BWhat changed
Optimized tensor placement and access over GPU, CPU and disk with linear-program-guided offloading and compression for high-throughput, latency-insensitive language-model generation.
Scientific scope
Optimized tensor placement and access over GPU, CPU and disk with linear-program-guided offloading and compression for high-throughput, latency-insensitive language-model generation. Throughput is amortized over large batches and can involve high latency, CPU memory and disk resources; it is not interactive one-token-per-second service on a standalone small GPU.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Throughput is amortized over large batches and can involve high latency, CPU memory and disk resources; it is not interactive one-token-per-second service on a standalone small GPU.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
- FlexGen: High-Throughput Generative Inference of Large Language Models with a Single GPU
- Sheng et al.: FlexGen throughput-oriented heterogeneous inference — author/proceedings record
- High-throughput Generative Inference of Large Language Models with a Single GPU
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction