skills.rest
Harries
HarriesCOMMUNITY
Explore all skills by this author →
Website𝕏@HarriesSteeleGitHub

slime-rl-training

Automate GRPO-style RL post-training workflows for large language models with Megatron-LM and SGLang.

Updated Apr 24, 2026
Search Tags:#llm#slime#rl-training#sglang#post-training#megatron-lm
One-click install
npx skills add https://github.com/Harries/hermes-agent --skill slime-rl-training-harries
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: slime-rl-training
Source: https://github.com/Harries/hermes-agent/tree/main/optional-skills/mlops/slime
Command: npx skills add https://github.com/Harries/hermes-agent --skill slime-rl-training-harries

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires sglang-router>=0.2.3, ray, torch>=2.0.0, transformers>=4.40.0.

What problem does it solve?

slime provides a structured guide for RL post-training of large language models using Megatron-LM with SGLang, helping teams implement and scale reinforcement learning-based fine-tuning.

Core Features & Use Cases

  • Training with Megatron-LM and SGLang for RL fine-tuning of GLM models.
  • Rollout generation and data buffering for high-throughput experiments.
  • Async training, evaluation workflows, and multi-turn/tooling support.

Quick Start

Launch a GRPO training workflow by sourcing a model script and running the slime training entry with the recommended arguments.

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 RL post-training for large language models using Megatron-LM and SGLang?▼

slime automates GRPO-style rollout training and asynchronous data buffering within a Megatron-LM and SGLang environment. It provides a structured workflow to manage high-throughput reinforcement learning experiments and scale distributed training runs.

Does slime support multi-turn interactions and tooling during RL fine-tuning?▼

Yes, slime supports multi-turn interactions with tooling during RL fine-tuning. It utilizes asynchronous data buffering across training runs to handle complex rollout generation and evaluation workflows for large language models.

What prerequisites are needed to run GRPO training workflows with slime?▼

Running GRPO training workflows requires a pre-configured Megatron-LM environment, SGLang integration, and slime model scripts. You also need dependencies including Ray, Torch, Transformers, and sglang-router to execute the rollout and reward arguments.

How do I launch a reinforcement learning training run for GLM models in slime?▼

To launch a reinforcement learning training run, source a pre-configured slime model script and execute the slime training entry point. You must provide recommended arguments for rollout generation, prompts, and rewards to start the GRPO workflow.

What is the best way to handle asynchronous data buffering during LLM rollout generation?▼

Using slime is an effective way to handle asynchronous data buffering during LLM rollout generation. It automates data buffering across training runs to maintain high throughput for Megatron-LM and SGLang fine-tuning experiments.

Related Skills

Explore More Software Engineering Skills→
afel6afel6

slime-rl-training

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

Community
Advanced
jleechanorgjleechanorg

slime-rl-training

Guide RL-based post-training of large language models with slime workflows.

Official
Advanced
aayushsoamaayushsoam
1

slime-rl-training

Automate LLM post-training with slime and Megatron-LM integration.

Community
Advanced
ihatesea69ihatesea69
3

slime-rl-training

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

Community
Advanced