Tulu 2 Dpo 7b

🚀 Tulu V2 DPO 7B

Tulu V2 DPO 7B is a fine - tuned language model, trained to be a helpful assistant. It offers a strong alternative to Llama 2 7b Chat.

🚀 Quick Start

Tulu is a series of language models designed to serve as helpful assistants. Tulu V2 DPO 7B is a fine - tuned version of Llama 2. It was trained on a mix of publicly available, synthetic, and human datasets using Direct Preference Optimization (DPO). This model is a great alternative to Llama 2 7b Chat.

For more details, read the paper: Camels in a Changing Climate: Enhancing LM Adaptation with Tulu 2.

✨ Features

Model description

Model Sources

• Repository: [https://github.com/allenai/open - instruct](https://github.com/allenai/open - instruct)
• DPO Recipe: The DPO recipe is from the [Zephyr Beta](https://huggingface.co/HuggingFaceH4/zephyr - 7b - beta) model
• Model Family: Other models and the dataset are found in the [Tulu V2 collection](https://huggingface.co/collections/allenai/tulu - v2 - suite - 6551b56e743e6349aab45101).

Performance

Input Format

The model is trained to use the following format (note the newlines):

<|user|>
Your message here!
<|assistant|>

⚠️ Important Note

For best results, format all inputs in this manner. Make sure to include a newline after <|assistant|>, as this can significantly affect generation quality.

Intended uses & limitations

The model was initially fine - tuned on a filtered and preprocessed [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. Then, it was further aligned with a Jax DPO trainer built on [EasyLM](https://github.com/young - geng/EasyLM) on the openbmb/UltraFeedback dataset, which contains 64k prompts and model completions ranked by GPT - 4.

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 used to train the base Llama 2 models were. 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 DPO training:

• learning_rate: 5e - 07
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 3.0

📚 Documentation

Citation

If you find Tulu 2 useful in your work, please cite it with:

@misc{ivison2023camels,
      title={Camels in a Changing Climate: Enhancing LM Adaptation with Tulu 2}, 
      author={Hamish Ivison and Yizhong Wang and Valentina Pyatkin and Nathan Lambert and Matthew Peters and Pradeep Dasigi and Joel Jang and David Wadden and Noah A. Smith and Iz Beltagy and Hannaneh Hajishirzi},
      year={2023},
      eprint={2311.10702},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

Model card adapted from [Zephyr Beta](https://huggingface.co/HuggingFaceH4/zephyr - 7b - beta/blob/main/README.md)