On unified-memory edge hardware, Ghidorah partitions Medusa-style speculative decoding across CPU and GPU with all-column weight splits, sparse ARM kernels, and profile-based tuning, reporting up to 7.6x decode speedup.
Specinfer: Accelerating large language model serving with tree-based speculative inference and ver- ification,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.DC 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Ghidorah: Fast LLM Inference on Edge with Speculative Decoding and Hetero-Core Parallelism
On unified-memory edge hardware, Ghidorah partitions Medusa-style speculative decoding across CPU and GPU with all-column weight splits, sparse ARM kernels, and profile-based tuning, reporting up to 7.6x decode speedup.