The Beaver reward model is a preference model trained using the PKU-SafeRLHF dataset. It can play a role in the safe RLHF algorithm, helping the Beaver model become more helpful.

Developed by: the PKU-Alignment Team.
Model Type: An auto-regressive language model based on the transformer architecture.
License: Non-commercial license.
Fine-tuned from model: LLaMA, Alpaca.

Model Sources

Repository: https://github.com/PKU-Alignment/safe-rlhf
Beaver: https://huggingface.co/PKU-Alignment/beaver-7b-v1.0
Dataset: https://huggingface.co/datasets/PKU-Alignment/PKU-SafeRLHF
Reward Model: https://huggingface.co/PKU-Alignment/beaver-7b-v1.0-reward
Cost Model: https://huggingface.co/PKU-Alignment/beaver-7b-v1.0-cost
Dataset Paper: https://arxiv.org/abs/2307.04657
Paper: https://arxiv.org/abs/2310.12773

How to Use the Reward Model

python

import torch
from transformers import AutoTokenizer
from safe_rlhf.models import AutoModelForScore

model = AutoModelForScore.from_pretrained('PKU-Alignment/beaver-7b-v1.0-reward', torch_dtype=torch.bfloat16, device_map='auto')
tokenizer = AutoTokenizer.from_pretrained('PKU-Alignment/beaver-7b-v1.0-reward')

input = 'BEGINNING OF CONVERSATION: USER: hello ASSISTANT:Hello! How can I help you today?'

input_ids = tokenizer(input, return_tensors='pt')
output = model(**input_ids)
print(output)

# ScoreModelOutput(
#     scores=tensor([[[-19.7500],
#          [-19.3750],
#          [-20.1250],
#          [-18.0000],
#          [-20.0000],
#          [-23.8750],
#          [-23.5000],
#          [-22.0000],
#          [-21.0000],
#          [-20.1250],
#          [-23.7500],
#          [-21.6250],
#          [-21.7500],
#          [-12.9375],
#          [ -6.4375],
#          [ -8.1250],
#          [ -7.3438],
#          [ -9.1875],
#          [-13.6250],
#          [-10.5625],
#          [ -9.9375],
#          [ -6.4375],
#          [ -6.0938],
#          [ -5.8438],
#          [ -6.6562],
#          [ -5.9688],
#          [ -9.1875],
#          [-11.4375]]], grad_fn=<ToCopyBackward0>),
#     end_scores=tensor([[-11.4375]], grad_fn=<ToCopyBackward0>),
#     last_hidden_state=tensor([[[ 0.7461, -0.6055, -0.4980,  ...,  0.1670,  0.7812, -0.3242],
#          [ 0.7383, -0.5391, -0.1836,  ..., -0.1396,  0.5273, -0.2256],
#          [ 0.6836, -0.7031, -0.3730,  ...,  0.2100,  0.5000, -0.6328],
#          ...,
#          [-1.7969,  1.0234,  1.0234,  ..., -0.8047,  0.2500, -0.8398],
#          [ 2.0469, -1.3203,  0.8984,  ..., -0.7734, -1.4141, -1.6797],
#          [ 4.3438, -0.6953,  0.9648,  ..., -0.1787,  0.6680, -3.0000]]],
#        dtype=torch.bfloat16, grad_fn=<ToCopyBackward0>),
#     end_last_hidden_state=tensor([[ 4.3438, -0.6953,  0.9648,  ..., -0.1787,  0.6680, -3.0000]],
#        dtype=torch.bfloat16, grad_fn=<ToCopyBackward0>),
#     end_index=tensor([27])
# )