skills.rest
markwang2658
markwang2658COMMUNITY
Explore all skills by this author →
GitHub

slime-rl-training

Plan and execute LLM RL post-training with the slime Megatron+SGLang framework.

31|3|Updated May 7, 2026
Search Tags:#framework#reinforcement-learning#slime#rl-training#sglang#post-training#megatron-lm
One-click install
npx skills add https://github.com/markwang2658/hermes-windows-native --skill slime-rl-training-markwang2658
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: slime-rl-training
Source: https://github.com/markwang2658/hermes-windows-native/tree/main/hermes-agent/optional-skills/mlops/slime
Command: npx skills add https://github.com/markwang2658/hermes-windows-native --skill slime-rl-training-markwang2658

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

Planning and executing LLM post-training with reinforcement learning using slime, enabling end-to-end RL workflows for GLM models and related architectures.

Core Features & Use Cases

  • Guided RL post-training setup with Megatron-LM and SGLang, including data buffer and rollout orchestration.
  • Pre-configured model scripts and workflow templates to accelerate experiments and reproducibility.
  • Use Case: Research teams validating RL-based fine-tuning pipelines on GLM families with controlled evaluation setups.

Quick Start

Execute a slime-rl-training session by configuring your data and model, then launch the training script to start RL post-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 set up RL post-training for LLMs using Megatron-LM and SGLang?▼

RL post-training with slime requires configuring model architectures, data buffers, and RL algorithms through structured arguments. You then execute pre-configured scripts to orchestrate rollouts and start training across Linux and macOS environments.

Can I customize reward functions during reinforcement learning fine-tuning?▼

Yes, slime supports optional custom generate and reward functions during reinforcement learning fine-tuning. You can integrate these through structured arguments to tailor the RL post-training pipeline to your specific research evaluation needs.

Does the slime RL training framework support GLM model families?▼

Yes, the slime RL training framework explicitly supports GLM model families. It provides pre-configured model scripts and workflow templates designed to accelerate reproducible RL experiments for GLM architectures.

What is needed to orchestrate data generation workflows for scalable RL experiments?▼

Orchestrating scalable RL experiments requires configuring data buffers and rollout mechanisms within the slime framework. This setup applies to research-grade data generation workflows and supports controlled evaluation across Linux and macOS environments.

Are there limitations when running RL training pipelines on macOS versus Linux?▼

The slime framework supports scalable RL experiments across both Linux and macOS environments. However, users should verify their specific hardware acceleration and distributed training configurations, as large-scale Megatron-LM deployments typically favor Linux environments.

Related Skills

Explore More Software Engineering Skills→
ovachieverovachiever
52

grpo-rl-training

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

Community
Advanced
NVlabsNVlabs
108

alpamayo1-x-rl

Automate RL post-training for Alpamayo VLM action models with Cosmos-RL and GRPO.

Official
Advanced
Clay-HHKClay-HHK
2

slime-rl-training

Automates RL post-training for LLMs with Megatron-LM and SGLang workflows.

Community
Advanced
pmcdowallpmcdowall

fine-tuning-with-trl

Align large language models via reinforcement learning and preference-based fine-tuning workflows.

Community
Advanced