PKU SEC Lab

HybriMoE. [DAC'25] Official implement of "HybriMoE: Hybrid CPU-GPU Scheduling and Cache Management for Efficient MoE Inference"

Awesome-PPML-Papers. This repo summarizes papers for efficient PPML across protocol, model, and system levels.

AdapMoE. Code release for AdapMoE accepted by ICCAD 2024

DYSL_VLA. Dysl-vla：Official code for "DySL-VLA: Efficient Vision-Language-Action Model Inference via Dynamic-Static Layer-Skipping for Robot Manipulation". A framework that dynamically skips VLA layers based on action importance, achieving 3.75× speedup and 85.7× training reduction on robotic manipulation tasks.

LightMamba. Code release for LightMamba accepted by DATE 2025

ODB-dLLM. Implementation of "Orchestrating Dual-Boundaries: An Arithmetic Intensity Inspired Acceleration Framework for Diffusion Language Models"

mpcvit. Code release for MPCViT accepted by ICCV 2023

EfficientNav. Python

KEEP_Embodied_Memory. KEEP：Official code for "KEEP: A KV-Cache-Centric Memory Management System for Efficient Embodied Planning". A memory management system optimized for embodied planning via static-dynamic memory construction and KV cache re-computation, reducing memory overhead and improving inference speed.

ReaLM_DAC25. Code release for ReaLM accepted by DAC 25

FPQVAR. This is the official implementation of the algorithm part of "FPQVAR: Floating Point Quantization for Visual Autoregressive Model with FPGA Hardware Co-design"

CryptoMoE. Python

SPEX. Python

TEAM-MoE-dLLM. Initial Implementation of "TEAM: Temporal-Spatial Consistency Guided Expert Activation for MoE Diffusion Language Model Acceleration".

dac-sdc-2023-designs. DAC SDC 2023 Winner Designs

NeurIPS24-PrivCirNet. [NeurIPS'24] Official implement of "PrivCirNet: Efficient Private Inference via Block Circulant Transformation"

MPCache. Python

dac-sdc-2024. Python

ICCAD24-HG-PIPE. FPGA-based hardware accelerator for Vision Transformer (ViT), with Hybrid-Grained Pipeline.