mirror of
https://github.com/hanzoai/papers.git
synced 2026-11-03 02:27:30 +00:00
19 lines
762 B
TeX
19 lines
762 B
TeX
\section{Related Work}
|
|
|
|
\label{sec:related}
|
|
% ============================================================================
|
|
|
|
\textbf{CrypTen}~\cite{crypten} and \textbf{TenSEAL}~\cite{tenseal} provide
|
|
FHE wrappers for PyTorch tensors but lack GPU acceleration and production
|
|
serving infrastructure.
|
|
|
|
\textbf{Concrete ML}~\cite{concreteml} by Zama compiles scikit-learn models to
|
|
FHE circuits using TFHE. Our system extends this approach with CKKS support for
|
|
neural networks and GPU acceleration via Candle.
|
|
|
|
\textbf{Jin}~\cite{hanzojin} handles multimodal routing for Hanzo models; FHE
|
|
inference integrates as an optional encrypted execution backend for
|
|
privacy-sensitive requests.
|
|
|
|
% ============================================================================
|