TERAIO speeds up LLM training by offloading inactive tensors to SSDs with a lifetime-aware, GPUDirect Storage based migration plan, reaching 1.47x over ZeRO-Offload/Infinity and 80.7% of ideal.
Efficient combination of rematerialization and offloading for training dnns.Advances in Neural Information Processing Systems, 34:23844–23857, 2021
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
baseline 1
citation-polarity summary
fields
cs.DC 1years
2025 1verdicts
CONDITIONAL 1roles
baseline 1polarities
baseline 1representative citing papers
citing papers explorer
-
Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage
TERAIO speeds up LLM training by offloading inactive tensors to SSDs with a lifetime-aware, GPUDirect Storage based migration plan, reaching 1.47x over ZeRO-Offload/Infinity and 80.7% of ideal.