Downloads · 30 days
156
10% of all-time downloads
allenai/tulu-v2.5-ppo-13b-uf-mean
tulu-v2.5-ppo-13b-uf-mean is a text generation model from allenai. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as apache-2.0.
<center <img src="https://huggingface.co/datasets/allenai/blog-images/resolve/main/tulu-2.5/tulu25banner.png" alt="Tulu 2.5 banner image" width="800px"/ </center
Downloads · 30 days
156
10% of all-time downloads
All-time downloads
1.5K
Public
Parameters
13B
26 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors26 GB · 100%
From the Hugging Face model README
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. 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 13B RM trained on the UltraFeedback data, and then re-used the same prompts during PPO training.
For more details, read the paper: Unpacking DPO and PPO: Disentangling Best Practices for Learning from Preference Feedback.
ultrafeedback_mean_aspects split.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 in the tokenizer implementing this template.
The model was initially fine-tuned on a filtered and preprocessed of the Tulu V2 mix dataset, 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 PPO trainer built on EasyLM on the dataset mentioned above.
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 for an example of this.
The following hyperparameters were used during PPO training:
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}
}