README.md
| 1 | --- |
| 2 | datasets: |
| 3 | - PKU-Alignment/PKU-SafeRLHF |
| 4 | language: |
| 5 | - en |
| 6 | tags: |
| 7 | - reinforcement-learning-from-human-feedback |
| 8 | - reinforcement-learning |
| 9 | - beaver |
| 10 | - safety |
| 11 | - llama |
| 12 | - ai-safety |
| 13 | - deepspeed |
| 14 | - rlhf |
| 15 | - alpaca |
| 16 | library_name: safe-rlhf |
| 17 | --- |
| 18 | |
| 19 | # 🦫 Beaver's Reward Model |
| 20 | |
| 21 | ## Model Details |
| 22 | |
| 23 | The Beaver reward model is a preference model trained using the [PKU-SafeRLHF](https://huggingface.co/datasets/PKU-Alignment/PKU-SafeRLHF) dataset. |
| 24 | It can play a role in the safe RLHF algorithm, helping the Beaver model become more helpful. |
| 25 | |
| 26 | - **Developed by:** the [PKU-Alignment](https://github.com/PKU-Alignment) Team. |
| 27 | - **Model Type:** An auto-regressive language model based on the transformer architecture. |
| 28 | - **License:** Non-commercial license. |
| 29 | - **Fine-tuned from model:** [LLaMA](https://arxiv.org/abs/2302.13971), [Alpaca](https://github.com/tatsu-lab/stanford_alpaca). |
| 30 | |
| 31 | ## Model Sources |
| 32 | |
| 33 | - **Repository:** <https://github.com/PKU-Alignment/safe-rlhf> |
| 34 | - **Beaver:** <https://huggingface.co/PKU-Alignment/beaver-7b-v1.0> |
| 35 | - **Dataset:** <https://huggingface.co/datasets/PKU-Alignment/PKU-SafeRLHF> |
| 36 | - **Reward Model:** <https://huggingface.co/PKU-Alignment/beaver-7b-v1.0-reward> |
| 37 | - **Cost Model:** <https://huggingface.co/PKU-Alignment/beaver-7b-v1.0-cost> |
| 38 | - **Dataset Paper:** <https://arxiv.org/abs/2307.04657> |
| 39 | - **Paper:** <https://arxiv.org/abs/2310.12773> |
| 40 | |
| 41 | ## How to Use the Reward Model |
| 42 | |
| 43 | ```python |
| 44 | import torch |
| 45 | from transformers import AutoTokenizer |
| 46 | from safe_rlhf.models import AutoModelForScore |
| 47 | |
| 48 | model = AutoModelForScore.from_pretrained('PKU-Alignment/beaver-7b-v1.0-reward', torch_dtype=torch.bfloat16, device_map='auto') |
| 49 | tokenizer = AutoTokenizer.from_pretrained('PKU-Alignment/beaver-7b-v1.0-reward') |
| 50 | |
| 51 | input = 'BEGINNING OF CONVERSATION: USER: hello ASSISTANT:Hello! How can I help you today?' |
| 52 | |
| 53 | input_ids = tokenizer(input, return_tensors='pt') |
| 54 | output = model(**input_ids) |
| 55 | print(output) |
| 56 | |
| 57 | # ScoreModelOutput( |
| 58 | # scores=tensor([[[-19.7500], |
| 59 | # [-19.3750], |
| 60 | # [-20.1250], |
| 61 | # [-18.0000], |
| 62 | # [-20.0000], |
| 63 | # [-23.8750], |
| 64 | # [-23.5000], |
| 65 | # [-22.0000], |
| 66 | # [-21.0000], |
| 67 | # [-20.1250], |
| 68 | # [-23.7500], |
| 69 | # [-21.6250], |
| 70 | # [-21.7500], |
| 71 | # [-12.9375], |
| 72 | # [ -6.4375], |
| 73 | # [ -8.1250], |
| 74 | # [ -7.3438], |
| 75 | # [ -9.1875], |
| 76 | # [-13.6250], |
| 77 | # [-10.5625], |
| 78 | # [ -9.9375], |
| 79 | # [ -6.4375], |
| 80 | # [ -6.0938], |
| 81 | # [ -5.8438], |
| 82 | # [ -6.6562], |
| 83 | # [ -5.9688], |
| 84 | # [ -9.1875], |
| 85 | # [-11.4375]]], grad_fn=<ToCopyBackward0>), |
| 86 | # end_scores=tensor([[-11.4375]], grad_fn=<ToCopyBackward0>), |
| 87 | # last_hidden_state=tensor([[[ 0.7461, -0.6055, -0.4980, ..., 0.1670, 0.7812, -0.3242], |
| 88 | # [ 0.7383, -0.5391, -0.1836, ..., -0.1396, 0.5273, -0.2256], |
| 89 | # [ 0.6836, -0.7031, -0.3730, ..., 0.2100, 0.5000, -0.6328], |
| 90 | # ..., |
| 91 | # [-1.7969, 1.0234, 1.0234, ..., -0.8047, 0.2500, -0.8398], |
| 92 | # [ 2.0469, -1.3203, 0.8984, ..., -0.7734, -1.4141, -1.6797], |
| 93 | # [ 4.3438, -0.6953, 0.9648, ..., -0.1787, 0.6680, -3.0000]]], |
| 94 | # dtype=torch.bfloat16, grad_fn=<ToCopyBackward0>), |
| 95 | # end_last_hidden_state=tensor([[ 4.3438, -0.6953, 0.9648, ..., -0.1787, 0.6680, -3.0000]], |
| 96 | # dtype=torch.bfloat16, grad_fn=<ToCopyBackward0>), |
| 97 | # end_index=tensor([27]) |
| 98 | # ) |
| 99 | ``` |
| 100 | |