File size: 10,987 Bytes
37212c0 4d3fa46 37212c0 4d3fa46 37212c0 4d3fa46 37212c0 4d3fa46 37212c0 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 |
---
model-index:
- name: allenai/llama-3-tulu-v2.5-8b-uf-mean-70b-uf-rm
results: []
datasets:
- allenai/tulu-2.5-preference-data
- allenai/tulu-v2-sft-mixture
language:
- en
base_model: allenai/llama-3-tulu-2-8b
license: apache-2.0
---
<center>
<img src="https://huggingface.co/datasets/allenai/blog-images/resolve/main/tulu-2.5/tulu_25_banner.png" alt="Tulu 2.5 banner image" width="800px"/>
</center>
# Model Card for Llama 3 Tulu V2.5 PPO 13B - UltraFeedback Mean w. 70B UltraFeedback RM
Tulu is a series of language models that are trained to act as helpful assistants.
Tulu V2.5 is a series of models trained using DPO and PPO starting from the [Tulu 2 suite](https://huggingface.co/collections/allenai/tulu-v2-suite-6551b56e743e6349aab45101).
This model is trained on the UltraFeedback dataset (using the per-aspect/fine-grained scores for deciding chosen and rejected) using PPO.
We used a 8B RM trained on the UltraFeedback dataset, and then used the UltraFeedback prompts during PPO training.
This is part of a small update to the original V2.5 suite, adding some Llama 3-based models. We add three models:
- [allenai/llama-3-tulu-v2.5-8b-uf-mean-8b-uf-rm](https://huggingface.co/allenai/tulu-v2.5-llama3-8b-uf-mean-8b-uf-rm)
- [allenai/llama-3-tulu-v2.5-llama3-8b-uf-mean-70b-uf-rm-mixed-prompts](https://huggingface.co/allenai/tulu-v2.5-llama3-8b-uf-mean-70b-uf-rm-mixed-prompts)
- [allenai/llama-3-tulu-v2.5-llama3-8b-uf-mean-70b-uf-rm](https://huggingface.co/allenai/tulu-v2.5-llama3-8b-uf-mean-70b-uf-rm-mixed-prompts) (best overall model, this model)
For more details, read the paper:
[Unpacking DPO and PPO: Disentangling Best Practices for Learning from Preference Feedback](https://arxiv.org/abs/2406.09279).
Built with Meta Llama 3! Note that Llama 3 is released under the Meta Llama 3 community license, included here under llama_3_license.txt.
## .Model description
- **Model type:** One model belonging to a suite of RLHF tuned chat models on a mix of publicly available, synthetic and human-created datasets.
- **Language(s) (NLP):** English
- **License:** Apache 2.0.
- **Finetuned from model:** [meta-llama/Llama-2-13b-hf](https://huggingface.co/meta-llama/Llama-2-13b-hf)
### Model Sources
- **Repository:** https://github.com/allenai/open-instruct
- **Dataset:** Data used to train this model can be found [here](https://huggingface.co/datasets/allenai/tulu-2.5-preference-data) - specifically the `ultrafeedback_mean_aspects` split. Only the prompts were used.
- **Model Family:** The collection of related models can be found [here](https://huggingface.co/collections/allenai/tulu-v25-suite-66676520fd578080e126f618).
- **Reward Model:** The reward model used during PPO training can be found [here](https://huggingface.co/allenai/llama-3-tulu-2-8b-uf-mean-rm), and the data used to train it [here](https://huggingface.co/datasets/allenai/tulu-2.5-preference-data) - specifically the `ultrafeedback_mean_aspects` split.
## Results
This is a model trained on Llama 3 as an update to the Tulu v2.5 suite.
For details on training and evaluation, read [our paper](https://arxiv.org/abs/2406.09279)!
| Model | Size | Alignment | GSM8k 8-shot CoT Acc. | AlpacaEval 2 Winrate (LC) |
|-|-|-|-|-|
| **Tulu V2.5 PPO Llama 3 70B (this model)** | 8B | PPO with 8B RM | 48.5 | **28.8** |
| **Tulu V2.5 PPO 13B** | 13B | PPO with 70B RM | 67.0 | 26.7 |
| **Tulu V2 DPO 13B** | 13B | DPO | 50.5 | 16.0 |
| **Tulu V2 SFT 13B** | 13B | - | 46.0 | 10.4 |
| **Tulu V2 DPO 70B** | 70B | DPO | **71.5** | 21.2 |
## Input Format
The model is trained to use the following format (note the newlines):
```
<|user|>
Your message here!
<|assistant|>
```
For best results, format all inputs in this manner. **Make sure to include a newline after `<|assistant|>`, this can affect generation quality quite a bit.**
We have included a [chat template](https://huggingface.co/docs/transformers/main/en/chat_templating) in the tokenizer implementing this template.
## Model Family
[Preference Data](https://huggingface.co/datasets/allenai/tulu-2.5-preference-data), [Prompts Data](https://huggingface.co/datasets/allenai/tulu-2.5-prompts) | DPO Models | PPO Models | Reward Models | Value Models |
|-------------|-------------|-------------|---------------|---------------|
| ultrafeedback_mean_aspects | [tulu-v2.5-dpo-13b-uf-mean](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-uf-mean) | [tulu-v2.5-ppo-13b-uf-mean-70b-uf-rm](https://huggingface.co/allenai/tulu-v2.5-ppo-13b-uf-mean-70b-uf-rm) | [tulu-v2.5-70b-uf-rm](https://huggingface.co/allenai/tulu-v2.5-70b-uf-rm) | [tulu-v2.5-ppo-13b-uf-mean-70b-uf-rm-value](https://huggingface.co/allenai/tulu-v2.5-ppo-13b-uf-mean-70b-uf-rm-value) |
| preference_big_mixture | = | [tulu-v2.5-ppo-13b-uf-mean-13b-mix-rm](https://huggingface.co/allenai/tulu-v2.5-ppo-13b-uf-mean-13b-mix-rm) | [tulu-v2.5-13b-preference-mix-rm](https://huggingface.co/allenai/tulu-v2.5-13b-preference-mix-rm) | [tulu-v2.5-ppo-13b-uf-mean-13b-mix-rm-value](https://huggingface.co/allenai/tulu-v2.5-ppo-13b-uf-mean-13b-mix-rm-value) |
| preference_big_mixture | = | [tulu-v2.5-ppo-13b-uf-mean-70b-mix-rm](https://huggingface.co/allenai/tulu-v2.5-ppo-13b-uf-mean-70b-mix-rm) | [tulu-v2.5-70b-preference-mix-rm](https://huggingface.co/allenai/tulu-v2.5-70b-preference-mix-rm) | [tulu-v2.5-ppo-13b-uf-mean-70b-mix-rm-value](https://huggingface.co/allenai/tulu-v2.5-ppo-13b-uf-mean-70b-mix-rm-value) |
| ultrafeedback_mean_aspects | = | [tulu-v2.5-ppo-13b-uf-mean](https://huggingface.co/allenai/tulu-v2.5-ppo-13b-uf-mean) | [tulu-v2.5-13b-uf-rm](https://huggingface.co/allenai/tulu-v2.5-13b-uf-rm) | [tulu-v2.5-ppo-13b-uf-mean-13b-uf-rm-value](https://huggingface.co/allenai/tulu-v2.5-ppo-13b-uf-mean-13b-uf-rm-value) |
| ultrafeedback_mean_aspects | = | [tulu-v2.5-ppo-13b-uf-mean-70b-uf-rm-mixed-prompts](https://huggingface.co/allenai/tulu-v2.5-ppo-13b-uf-mean-70b-uf-rm-mixed-prompts) | [tulu-v2.5-70b-uf-rm](https://huggingface.co/allenai/tulu-v2.5-70b-uf-rm) * with extra prompts | [tulu-v2.5-ppo-13b-uf-mean-70b-uf-rm-mixed-prompts-value](https://huggingface.co/allenai/tulu-v2.5-ppo-13b-uf-mean-70b-uf-rm-mixed-prompts-value) |
| hh_rlhf_60k | [tulu-v2.5-dpo-13b-hh-rlhf-60k](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-hh-rlhf-60k) | [tulu-v2.5-ppo-13b-hh-rlhf-60k](https://huggingface.co/allenai/tulu-v2.5-ppo-13b-hh-rlhf-60k) | [tulu-v2.5-13b-hh-rlhf-60k-rm](https://huggingface.co/allenai/tulu-v2.5-13b-hh-rlhf-60k-rm) | |
| chatbot_arena_2023 | [tulu-v2.5-dpo-13b-chatbot-arena-2023](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-chatbot-arena-2023) | [tulu-v2.5-ppo-13b-chatbot-arena-2023](https://huggingface.co/allenai/tulu-v2.5-ppo-13b-chatbot-arena-2023) | [tulu-v2.5-13b-chatbot-arena-2023-rm](https://huggingface.co/allenai/tulu-v2.5-13b-chatbot-arena-2023-rm) | |
| stack_exchange_60k | [tulu-v2.5-dpo-13b-stackexchange-60k](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-stackexchange-60k) | [tulu-v2.5-ppo-13b-stackexchange-60k](https://huggingface.co/allenai/tulu-v2.5-ppo-13b-stackexchange-60k) | [tulu-v2.5-13b-stackexchange-60k-rm](https://huggingface.co/allenai/tulu-v2.5-13b-stackexchange-60k-rm) | |
| nectar_60k | N/A | [tulu-v2.5-ppo-13b-nectar-60k](https://huggingface.co/allenai/tulu-v2.5-ppo-13b-nectar-60k) | [tulu-v2.5-13b-nectar-60k-rm](https://huggingface.co/allenai/tulu-v2.5-13b-nectar-60k-rm) | |
| nectar | [tulu-v2.5-dpo-13b-nectar](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-nectar) | | | |
| helpsteer | [tulu-v2.5-dpo-13b-helpsteer](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-helpsteer) | | | |
| shp2 | [tulu-v2.5-dpo-13b-shp2](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-shp2) | | | |
| stack_exchange_paired | [tulu-v2.5-dpo-13b-stackexchange](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-stackexchange) | | | |
| ultrafeedback_overall | [tulu-v2.5-dpo-13b-uf-overall](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-uf-overall) | | | |
| capybara | [tulu-v2.5-dpo-13b-capybara](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-capybara) | | | |
| prm800k_pairs_phase2 | [tulu-v2.5-dpo-13b-prm-phase-2](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-prm-phase-2) | | | |
| hh_rlhf | [tulu-v2.5-dpo-13b-hh-rlhf](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-hh-rlhf) | | | |
| chatbot_arena_2024 | [tulu-v2.5-dpo-13b-chatbot-arena-2024](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-chatbot-arena-2024) | | | |
| alpaca_farm_human_pref | [tulu-v2.5-dpo-13b-alpacafarm-human-pref](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-alpacafarm-human-pref) | | | |
| alpaca_farm_gpt4_pref | [tulu-v2.5-dpo-13b-alpacafarm-gpt4-pref](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-alpacafarm-gpt4-pref) | | | |
| orca_dpo_pairs | [tulu-v2.5-dpo-13b-argilla-orca-pairs](https://huggingface.co/allenai/tulu-v2.5-dpo-13b-argilla-orca-pairs) | | | |
*The extra prompts are all the prompts in the prompts dataset. Default only uses the split `ultrafeedback_prompts`.
## Intended uses & limitations
The model was initially fine-tuned on a filtered and preprocessed of the [Tulu V2 mix dataset](https://huggingface.co/datasets/allenai/tulu-v2-sft-mixture), which contains a diverse range of human created instructions and synthetic dialogues generated primarily by other LLMs.
We then further aligned the model with a [Jax DPO trainer](https://github.com/hamishivi/EasyLM/blob/main/EasyLM/models/llama/llama_train_dpo.py) built on [EasyLM](https://github.com/young-geng/EasyLM) on the dataset mentioned above.
## Bias, Risks, and Limitations
The Tulu models have not been aligned to generate safe completions within the RLHF phase or deployed with in-the-loop filtering of responses like ChatGPT, so the model can produce problematic outputs (especially when prompted to do so).
It is also unknown what the size and composition of the corpus was used to train the base Llama 2 models, however it is likely to have included a mix of Web data and technical sources like books and code. See the [Falcon 180B model card](https://huggingface.co/tiiuae/falcon-180B#training-data) for an example of this.
### Training hyperparameters
The following hyperparameters were used during PPO training:
- learning_rate: 1e-06
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 1.0
- KL penalty coefficient: 0.05
## Citation
If you find Tulu 2.5 is useful in your work, please cite it with:
```
@misc{ivison2024unpacking,
title={{Unpacking DPO and PPO: Disentangling Best Practices for Learning from Preference Feedback}},
author={{Hamish Ivison and Yizhong Wang and Jiacheng Liu and Ellen Wu and Valentina Pyatkin and Nathan Lambert and Yejin Choi and Noah A. Smith and Hannaneh Hajishirzi}}
year={2024},
eprint={2406.09279},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
``` |