Tutorials

Hands-on tutorials to help you master Flux for various RLHF training scenarios.

---

Learning Path

graph LR
    A[Basic RLHF] --> B[Custom Rewards]
    B --> C[Multi-GPU]
    C --> D[DPO Training]
    D --> E[Adaptive Async]
    E --> F[Production]

---

Beginner Tutorials

• Basic RLHF Training

  ---

  Complete walkthrough of training an LLM with RLHF using Flux.

  Time: 30 minutes Prerequisites: Flux installed

  Start Tutorial

• Custom Reward Functions

  ---

  Learn to create custom reward functions for your specific task.

  Time: 20 minutes Prerequisites: Basic RLHF tutorial

  Start Tutorial

---

Intermediate Tutorials

• Multi-GPU Training

  ---

  Scale your training across multiple GPUs on a single node.

  Time: 45 minutes Prerequisites: Basic training working

  Start Tutorial

• Fine-tuning with DPO

  ---

  Use Direct Preference Optimization for preference learning.

  Time: 30 minutes Prerequisites: Preference data available

  Start Tutorial

---

Advanced Tutorials

• Adaptive Async in Practice

  ---

  Deep dive into configuring and monitoring adaptive async control.

  Time: 60 minutes Prerequisites: Multi-GPU training

  Start Tutorial

• Production Deployment

  ---

  Deploy Flux training at scale with monitoring and fault tolerance.

  Time: 90 minutes Prerequisites: All previous tutorials

  Start Tutorial

---

Quick Reference

Tutorial	Difficulty	Time	Key Topics
Basic RLHF	Beginner	30 min	FluxTrainer, GRPO, basic config
Custom Rewards	Beginner	20 min	RewardFunction, FunctionReward
Multi-GPU	Intermediate	45 min	TP, DP, distributed training
DPO Training	Intermediate	30 min	DPO algorithm, preference data
Adaptive Async	Advanced	60 min	PID tuning, staleness monitoring
Production	Advanced	90 min	Monitoring, checkpoints, scaling

---

Tutorial Format

Each tutorial follows a consistent structure:

1. Overview - What you'll learn
2. Prerequisites - What you need before starting
3. Setup - Environment and data preparation
4. Step-by-Step - Detailed instructions
5. Verification - How to know it worked
6. Troubleshooting - Common issues and solutions
7. Next Steps - Where to go from here

---

Sample Projects

Complete example projects you can clone and run:

Math Reasoning (GSM8K)

Train a model for mathematical reasoning using GRPO.

git clone https://github.com/flux-team/flux-examples
cd flux-examples/math-reasoning
./run.sh

Code Generation (HumanEval)

Train a model for code generation with custom evaluation.

git clone https://github.com/flux-team/flux-examples
cd flux-examples/code-generation
./run.sh

Chat Assistant (UltraChat)

Build a general-purpose chat assistant.

git clone https://github.com/flux-team/flux-examples
cd flux-examples/chat-assistant
./run.sh

---

Video Tutorials

Coming soon! Subscribe to our YouTube channel for video walkthroughs.

---

Community Tutorials

Have you written a tutorial about Flux? Submit it here and we'll feature it!

---

Getting Help

Stuck on a tutorial? Here's how to get help:

1. Check the FAQ
2. Search GitHub Issues
3. Ask on Discord
4. Open a new issue with the tutorial-help label