang's inference capabilities.

How do I launch a GRPO training job for a qwen3-4B model?▼

Launch a GRPO training job for a qwen3-4B model by executing the framework with your provided training data. The system uses the qwen3-4B model configuration to initiate the reinforcement learning process.

Can this framework handle multi-turn agentic workflows and custom reward models?▼

The framework handles multi-turn agentic workflows and supports flexible reward model integration. It satisfies requirements for distributed training while managing complex interactions during rollout generation.

What dependencies are required for distributed LLM reinforcement learning?▼

Distributed LLM reinforcement learning requires sglang-router, ray, torch, and transformers as dependencies. These components enable the distributed training and inference infrastructure needed for the framework.

Related Skills

Explore More Software Engineering Skills→
richardnguyen0715richardnguyen0715

slime-rl-training

Train LLMs with reinforcement learning using Megatron-LM and SGLang.

Community
Advanced
zeronx798zeronx798

slime-rl-training

Integrates Megatron-LM and SGLang for RL post-training of LLMs.

Community
Advanced
matthew-johnsonmatthew-johnson

slime-rl-training

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

Community
Advanced
zangjeicyzangjeicy

slime-rl-training

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

Community
Advanced