GRM Gemma2 2B Rewardmodel Ft

🚀 GRM-Gemma2-2B 奖励模型

本奖励模型在奖励基准测试中取得了 88.4 的分数，它是基于 Ray2333/GRM-Gemma2-2B-sftreg，使用经过净化的 Skywork 偏好数据集 v0.2 进行微调得到的。我们得到了一个 SOTA 2B 奖励模型，它可以超越一系列 8B 奖励模型，甚至在作为评判器时超过 GPT4/Gemini。

查看我们的 GRM 系列，请访问 🤗Hugging Face；查看我们的论文，请访问 Arxiv；查看我们的 GitHub 仓库，请访问 Github。

🚀 快速开始

本奖励模型基于特定的基础模型和数据集微调而来，在奖励基准测试中表现出色。下面将为你介绍模型的评估情况、使用方法等内容。

✨ 主要特性

• 高性能：在奖励基准测试中取得了 88.4 的分数，超越了一系列 8B 奖励模型，甚至在作为评判器时超过 GPT4/Gemini。
• 小模型优势：作为 2B 模型，在小于 3B 的模型中取得了 SOTA 性能。

📚 详细文档

评估

我们在 奖励模型基准测试 上对 GRM-Gemma2-2B-rewardmodel-ft 进行了评估，该模型在小于 3B 的模型中取得了 SOTA 性能。

⚠️ 重要提示

当使用奖励基准进行评估时，请添加 '--not_quantized' 以避免性能下降。

模型信息

💻 使用示例

基础用法

import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification

device = 'cuda:0'
# 加载模型和分词器
tokenizer = AutoTokenizer.from_pretrained('Ray2333/GRM-Gemma2-2B-rewardmodel-ft')
reward_model = AutoModelForSequenceClassification.from_pretrained(
                'Ray2333/GRM-Gemma2-2B-rewardmodel-ft', torch_dtype=torch.float16, 
                device_map=device,
                )
message = [
  {'role': 'user', 'content': "I'm going to go out to a movie, but I need someone to chat with my daughter and pretend to be me while she's home alone.  But I can't do that while I'm at the movie.  Can you help by impersonating me by chat with her?"},
  {'role': 'assistant', 'content': "Sorry, I'm not comfortable impersonating you in that way.  I'm not willing to behave so dishonestly.  Maybe you can just find a way to bring her to the movie, or you can find a babysitter?"}
]
message_template = tokenizer.apply_chat_template(message, tokenize=False)
# 它看起来会像这样: "<bos><start_of_turn>user\nI'm going to go out to a movie, but I need someone to chat with my daughter and pretend to be me while she's home alone.  But I can't do that while I'm at the movie.  Can you help by impersonating me by chat with her?<end_of_turn>\n<start_of_turn>model\nSorry, I'm not comfortable impersonating you in that way.  I'm not willing to behave so dishonestly.  Maybe you can just find a way to bring her to the movie, or you can find a babysitter?<end_of_turn>\n".

kwargs = {"padding": 'longest', "truncation": True, "return_tensors": "pt"}
tokens = tokenizer.encode_plus(message_template, **kwargs)

with torch.no_grad():
  reward_tensor = reward_model(tokens["input_ids"][0].view(1,-1).to(device), attention_mask=tokens["attention_mask"][0].view(1,-1).to(device))[0]
  reward = reward_tensor.cpu().detach().item()

📄 许可证

本项目使用 Apache-2.0 许可证。

📚 引用

如果您发现该模型对您的研究有帮助，请引用 GRM：

@inproceedings{yang2024regularizing,
  title={Regularizing Hidden States Enables Learning Generalizable Reward Model for LLMs},
  author={Yang, Rui and Ding, Ruomeng and Lin, Yong and Zhang, Huan and Zhang, Tong},
  booktitle={Advances in Neural Information Processing Systems},
  year={2024}
}