Flux Attention: Context-Aware Hybrid Attention for Efficient LLMs Inference
-
Updated
Apr 18, 2026 - Python
Flux Attention: Context-Aware Hybrid Attention for Efficient LLMs Inference
An simple web application that detects deepfake content in uploaded videos using AI-powered analysis with real-time monitoring and diagnostics.
Hybrid Mamba-2 + Transformer 2.94B LLM (Nemotron-H style) — Korean 3B model pretrained from scratch on 7× NVIDIA B200 GPUs with SFT + DPO alignment
A secure, fast, and browser-based file sharing system built using Python & Django. It supports direct peer-to-peer transfer and a backup Upload & Share mode, so file sharing always works no matter the network.
Site-to-Site VPN connection between AWS and on premises environment
From On-Premise to Cloud
Hybrid 1D+2D fusion for induction motor startup current fault diagnosis (reassigned spectrogram + FiLM)
Official implementation of "CoSwin: Convolution Enhanced Hierarchical Shifted Window Attention For Small-Scale Vision" (PyTorch version)
Hybrid Image Enhancement platform combining FastAPI and Streamlit for customizable multi-stage pipelines using deep learning. Supports flexible deblurring and dehazing workflows, user authentication, and output history. Easily extendable architecture with pre-trained model weights for advanced image restoration.
Hybrid data governance solution that integrates on-premise databases with Amazon DataZone using AWS CDK. Enables centralized data cataloging, access management, and automated notifications for unmanaged assets in hybrid cloud environments using events and a serverless solution.
A machine learning solution for classifying emotions in speech audio using hybrid deep learning (CNN-LSTM) and gradient boosting (XGBoost).
ESGP: A Hybrid ESN Architecture with Gated Dynamics
🎙️ Intelligent Discord bot that automatically records voice conversations in high-quality MP3. Hybrid Python/Node.js architecture with email notifications and cross-platform support
ZK Hybrid Architecture (ZHA) – A scalable, opinionated .NET 8 architecture that blends Clean Architecture, DDD, Modular Monolith, and Microservices. Designed to help developers start simple and gradually scale their applications into enterprise-ready distributed systems with Shared Kernel, API Gateway, and service isolation.
Pipeline híbrida de extração de dados com IA. Uma abordagem mista (Heurística + LLM) projetada para alta performance. Implementa arquitetura 'Heuristic-First' com padrão Watchdog cumulativo para garantir latência <10s e otimização de custo, mantendo acurácia >80% em layouts não estruturados.
A modular Google ADK agent project demonstrating a hybrid NLP architecture. It uses the speed and determinism of spaCy for structured entity extraction (NER) and an ADK LlmAgent (Gemini) for high-level reasoning and suggesting communication routing (e.g., Sales, Finance).
Hybrid vision backbone: VOLO-style Outlook local mixing + MaxViT grid attention + MBConv. Two variants (front-only vs per-block Outlook), FP16 training CLI with CutMix, and reproducible CIFAR-100 64×64 benchmarks.
Violence Detection
A novel hybrid architecture combining CNNs and Swin Transformers, to address limitations in 3D voxel reconstruction.
Implementation of GPT2 using ChebyKAN
Add a description, image, and links to the hybrid-architecture topic page so that developers can more easily learn about it.
To associate your repository with the hybrid-architecture topic, visit your repo's landing page and select "manage topics."