⚠️

重要前提

安装AI Skills的关键前提是：必须科学上网，且开启TUN模式，这一点至关重要，直接决定安装能否顺利完成，在此郑重提醒三遍：科学上网，科学上网，科学上网。查看完整安装教程 →

SimPO 简单偏好优化：无需参考模型的AI对齐方法，性能优于DPO，快速上手指南

simpo-training by orchestra-research/ai-research-skills

65 周安装量

5,500 GitHub Stars

GitHub

安装命令

npx skills add https://github.com/orchestra-research/ai-research-skills --skill simpo-training

AI/机器学习 PyTorch 自然语言处理

🇨🇳中文介绍

SimPO - 简单偏好优化

快速开始

SimPO 是一种无需参考模型的偏好优化方法，其性能优于 DPO 且无需参考模型。

安装：

# 创建环境
conda create -n simpo python=3.10 && conda activate simpo

# 安装 PyTorch 2.2.2
# 访问：https://pytorch.org/get-started/locally/

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

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

训练 (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

常见工作流

工作流 1：从基础模型开始训练 (Mistral 7B)

配置 (mistral-7b-base-simpo.yaml)：

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

# 数据集
dataset_mixer:
  HuggingFaceH4/ultrafeedback_binarized: 1.0
dataset_splits:
  - train_prefs
  - test_prefs

# SimPO 超参数
beta: 2.0                  # 奖励缩放 (2.0-10.0)
gamma_beta_ratio: 0.5       # 目标边界 (0-1)
loss_type: sigmoid          # sigmoid 或 hinge
sft_weight: 0.0             # 可选的 SFT 正则化

# 训练
learning_rate: 5e-7         # 关键：3e-7 到 1e-6
num_train_epochs: 1
per_device_train_batch_size: 1
gradient_accumulation_steps: 8

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

广告位招租

在这里展示您的产品或服务

触达数万 AI 开发者，精准高效

联系我们

工作流 2：微调指令模型 (Llama 3 8B)

配置 (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             # 添加 SFT 损失以保持能力

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

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

工作流 3：推理密集型任务 (较低学习率)

适用于数学/代码任务：

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

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

beta: 5.0                   # 更高的值以获得更强的信号
gamma_beta_ratio: 0.7       # 更大的边界
learning_rate: 3e-7         # 推理任务使用较低的学习率
sft_weight: 0.0

num_train_epochs: 1
per_device_train_batch_size: 1
gradient_accumulation_steps: 16

何时使用及其替代方案

在以下情况使用 SimPO：

希望训练比 DPO 更简单（无需参考模型）
拥有偏好数据（选中/拒绝配对）
需要比 DPO 更好的性能
计算资源有限
单节点训练足够

SimPO：最简单，性能最佳，无需参考模型
DPO：需要参考模型基线，更保守
PPO：最大控制，需要奖励模型，设置复杂
GRPO：内存高效的 RL，无需评判器

在以下情况使用替代方案：

OpenRLHF：多节点分布式训练，PPO/GRPO
TRL：需要在一个框架中使用多种方法
DPO：用于建立基线比较

问题：损失发散

learning_rate: 3e-7  # 从 5e-7 降低

beta: 1.0  # 从 2.0 降低

问题：模型遗忘能力

添加 SFT 正则化：

sft_weight: 0.1  # 添加 SFT 损失组件

问题：偏好区分度差

增加 beta 和边界：

beta: 5.0            # 从 2.0 增加
gamma_beta_ratio: 0.8  # 从 0.5 增加

问题：训练期间内存不足

减小批次大小：

per_device_train_batch_size: 1
gradient_accumulation_steps: 16  # 保持有效批次大小

启用梯度检查点：

gradient_checkpointing: true

损失函数：关于 sigmoid 与 hinge 损失、数学公式以及何时使用每种损失，请参阅 references/loss-functions.md。

超参数调优：关于 beta、gamma、学习率选择指南以及特定模型大小的建议，请参阅 references/hyperparameters.md。

数据集准备：关于偏好数据格式、质量过滤和自定义数据集创建，请参阅 references/datasets.md。

GPU：推荐 NVIDIA A100/H100
显存：
- 7B 模型：1× A100 40GB (DeepSpeed ZeRO-3)
- 8B 模型：2× A100 40GB
- 70B 模型：8× A100 80GB
单节点：DeepSpeed ZeRO-3 足够
混合精度：推荐 BF16

DeepSpeed ZeRO-3 (默认配置)
梯度检查点
Flash Attention 2

🇺🇸English

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

Paper: https://arxiv.org/abs/2405.14734 (NeurIPS 2024)
GitHub: https://github.com/princeton-nlp/SimPO
Models: https://huggingface.co/princeton-nlp
Alignment Handbook: https://github.com/huggingface/alignment-handbook

Weekly Installs

Repository

orchestra-resea…h-skills

GitHub Stars

5.5K

First Seen

Feb 7, 2026

Security Audits

Gen Agent Trust HubPass SocketPass SnykWarn

Installed on

opencode55

cursor55

codex54

gemini-cli53

github-copilot52

claude-code52

超能力技能使用指南：AI助手技能调用优先级与工作流程详解

53,700 周安装