hazyresearch/based-1b-50b

Model Card

This model is pretrained Based model.

As a quality reference, we include a pretrained Mamba model provided here: https://huggingface.co/hazyresearch/mamba-1b-50b and a pretrained attention (Llama architecture) model provided here: https://huggingface.co/hazyresearch/attn-1b-50bn

All three checkpoints are pretrained on 50Bn tokens of the Pile in the exact same data order using next token prediction.

A WandB report for training is here: https://api.wandb.ai/links/hazy-research/ggo9rst2

Model Sources

The model implementation and training code that produced the model are provided here: https://github.com/HazyResearch/based

Uses

The purpose of this work is to evaluate the language modeling quality of a new efficient architecture, Based.

We include a series of benchmarks that you can use to evaluate quality:

• FDA: https://huggingface.co/datasets/hazyresearch/based-fda
• SWDE: https://huggingface.co/datasets/hazyresearch/based-swde
• SQUAD: https://huggingface.co/datasets/hazyresearch/based-squad

Citation

Please consider citing this paper if you use our work:

@article{arora2024simple,
  title={Simple linear attention language models balance the recall-throughput tradeoff},
  author={Arora, Simran and Eyuboglu, Sabri and Zhang, Michael and Timalsina, Aman and Alberti, Silas and Zinsley, Dylan and Zou, James and Rudra, Atri and Ré, Christopher},
  journal={arXiv:2402.18668},
  year={2024}
}

Please reach out to simarora@stanford.edu, eyuboglu@stanford.edu, and mzhang20@stanford.edu with questions.