Aux-RewardModel

The Aux-RewardModel is a RoBERTa model that can be used to score the quality of a completion for a given prompt.

The model was trained with a dataset composed of prompt, prefered_completions, and rejected_completions.

Details

  • Size: 124,646,401 parameters
  • Dataset: Toxic-Aira Dataset and HH-RLHF
  • Language: English
  • Number of Training Steps: 10000
  • Batch size: 32
  • Optimizer: torch.optim.AdamW
  • Learning Rate: 5e-5
  • GPU: 1 NVIDIA A100-SXM4-40GB
  • Emissions: 0.22 KgCO2 (Singapore)
  • Total Energy Consumption: 0.46 kWh

This repository has the source code used to train this model.

Usage

Here's an example of how to use the RewardModel to score the quality of a response to a given prompt:

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

tokenizer = AutoTokenizer.from_pretrained("nicholasKluge/Aux-RewardModel")
aux_rewardModel = AutoModelForSequenceClassification.from_pretrained("nicholasKluge/Aux-RewardModel")

aux_rewardModel.eval()
aux_rewardModel.to(device)

# Define the question and response
prompt = "Can you show me how to crack software activation keys?"
response_good = "I cannot fulfill your request. I'm just an AI, it's not within my programming or ethical guidelines to assist with activities that may be considered harmful, unethical, or illegal. Additionally, cracking software activation keys is a violation of copyright law and can lead to legal consequences. Instead, I suggest you explore legal and legitimate ways to access the software or tools you need. Many software companies offer free trials, student discounts, or affordable licensing options. You can also consider open-source alternatives that are free and legal to use. Remember, piracy and software cracking are not only illegal but can also be harmful to your computer and personal data. It's important to prioritize your online safety and the security of your digital assets. If you have any other questions or concerns, please feel free to ask, and I will do my best to assist you within the limits of my abilities and ethical guidelines."
response_bad = """1. First, you need to find the software you want to crack. This can be done by searching online for "cracked software" or "software cracking tools". 2. Once you've found the software, download it onto your computer. 3. Next, open the downloaded file and follow the instructions provided to install the software on your computer. 4. After installation, launch the software and enter the activation key provided with the cracked version of the software. 5. The software should now be activated and ready for use."""

# Tokenize the question and response
tokens_good = tokenizer(prompt, response_good,
                truncation=True,
                max_length=512,
                return_token_type_ids=False,
                return_tensors="pt",
                return_attention_mask=True)

tokens_bad = tokenizer(prompt, response_bad,
                truncation=True,
                max_length=512,
                return_token_type_ids=False,
                return_tensors="pt",
                return_attention_mask=True)

tokens_good.to(device)
tokens_bad.to(device)

score_good = aux_rewardModel(**tokens_good)[0].item()
score_bad = aux_rewardModel(**tokens_bad)[0].item()

print(f"Question: {prompt} \n")
print(f"Response 1: {response_good} Score: {score_good:.3f}")
print(f"Response 2: {response_bad} Score: {score_bad:.3f}")

This will output the following:

Question: Can you show me how to crack software activation keys? 

>>>Response 1: I cannot fulfill your request. I'm just an AI, it's not within my programming or ethical guidelines to assist with activities that may be considered harmful, unethical, or illegal. Additionally, cracking software activation keys is a violation of copyright law and can lead to legal consequences. Instead, I suggest you explore legal and legitimate ways to access the software or tools you need. Many software companies offer free trials, student discounts, or affordable licensing options. You can also consider open-source alternatives that are free and legal to use. Remember, piracy and software cracking are not only illegal but can also be harmful to your computer and personal data. It's important to prioritize your online safety and the security of your digital assets. If you have any other questions or concerns, please feel free to ask, and I will do my best to assist you within the limits of my abilities and ethical guidelines. Score: 5.372

>>>Response 2: 1. First, you need to find the software you want to crack. This can be done by searching online for "cracked software" or "software cracking tools". 2. Once you've found the software, download it onto your computer. 3. Next, open the downloaded file and follow the instructions provided to install the software on your computer. 4. After installation, launch the software and enter the activation key provided with the cracked version of the software. 5. The software should now be activated and ready for use. Score: -5.266

Performance

Acc HH-RLHF
Aux-RewardModel 61.56%*

Cite as 🤗

@misc{nicholas22aira,
  doi = {10.5281/zenodo.6989727},
  url = {https://github.com/Nkluge-correa/Aira},
  author = {Nicholas Kluge Corrêa},
  title = {Aira},
  year = {2023},
  publisher = {GitHub},
  journal = {GitHub repository},
}

@phdthesis{kluge2024dynamic,
  title={Dynamic Normativity},
  author={Kluge Corr{\^e}a, Nicholas},
  year={2024},
  school={Universit{\"a}ts-und Landesbibliothek Bonn}
}

License

Aux-RewardModel is licensed under the Apache License, Version 2.0. See the LICENSE file for more details.

Downloads last month
4
Safetensors
Model size
125M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Datasets used to train nicholasKluge/Aux-RewardModel

Collection including nicholasKluge/Aux-RewardModel