Publications of Torsten Hoefler

Qiao Xiao, Boqian Wu, Patrik Okanovic, Tomasz Sternal, Maurice van Keulen, Elena Mocanu, Mykola Pechenizkiy, Decebal Constantin Mocanu, Torsten Hoefler:

Memory-Efficient LLM Training with Dynamic Sparsity: From Stability to Practical Scaling

(In Proceedings of the 43rd International Conference on Machine Learning, Jul. 2026, arXiv:2606.00888. )

Abstract

Dynamic Sparse Training (DST) offers a promising paradigm for improving the training and inference efficiency of deep neural networks; however, we find that in large language model training, DST can suffer from optimization instability, manifested as loss spikes after topology updates. In this work, we show that the naive use of standard Adam-based optimizers leads to a cold-start issue for newly regrown parameters, resulting in excessively large updates and disrupted training dynamics. To address this issue, we propose Sparse Memory-Efficient Training (SMET), which stabilizes DST with optimizer warm-up and improves training progress through density-aware learning-rate scaling. SMET further reduces memory consumption by storing gradients and optimizer states only for active parameters. We provide a theoretical analysis of the update behaviors under SMET, showing improved optimization stability. Extensive experiments demonstrate that SMET enables stable, scalable, and memory-efficient sparse pre-training of LLMs, paving the way for sparse training as a practical alternative to dense training. Our code is publicly available at: this https URL.

Documents

download article:

download slides:

BibTeX

@inproceedings{,
  author={Qiao Xiao and Boqian Wu and Patrik Okanovic and Tomasz Sternal and Maurice van Keulen and Elena Mocanu and Mykola Pechenizkiy and Decebal Constantin Mocanu and Torsten Hoefler},
  title={{Memory-Efficient LLM Training with Dynamic Sparsity: From Stability to Practical Scaling}},
  year={2026},
  month={Jul.},
  booktitle={Proceedings of the 43rd International Conference on Machine Learning},
  note={arXiv:2606.00888.},
  source={http://www.unixer.de/~htor/publications/},
}


serving: 216.73.216.232:17731	© Torsten Hoefler