AGENT SKILL

simpo-training

Simple Preference Optimization for LLM alignment. Reference-free alternative to DPO with better performance (+6.4 points on AlpacaEval 2.0). No reference model needed, more efficient than DPO. Use for preference alignment when want simpler, faster training than DPO/PPO.

License 许可
MIT
最近更新
2026-11-01
Stars
★ 0
Forks
0

01 前言 Front Matter

name
simpo-training
author
Orchestra Research
license
MIT
version
1.0.0
metadata
{ "hermes": { "tags": [ "Post-Training", "SimPO", "Preference Optimization", "Alignment", "DPO Alternative", "Reference-Free", "LLM Alignment", "Efficient Training" ] } }
platforms
[ "linux", "macos", "windows" ]
description
Simple Preference Optimization for LLM alignment. Reference-free alternative to DPO with better performance (+6.4 points on AlpacaEval 2.0). No reference model needed, more efficient than DPO. Use for preference alignment when want simpler, faster training than DPO/PPO.
dependencies
[ "torch", "transformers", "datasets", "trl", "accelerate" ]

02 资源文件 3

03 SKILL.md 正文

SimPO - Simple Preference Optimization

Quick start

SimPO is a reference-free preference optimization method that outperforms DPO without needing a reference model.

Installation:

# Create environment
conda create -n simpo python=3.10 && conda activate simpo

# Install PyTorch 2.2.2
# Visit: https://pytorch.org/get-started/locally/

# Install alignment-handbook
git clone https://github.com/huggingface/alignment-handbook.git
cd alignment-handbook
python -m pip install .

# Install Flash Attention 2
python -m pip install flash-attn --no-build-isolation

Training (Mistral 7B):

ACCELERATE_LOG_LEVEL=info accelerate launch \
  --config_file accelerate_configs/deepspeed_zero3.yaml \
  scripts/run_simpo.py \
  training_configs/mistral-7b-base-simpo.yaml

Common workflows

Workflow 1: Train from base model (Mistral 7B)

Config (mistral-7b-base-simpo.yaml):

# Model
model_name_or_path: mistralai/Mistral-7B-v0.1
torch_dtype: bfloat16

# Dataset
dataset_mixer:
  HuggingFaceH4/ultrafeedback_binarized: 1.0
dataset_splits:
  - train_prefs
  - test_prefs

# SimPO hyperparameters
beta: 2.0                  # Reward scaling (2.0-10.0)
gamma_beta_ratio: 0.5       # Target margin (0-1)
loss_type: sigmoid          # sigmoid or hinge
sft_weight: 0.0             # Optional SFT regularization

# Training
learning_rate: 5e-7         # Critical: 3e-7 to 1e-6
num_train_epochs: 1
per_device_train_batch_size: 1
gradient_accumulation_steps: 8

# Output
output_dir: ./outputs/mistral-7b-simpo

Launch training:

accelerate launch --config_file accelerate_configs/deepspeed_zero3.yaml \
  scripts/run_simpo.py training_configs/mistral-7b-base-simpo.yaml

Workflow 2: Fine-tune instruct model (Llama 3 8B)

Config (llama3-8b-instruct-simpo.yaml):

model_name_or_path: meta-llama/Meta-Llama-3-8B-Instruct

dataset_mixer:
  argilla/ultrafeedback-binarized-preferences-cleaned: 1.0

beta: 2.5
gamma_beta_ratio: 0.5
learning_rate: 5e-7
sft_weight: 0.1             # Add SFT loss to preserve capabilities

num_train_epochs: 1
per_device_train_batch_size: 2
gradient_accumulation_steps: 4
output_dir: ./outputs/llama3-8b-simpo

Launch:

accelerate launch --config_file accelerate_configs/deepspeed_zero3.yaml \
  scripts/run_simpo.py training_configs/llama3-8b-instruct-simpo.yaml

Workflow 3: Reasoning-intensive tasks (lower LR)

For math/code tasks:

model_name_or_path: deepseek-ai/deepseek-math-7b-base

dataset_mixer:
  argilla/distilabel-math-preference-dpo: 1.0

beta: 5.0                   # Higher for stronger signal
gamma_beta_ratio: 0.7       # Larger margin
learning_rate: 3e-7         # Lower LR for reasoning
sft_weight: 0.0

num_train_epochs: 1
per_device_train_batch_size: 1
gradient_accumulation_steps: 16

When to use vs alternatives

Use SimPO when:

  • Want simpler training than DPO (no reference model)
  • Have preference data (chosen/rejected pairs)
  • Need better performance than DPO
  • Limited compute resources
  • Single-node training sufficient

Algorithm selection:

  • SimPO: Simplest, best performance, no reference model
  • DPO: Need reference model baseline, more conservative
  • PPO: Maximum control, need reward model, complex setup
  • GRPO: Memory-efficient RL, no critic

Use alternatives instead:

  • OpenRLHF: Multi-node distributed training, PPO/GRPO
  • TRL: Need multiple methods in one framework
  • DPO: Established baseline comparison

Common issues

Issue: Loss divergence

Reduce learning rate:

learning_rate: 3e-7  # Reduce from 5e-7

Reduce beta:

beta: 1.0  # Reduce from 2.0

Issue: Model forgets capabilities

Add SFT regularization:

sft_weight: 0.1  # Add SFT loss component

Issue: Poor preference separation

Increase beta and margin:

beta: 5.0            # Increase from 2.0
gamma_beta_ratio: 0.8  # Increase from 0.5

Issue: OOM during training

Reduce batch size:

per_device_train_batch_size: 1
gradient_accumulation_steps: 16  # Maintain effective batch

Enable gradient checkpointing:

gradient_checkpointing: true

Advanced topics

Loss functions: See references/loss-functions.md for sigmoid vs hinge loss, mathematical formulations, and when to use each.

Hyperparameter tuning: See references/hyperparameters.md for beta, gamma, learning rate selection guide, and model-size-specific recommendations.

Dataset preparation: See references/datasets.md for preference data formats, quality filtering, and custom dataset creation.

Hardware requirements

  • GPU: NVIDIA A100/H100 recommended
  • VRAM:
    • 7B model: 1× A100 40GB (DeepSpeed ZeRO-3)
    • 8B model: 2× A100 40GB
    • 70B model: 8× A100 80GB
  • Single-node: DeepSpeed ZeRO-3 sufficient
  • Mixed precision: BF16 recommended

Memory optimization:

  • DeepSpeed ZeRO-3 (default config)
  • Gradient checkpointing
  • Flash Attention 2

Resources

04 同仓库更多 Skills

hikmet4434
hikmet4434/claude-skills

yeet

Use only when the user explicitly asks to stage, commit, push, and open a GitHub pull request in one flow using the GitHub CLI (`gh`).

★ 0
hikmet4434
hikmet4434/claude-skills

yayin-seo-geo-yasal-denetim

Bir web uygulaması canlıya çıkmadan veya iş teslimi biterken resmî/kurumsal sayfaları, SEO, GEO, indeksleme ve canlı çıktıyı denetler; eksikleri kodda tamamlar ve üretimde kanıtlar. Uygulama bitti, canlıya hazır mı, yayına al, launch checklist, release readiness, SEO audit, GEO audit, legal pages, privacy policy, terms, cookies, KVKK, robots, sitemap, canonical, hreflang, JSON-LD, llms.txt ifadelerinde kullan. Yeni veya önemli ölçüde değişmiş halka açık web uygulamasının son teslim aşamasında kullanıcı ayrıca istemese de uygula. Güvenlik sızma testi için güvenlik skillini ayrıca kullan.

★ 0
hikmet4434
hikmet4434/claude-skills

xurl

X/Twitter via xurl CLI: post, search, DM, media, v2 API.

★ 0
hikmet4434
hikmet4434/claude-skills

wrangler

Cloudflare Workers CLI for deploying, developing, and managing Workers, KV, R2, D1, Vectorize, Hyperdrive, Workers AI, Containers, Queues, Workflows, Pipelines, and Secrets Store. Load before running wrangler commands to ensure correct syntax and best practices. Biases towards retrieval from Cloudflare docs over pre-trained knowledge.

★ 0
hikmet4434
hikmet4434/claude-skills

winui-app

Bootstrap, develop, and design modern WinUI 3 desktop applications with C# and the Windows App SDK using official Microsoft guidance, WinUI Gallery patterns, Windows App SDK samples, and CommunityToolkit components. Use when creating a brand new app, preparing a machine for WinUI, reviewing, refactoring, planning, troubleshooting, environment-checking, or setting up WinUI 3 XAML, controls, navigation, windowing, theming, accessibility, responsiveness, performance, deployment, or related Windows app design and development work.

★ 0
hikmet4434
hikmet4434/claude-skills

whisper

OpenAI's general-purpose speech recognition model. Supports 99 languages, transcription, translation to English, and language identification. Six model sizes from tiny (39M params) to large (1550M params). Use for speech-to-text, podcast transcription, or multilingual audio processing. Best for robust, multilingual ASR.

★ 0