@inproceedings{oneill-dutta-2023-self,
    title = "Self-Distilled Quantization: Achieving High Compression Rates in Transformer-Based Language Models",
    author = "O{'}Neill, James  and
      Dutta, Sourav",
    editor = "Rogers, Anna  and
      Boyd-Graber, Jordan  and
      Okazaki, Naoaki",
    booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)",
    month = jul,
    year = "2023",
    address = "Toronto, Canada",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/fix-sig-urls/2023.acl-short.114/",
    doi = "10.18653/v1/2023.acl-short.114",
    pages = "1329--1339",
    abstract = "We investigate the effects of post-training quantization and quantization-aware training on the generalization of Transformer language models. We present a new method called self-distilled quantization (SDQ) that minimizes accumulative quantization errors and outperforms baselines. We apply SDQ to multilingual models XLM-R$_{\text{Base}}$ and InfoXLM$_{\text{Base}}$ and demonstrate that both models can be reduced from 32-bit floating point weights to 8-bit integer weights while maintaining a high level of performance on the XGLUE benchmark. Our results also highlight the challenges of quantizing multilingual models, which must generalize to languages they were not fine-tuned on."
}