Abstract
Deterministic data loading is important for foundation model development: model researchers need confidence that differences they observe across costly ablations are caused by the parameter they changed rather than non-determinism in the training data sequence. The data loader must provide elastic determinism, i.e., a deterministic sequence of global training data batches despite changes to the GPU topology across runs (e.g., due to GPU scarcity), frequent checkpoint-resume cycles, and different data processing execution backends. Achieving this is difficult because modern foundation model data pipelines tokenize, pack, and mix samples online, introducing stateful n-to-m transformations that break sample indexing. Existing data loaders largely assume indexable 1-to-1 pipelines, and the common workaround of offline materialization is expensive and, for some modalities such as video, infeasible. We present Zephon, a data loader for foundation models that supports online, stateful pipelines while providing elastic determinism and efficient resumption from checkpoints. It partitions the global stream into topology-independent lanes, serializes ordering decisions while parallelizing stateless work on interchangeable backends, and checkpoints only bounded in-flight state so recovery cost does not grow with training progress. We evaluate Zephon on text and vision-language workloads and show that it achieves competitive throughput while providing a combination of guarantees that no existing loader offers for online, stateful pipelines.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Böther, M., Wills, J., Robroek, T., Xu, S., Burstein, P., Zayas, D., Blakeney, C., Joshi, S., Yin, H., Adiga, R., Mongstad, H., Merrick, L., Maini, P., Morcos, A., Leavitt, M., Klimovic, A., & Gaza, B. (2026). Zephon: Elastic Determinism for Online, Stateful Foundation Model Data Loading Pipelines. https://omanscience.com/en/articles/zephon-elastic-determinism-for-online-stateful-foundation-model-data-loading-pipelines
MLA 9
Böther, Maximilian, et al. "Zephon: Elastic Determinism for Online, Stateful Foundation Model Data Loading Pipelines." https://omanscience.com/en/articles/zephon-elastic-determinism-for-online-stateful-foundation-model-data-loading-pipelines.
Chicago (author–date)
Böther, Maximilian, Josh Wills, Ties Robroek, Sonnet Xu, Paul Burstein, Daniel Zayas, Cody Blakeney, Siddharth Joshi, Haoli Yin, Rishabh Adiga, Haakon Mongstad, Luke Merrick, Pratyush Maini, Ari Morcos, Matthew Leavitt, Ana Klimovic, and Bogdan Gaza. 2026. "Zephon: Elastic Determinism for Online, Stateful Foundation Model Data Loading Pipelines." https://omanscience.com/en/articles/zephon-elastic-determinism-for-online-stateful-foundation-model-data-loading-pipelines.
Harvard
Böther, M., Wills, J., Robroek, T., Xu, S., Burstein, P., Zayas, D., Blakeney, C., Joshi, S., Yin, H., Adiga, R., Mongstad, H., Merrick, L., Maini, P., Morcos, A., Leavitt, M., Klimovic, A. and Gaza, B. (2026) 'Zephon: Elastic Determinism for Online, Stateful Foundation Model Data Loading Pipelines', Available at: https://omanscience.com/en/articles/zephon-elastic-determinism-for-online-stateful-foundation-model-data-loading-pipelines.
Vancouver
Böther M, Wills J, Robroek T, Xu S, Burstein P, Zayas D, et al. Zephon: Elastic Determinism for Online, Stateful Foundation Model Data Loading Pipelines. https://omanscience.com/en/articles/zephon-elastic-determinism-for-online-stateful-foundation-model-data-loading-pipelines
IEEE
M. Böther, J. Wills, T. Robroek, S. Xu, P. Burstein, D. Zayas, C. Blakeney, S. Joshi, H. Yin, R. Adiga, H. Mongstad, L. Merrick, P. Maini, A. Morcos, M. Leavitt, A. Klimovic, and B. Gaza, "Zephon: Elastic Determinism for Online, Stateful Foundation Model Data Loading Pipelines," https://omanscience.com/en/articles/zephon-elastic-determinism-for-online-stateful-foundation-model-data-loading-pipelines.