Endor: Hardware-Friendly Sparse Format for Offloaded LLM Inference.
Donghyeon JooRamyad HadidiSoheil FeiziBahar AsgariPublished in: CoRR (2024)
Keyphrases
- low cost
- hardware and software
- real time
- image processing
- databases
- inference process
- bayesian inference
- high dimensional
- multimedia
- probabilistic inference
- random fields
- bayesian networks
- computing power
- bayesian model
- vlsi implementation
- signal processing
- personal computer
- computing systems
- metadata
- grammatical inference
- single chip
- control program