Recurrent Memory Transformer implementation for Hugging Face models

RMT resources

[ paper ] [ code ] Recurrent Memory Transformer implementation and various training examples.

[ paper ] [ code ] BABILong - a long-context benchmark that supports 20 tasks and various sources of background text.

[ code ] Evaluate your long-context LLM on BABILong!

RMT is a memory-augmented segment-level recurrent Transformer. We implement our memory mechanism as a wrapper for any Hugging Face model by adding special memory tokens to the input sequence. The model is trained to control both memory operations and sequence representations processing.

Example: LM with RMT

Installation

pip install -e .

This command will install lm_experiments_tools with only required packages for Trainer and tools.

lm_experiments_tools Trainer supports gradient accumulation, logging to tensorboard, saving the best models based on metrics, custom metrics and data transformations support.

Install requirements for all experiments

Full requirements for all experiments are specified in requirements.txt. Install requirements after cloning the repo:

pip install -r requirements.txt

Citation

If you find our work useful, please consider citing the RMT papers:

@inproceedings{
        bulatov2022recurrent,
        title={Recurrent Memory Transformer},
        author={Aydar Bulatov and Yuri Kuratov and Mikhail Burtsev},
        booktitle={Advances in Neural Information Processing Systems},
        editor={Alice H. Oh and Alekh Agarwal and Danielle Belgrave and Kyunghyun Cho},
        year={2022},
        url={https://openreview.net/forum?id=Uynr3iPhksa}
}

@misc{bulatov2023scaling,
      title={Scaling Transformer to 1M tokens and beyond with RMT}, 
      author={Aydar Bulatov and Yuri Kuratov and Mikhail S. Burtsev},
      year={2023},
      eprint={2304.11062},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

@misc{kuratov2024search,
      title={In Search of Needles in a 11M Haystack: Recurrent Memory Finds What LLMs Miss}, 
      author={Yuri Kuratov and Aydar Bulatov and Petr Anokhin and Dmitry Sorokin and Artyom Sorokin and Mikhail Burtsev},
      year={2024},
      eprint={2402.10790},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}