AGENT SKILL

fine-tuning-with-trl

Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.

License 许可
MIT
最近更新
2026-06-24
Stars
★ 3
Forks
1

01 前言 Front Matter

name
fine-tuning-with-trl
author
Orchestra Research
license
MIT
version
1.0.0
metadata
{ "hermes": { "tags": [ "Post-Training", "TRL", "Reinforcement Learning", "Fine-Tuning", "SFT", "DPO", "PPO", "GRPO", "RLHF", "Preference Alignment", "HuggingFace" ] } }
description
Fine-tune LLMs using reinforcement learning with TRL - SFT for instruction tuning, DPO for preference alignment, PPO/GRPO for reward optimization, and reward model training. Use when need RLHF, align model with preferences, or train from human feedback. Works with HuggingFace Transformers.
dependencies
[ "trl", "transformers", "datasets", "peft", "accelerate", "torch" ]

03 SKILL.md 正文

TRL - Transformer Reinforcement Learning

Quick start

TRL provides post-training methods for aligning language models with human preferences.

Installation:

pip install trl transformers datasets peft accelerate

Supervised Fine-Tuning (instruction tuning):

from trl import SFTTrainer

trainer = SFTTrainer(
    model="Qwen/Qwen2.5-0.5B",
    train_dataset=dataset,  # Prompt-completion pairs
)
trainer.train()

DPO (align with preferences):

from trl import DPOTrainer, DPOConfig

config = DPOConfig(output_dir="model-dpo", beta=0.1)
trainer = DPOTrainer(
    model=model,
    args=config,
    train_dataset=preference_dataset,  # chosen/rejected pairs
    processing_class=tokenizer
)
trainer.train()

Common workflows

Workflow 1: Full RLHF pipeline (SFT → Reward Model → PPO)

Complete pipeline from base model to human-aligned model.

Copy this checklist:

RLHF Training:
- [ ] Step 1: Supervised fine-tuning (SFT)
- [ ] Step 2: Train reward model
- [ ] Step 3: PPO reinforcement learning
- [ ] Step 4: Evaluate aligned model

Step 1: Supervised fine-tuning

Train base model on instruction-following data:

from transformers import AutoModelForCausalLM, AutoTokenizer
from trl import SFTTrainer, SFTConfig
from datasets import load_dataset

# Load model
model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-0.5B")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-0.5B")

# Load instruction dataset
dataset = load_dataset("trl-lib/Capybara", split="train")

# Configure training
training_args = SFTConfig(
    output_dir="Qwen2.5-0.5B-SFT",
    per_device_train_batch_size=4,
    num_train_epochs=1,
    learning_rate=2e-5,
    logging_steps=10,
    save_strategy="epoch"
)

# Train
trainer = SFTTrainer(
    model=model,
    args=training_args,
    train_dataset=dataset,
    tokenizer=tokenizer
)
trainer.train()
trainer.save_model()

Step 2: Train reward model

Train model to predict human preferences:

from transformers import AutoModelForSequenceClassification
from trl import RewardTrainer, RewardConfig

# Load SFT model as base
model = AutoModelForSequenceClassification.from_pretrained(
    "Qwen2.5-0.5B-SFT",
    num_labels=1  # Single reward score
)
tokenizer = AutoTokenizer.from_pretrained("Qwen2.5-0.5B-SFT")

# Load preference data (chosen/rejected pairs)
dataset = load_dataset("trl-lib/ultrafeedback_binarized", split="train")

# Configure training
training_args = RewardConfig(
    output_dir="Qwen2.5-0.5B-Reward",
    per_device_train_batch_size=2,
    num_train_epochs=1,
    learning_rate=1e-5
)

# Train reward model
trainer = RewardTrainer(
    model=model,
    args=training_args,
    processing_class=tokenizer,
    train_dataset=dataset
)
trainer.train()
trainer.save_model()

Step 3: PPO reinforcement learning

Optimize policy using reward model:

python -m trl.scripts.ppo \
    --model_name_or_path Qwen2.5-0.5B-SFT \
    --reward_model_path Qwen2.5-0.5B-Reward \
    --dataset_name trl-internal-testing/descriptiveness-sentiment-trl-style \
    --output_dir Qwen2.5-0.5B-PPO \
    --learning_rate 3e-6 \
    --per_device_train_batch_size 64 \
    --total_episodes 10000

Step 4: Evaluate

from transformers import pipeline

# Load aligned model
generator = pipeline("text-generation", model="Qwen2.5-0.5B-PPO")

# Test
prompt = "Explain quantum computing to a 10-year-old"
output = generator(prompt, max_length=200)[0]["generated_text"]
print(output)

Workflow 2: Simple preference alignment with DPO

Align model with preferences without reward model.

Copy this checklist:

DPO Training:
- [ ] Step 1: Prepare preference dataset
- [ ] Step 2: Configure DPO
- [ ] Step 3: Train with DPOTrainer
- [ ] Step 4: Evaluate alignment

Step 1: Prepare preference dataset

Dataset format:

{
  "prompt": "What is the capital of France?",
  "chosen": "The capital of France is Paris.",
  "rejected": "I don't know."
}

Load dataset:

from datasets import load_dataset

dataset = load_dataset("trl-lib/ultrafeedback_binarized", split="train")
# Or load your own
# dataset = load_dataset("json", data_files="preferences.json")

Step 2: Configure DPO

from trl import DPOConfig

config = DPOConfig(
    output_dir="Qwen2.5-0.5B-DPO",
    per_device_train_batch_size=4,
    num_train_epochs=1,
    learning_rate=5e-7,
    beta=0.1,  # KL penalty strength
    max_prompt_length=512,
    max_length=1024,
    logging_steps=10
)

Step 3: Train with DPOTrainer

from transformers import AutoModelForCausalLM, AutoTokenizer
from trl import DPOTrainer

model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct")

trainer = DPOTrainer(
    model=model,
    args=config,
    train_dataset=dataset,
    processing_class=tokenizer
)

trainer.train()
trainer.save_model()

CLI alternative:

trl dpo \
    --model_name_or_path Qwen/Qwen2.5-0.5B-Instruct \
    --dataset_name argilla/Capybara-Preferences \
    --output_dir Qwen2.5-0.5B-DPO \
    --per_device_train_batch_size 4 \
    --learning_rate 5e-7 \
    --beta 0.1

Workflow 3: Memory-efficient online RL with GRPO

Train with reinforcement learning using minimal memory.

Copy this checklist:

GRPO Training:
- [ ] Step 1: Define reward function
- [ ] Step 2: Configure GRPO
- [ ] Step 3: Train with GRPOTrainer

Step 1: Define reward function

def reward_function(completions, **kwargs):
    """
    Compute rewards for completions.

    Args:
        completions: List of generated texts

    Returns:
        List of reward scores (floats)
    """
    rewards = []
    for completion in completions:
        # Example: reward based on length and unique words
        score = len(completion.split())  # Favor longer responses
        score += len(set(completion.lower().split()))  # Reward unique words
        rewards.append(score)
    return rewards

Or use a reward model:

from transformers import pipeline

reward_model = pipeline("text-classification", model="reward-model-path")

def reward_from_model(completions, prompts, **kwargs):
    # Combine prompt + completion
    full_texts = [p + c for p, c in zip(prompts, completions)]
    # Get reward scores
    results = reward_model(full_texts)
    return [r["score"] for r in results]

Step 2: Configure GRPO

from trl import GRPOConfig

config = GRPOConfig(
    output_dir="Qwen2-GRPO",
    per_device_train_batch_size=4,
    num_train_epochs=1,
    learning_rate=1e-5,
    num_generations=4,  # Generate 4 completions per prompt
    max_new_tokens=128
)

Step 3: Train with GRPOTrainer

from datasets import load_dataset
from trl import GRPOTrainer

# Load prompt-only dataset
dataset = load_dataset("trl-lib/tldr", split="train")

trainer = GRPOTrainer(
    model="Qwen/Qwen2-0.5B-Instruct",
    reward_funcs=reward_function,  # Your reward function
    args=config,
    train_dataset=dataset
)

trainer.train()

CLI:

trl grpo \
    --model_name_or_path Qwen/Qwen2-0.5B-Instruct \
    --dataset_name trl-lib/tldr \
    --output_dir Qwen2-GRPO \
    --num_generations 4

When to use vs alternatives

Use TRL when:

  • Need to align model with human preferences
  • Have preference data (chosen/rejected pairs)
  • Want to use reinforcement learning (PPO, GRPO)
  • Need reward model training
  • Doing RLHF (full pipeline)

Method selection:

  • SFT: Have prompt-completion pairs, want basic instruction following
  • DPO: Have preferences, want simple alignment (no reward model needed)
  • PPO: Have reward model, need maximum control over RL
  • GRPO: Memory-constrained, want online RL
  • Reward Model: Building RLHF pipeline, need to score generations

Use alternatives instead:

  • HuggingFace Trainer: Basic fine-tuning without RL
  • Axolotl: YAML-based training configuration
  • LitGPT: Educational, minimal fine-tuning
  • Unsloth: Fast LoRA training

Common issues

Issue: OOM during DPO training

Reduce batch size and sequence length:

config = DPOConfig(
    per_device_train_batch_size=1,  # Reduce from 4
    max_length=512,  # Reduce from 1024
    gradient_accumulation_steps=8  # Maintain effective batch
)

Or use gradient checkpointing:

model.gradient_checkpointing_enable()

Issue: Poor alignment quality

Tune beta parameter:

# Higher beta = more conservative (stays closer to reference)
config = DPOConfig(beta=0.5)  # Default 0.1

# Lower beta = more aggressive alignment
config = DPOConfig(beta=0.01)

Issue: Reward model not learning

Check loss type and learning rate:

config = RewardConfig(
    learning_rate=1e-5,  # Try different LR
    num_train_epochs=3  # Train longer
)

Ensure preference dataset has clear winners:

# Verify dataset
print(dataset[0])
# Should have clear chosen > rejected

Issue: PPO training unstable

Adjust KL coefficient:

config = PPOConfig(
    kl_coef=0.1,  # Increase from 0.05
    cliprange=0.1  # Reduce from 0.2
)

Advanced topics

SFT training guide: See references/sft-training.md for dataset formats, chat templates, packing strategies, and multi-GPU training.

DPO variants: See references/dpo-variants.md for IPO, cDPO, RPO, and other DPO loss functions with recommended hyperparameters.

Reward modeling: See references/reward-modeling.md for outcome vs process rewards, Bradley-Terry loss, and reward model evaluation.

Online RL methods: See references/online-rl.md for PPO, GRPO, RLOO, and OnlineDPO with detailed configurations.

Hardware requirements

  • GPU: NVIDIA (CUDA required)
  • VRAM: Depends on model and method
    • SFT 7B: 16GB (with LoRA)
    • DPO 7B: 24GB (stores reference model)
    • PPO 7B: 40GB (policy + reward model)
    • GRPO 7B: 24GB (more memory efficient)
  • Multi-GPU: Supported via accelerate
  • Mixed precision: BF16 recommended (A100/H100)

Memory optimization:

  • Use LoRA/QLoRA for all methods
  • Enable gradient checkpointing
  • Use smaller batch sizes with gradient accumulation

Resources

? Связи

  • [[000 - Map of Maps]] — REPO_DOCS
  • [[000 - Map of Maps]] — Map of Maps

04 同仓库更多 Skills

ibragimov-oasis
ibragimov-oasis/vibe-coder

ckm:ui-styling

Create beautiful, accessible user interfaces with shadcn/ui components (built on Radix UI + Tailwind), Tailwind CSS utility-first styling, and canvas-based visual designs. Use when building user interfaces, implementing design systems, creating responsive layouts, adding accessible components (dialogs, dropdowns, forms, tables), customizing themes and colors, implementing dark mode, generating visual designs and posters, or establishing consistent styling patterns across applications.

设计师视觉设计
★ 3
ibragimov-oasis
ibragimov-oasis/vibe-coder

ckm:design

Comprehensive design skill: brand identity, design tokens, UI styling, logo generation (55 styles, Gemini AI), corporate identity program (50 deliverables, CIP mockups), HTML presentations (Chart.js), banner design (22 styles, social/ads/web/print), icon design (15 styles, SVG, Gemini 3.1 Pro), social photos (HTML→screenshot, multi-platform). Actions: design logo, create CIP, generate mockups, build slides, design banner, generate icon, create social photos, social media images, brand identity, design system. Platforms: Facebook, Twitter, LinkedIn, YouTube, Instagram, Pinterest, TikTok, Threads, Google Ads.

设计师演示文稿视觉设计
★ 3
ibragimov-oasis
ibragimov-oasis/vibe-coder

ckm:design-system

Token architecture, component specifications, and slide generation. Three-layer tokens (primitive→semantic→component), CSS variables, spacing/typography scales, component specs, strategic slide creation. Use for design tokens, systematic design, brand-compliant presentations.

设计师视觉设计
★ 3
ibragimov-oasis
ibragimov-oasis/vibe-coder

ckm:banner-design

Design banners for social media, ads, website heroes, creative assets, and print. Multiple art direction options with AI-generated visuals. Actions: design, create, generate banner. Platforms: Facebook, Twitter/X, LinkedIn, YouTube, Instagram, Google Display, website hero, print. Styles: minimalist, gradient, bold typography, photo-based, illustrated, geometric, retro, glassmorphism, 3D, neon, duotone, editorial, collage. Uses ui-ux-pro-max, frontend-design, ai-artist, ai-multimodal skills.

设计师前端开发视觉设计
★ 3
ibragimov-oasis
ibragimov-oasis/vibe-coder

seo

Comprehensive SEO analysis for any website or business type. Performs full site audits, single-page deep analysis, technical SEO checks (crawlability, indexability, Core Web Vitals with INP), schema markup detection/validation/generation, content quality assessment (E-E-A-T framework per Dec 2025 update extending to all competitive queries), image optimization, sitemap analysis, and Generative Engine Optimization (GEO) for AI Overviews, ChatGPT, and Perplexity citations. Analyzes AI crawler accessibility (GPTBot, ClaudeBot, PerplexityBot), llms.txt compliance, brand mention signals, and passage-level citability. Industry detection for SaaS, e-commerce, local business, publishers, agencies. Triggers on: "SEO", "audit", "schema", "Core Web Vitals", "sitemap", "E-E-A-T", "AI Overviews", "GEO", "technical SEO", "content quality", "page speed", "structured data".

市场营销营销推广
★ 3
ibragimov-oasis
ibragimov-oasis/vibe-coder

seo-technical

Technical SEO audit across 9 categories: crawlability, indexability, security, URL structure, mobile, Core Web Vitals, structured data, JavaScript rendering, and IndexNow protocol. Use when user says "technical SEO", "crawl issues", "robots.txt", "Core Web Vitals", "site speed", or "security headers".

市场营销营销推广
★ 3