skills.rest
jleechanorg
jleechanorgOFFICIAL
Explore all skills by this author →
GitHub

slime-rl-training

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

Updated Jun 28, 2026
Search Tags:#llm#rollout#slime#sglang#megatron-lm#rl-post-training#data-buffer
One-click install
npx skills add https://github.com/jleechanorg/hermes-agent --skill slime-rl-training-jleechanorg
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: slime-rl-training
Source: https://github.com/jleechanorg/hermes-agent/tree/main/optional-skills/mlops/slime
Command: npx skills add https://github.com/jleechanorg/hermes-agent --skill slime-rl-training-jleechanorg

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

slime provides structured guidance for RL post-training of large language models using Megatron+SGLang, simplifying experimentation, reproducibility, and deployment.

Core Features & Use Cases

  • Training workflows: standard GRPO, asynchronous, and multi-turn training with tool use.
  • Data management: a flexible data buffer, rollout orchestration, and evaluation configurations.
  • Model coverage: supports GLM-4.x, Qwen3, DeepSeek V3/R1, Llama 3 with pre-configured scripts.

Quick Start

Launch slime RL post-training with the included prompts and model scripts to begin a baseline experiment.

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 Megatron and SGLang?▼

Run RL post-training for large language models using slime to orchestrate structured workflows, data buffers, and rollout tooling with Megatron and SGLang. It provides pre-configured scripts to easily launch baseline experiments.

Can I use slime to train Qwen3, DeepSeek, and Llama3 models?▼

Yes, slime supports RL post-training for Qwen3, DeepSeek V3/R1, Llama3, and GLM-4.x model families. It includes pre-configured scripts to ensure robust and reproducible experiments across these architectures.

What's the best way to implement asynchronous or multi-turn GRPO training?▼

Implement asynchronous and multi-turn GRPO training using slime's structured workflows. It provides rollout orchestration and tool use configurations to manage these complex training cycles effectively.

How do I manage data prompts and reward modeling for LLM rollouts?▼

Manage data prompts and reward modeling for LLM rollouts using slime's flexible data buffer. It enforces configurable prompts and reproducibility safeguards to ensure robust experimentation.

Why do I need a data buffer for LLM reinforcement learning experiments?▼

You need a data buffer for LLM reinforcement learning to manage rollout orchestration and evaluation configurations. Slime uses this buffer to enforce reproducibility safeguards and maintain structured training workflows.

Related Skills

Explore More Software Engineering Skills→
dawsonblockdawsonblock

slime-rl-training

Implement RL post-training for GLM models with slime, Megatron-LM, and SGLang.

Community
Advanced
guccangguccang
3

evaluating-llms-harness

Evaluate large language models against standardized benchmarks using lm-evaluation-harness.

Community
Advanced
gqf2008gqf2008

slime-rl-training

Integrate Megatron-LM training with SGLang rollout for RL fine-tuning.

Community
Advanced
CUexterCUexter
2

slime-rl-training

Run RL-based LLM post-training with Megatron-LM and SGLang for scalable rollout and GRPO workflows.

Community
Advanced