🚀 hh-rlhf
本模型是基於 vicgalle/gpt2-open-instruct-v1 在 Anthropic/hh-rlhf 數據集的一個子集(15k)上進行微調的版本。它在對話場景中,對“Human:”或“Assistant:”提示的響應效果較好,尤其適合較短的回覆。在評估集上,該模型取得了如下結果:
🚀 快速開始
本模型是 vicgalle/gpt2-open-instruct-v1 在 Anthropic/hh-rlhf 數據集的一個子集(15k)上的微調版本。以下是一個簡單的對話示例,展示了模型的使用方式:
Human: Insane clown posse says...
Human: Should we look for a woman?
Assistant: It’s okay if you’re having a tough time finding what you are looking for. It’s a common question people might come up with for an argument or misunderstanding. What are you looking for, and what kind of woman would you have?
Human: Are you trying to find someone to argue
✨ 主要特性
- 對話響應良好:該模型在對話場景中,對“Human:”或“Assistant:”提示的響應效果較好,尤其適合較短的回覆。
- 特定場景研究價值:旨在研究指令模型和提示之間的交集,專注於微妙的提示交互。
📚 詳細文檔
模型描述
GPT2 open instruct 在 open-instruct 數據集上進行了全面訓練。它將一個語言模型(LM)頭重新設計為部分 rhlf 適配器,並進行了微妙的強化。
預期用途與限制
本模型旨在研究指令模型和提示之間的交集,專注於微妙的提示交互。不過,目前可能需要進行大幅改進。
訓練和評估數據
Train dataset size: 15000
Test dataset size: 500
Dataset({
features: ['chosen', 'rejected'],
num_rows: 15000
})
Dataset({
features: ['chosen', 'rejected'],
num_rows: 500
})
訓練過程
訓練超參數
訓練過程中使用了以下超參數:
- 學習率:0.0005
- 訓練批次大小:2
- 評估批次大小:1
- 隨機種子:42
- 優化器:Adam(β1 = 0.9,β2 = 0.999,ε = 1e-08)
- 學習率調度器類型:線性
- 學習率調度器預熱步數:100
- 訓練輪數:4
訓練結果
訓練損失 |
輪數 |
步數 |
驗證損失 |
2.3108 |
1.0 |
7500 |
2.1799 |
2.265 |
2.0 |
15000 |
2.1632 |
2.2507 |
3.0 |
22500 |
2.1567 |
2.2519 |
4.0 |
30000 |
2.1534 |
框架版本
- Transformers 4.31.0
- Pytorch 2.0.1+cu118
- Datasets 2.13.1
- Tokenizers 0.13.3
📄 許可證
本模型採用 MIT 許可證。
⚠️ 重要提示
該模型的回覆有時可能包含未經審查的言論和暴力內容,尤其是在提問時。在用於其他文本用途時,需要進行審查。建議將生成長度控制在合理範圍內,以獲得更好的回覆效果。