HMI serves up to 10,000 tenant-specific BERT and GPT instances on one GPU by replacing lower-layer computation with trigram lookup tables and sharing upper layers with per-task adapters.
In: 2020 ACM/IEEE 47th An- nual International Symposium on Computer Architecture (ISCA), pp
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
HMI: Hierarchical Knowledge Management for Efficient Multi-Tenant Inference in Pretrained Language Models
HMI serves up to 10,000 tenant-specific BERT and GPT instances on one GPU by replacing lower-layer computation with trigram lookup tables and sharing upper layers with per-task adapters.