Juan Miguel De Haro, Rubén Cano, et al.
IPDPS 2022
We introduce a novel run-time method for significantly reducing the accuracy loss associated with quantizing BERT-like models to 8-bit integers. Existing methods for quantizing models either modify the training procedure, or they require an additional calibration step to adjust parameters that also requires a selected held-out dataset. Our method permits taking advantage of quantization without the need for these adjustments. We present results on several NLP tasks demonstrating the usefulness of this technique.
Juan Miguel De Haro, Rubén Cano, et al.
IPDPS 2022
Jinghan Huang, Hyungyo Kim, et al.
MICRO 2025
Irem Boybat-Kara
IEDM 2023
Laura Bégon-Lours, Mattia Halter, et al.
MRS Spring Meeting 2023