Search Results for "network architecture"
Sort By:
Showing 56 open source projects for "network architecture"
View related business solutions-
Streamline Azure Security with Palo Alto Networks VM-SeriesImprove your security posture and reduce incident response time. Use the VM-Series to natively analyze Azure traffic and dynamically drive policy updates based on workload changes.
-
Build Agents and Models on One PlatformGemini Enterprise Agent Platform is Google Cloud's comprehensive platform for developers to build, scale, govern, and optimize agents and models. Choose from Google's most advanced models and third-party models like Anthropic's Claude Model Family.
-
1
xLSTM
Neural Network architecture based on ideas of the original LSTM
xLSTM is an open-source machine learning architecture that reimagines the classic Long Short-Term Memory (LSTM) network for modern large-scale language modeling and sequence processing tasks. The project introduces a new recurrent neural network design that incorporates exponential gating mechanisms and enhanced memory structures to overcome limitations of traditional LSTM models. -
2
BitNet
BitNet: Scaling 1-bit Transformers for Large Language Models
BitNet is a machine learning research implementation that explores extremely low-precision neural network architectures designed to dramatically reduce the computational cost of large language models. The project implements the BitNet architecture described in research on scaling transformer models using extremely low-bit quantization techniques. In this approach, neural network weights are quantized to approximately one bit per parameter, allowing models to operate with far lower memory usage than traditional 16-bit or 32-bit neural networks. ... -
3
MatMul-Free LM
Implementation for MatMul-free LM
MatMul-Free LM is an experimental implementation of a large language model architecture designed to eliminate traditional matrix multiplication operations used in transformer networks. Since matrix multiplication is one of the most computationally expensive components of modern language models, the project explores alternative computational strategies that reduce hardware requirements while maintaining comparable performance. The architecture relies on quantization-aware training and... -
4
graphify
AI coding assistant skill (Claude Code, Codex, OpenCode, OpenClaw)
...It focuses on building visual models such as nodes and edges that represent entities and their connections, making complex datasets easier to interpret. The system likely supports dynamic updates, allowing graphs to evolve as data changes or new inputs are introduced. It is particularly useful in domains such as network analysis, knowledge graphs, and system architecture visualization. The architecture emphasizes flexibility, enabling users to customize how data is mapped and displayed. It may also include analytical features to explore patterns, clusters, or anomalies within the graph. Overall, graphify serves as a bridge between raw data and visual insight. -
MongoDB Atlas runs apps anywhereMongoDB Atlas gives you the freedom to build and run modern applications anywhere—across AWS, Azure, and Google Cloud. With global availability in over 115 regions, Atlas lets you deploy close to your users, meet compliance needs, and scale with confidence across any geography.
-
5
MuseGAN
An AI for Music Generation
...Instead of generating raw audio, the model operates on piano-roll representations of music, which encode notes as time-pitch matrices for each instrument track. This representation allows the neural network to capture rhythmic patterns, harmonic relationships, and structural dependencies across instruments. The architecture is based on convolutional GAN models that learn temporal musical structure and inter-track relationships from training data. The project was trained using the Lakh Pianoroll Dataset, a large collection of multitrack musical sequences derived from MIDI files. -
6
Cactus Needle
26m function call model that runs on incredibly small devices
Needle is an experimental 26-million-parameter function-calling model designed to run on extremely small devices such as phones, watches, glasses, and low-power personal AI hardware. It is based on a Simple Attention Network architecture and was distilled from a much larger model to focus on fast, compact tool-use behavior. The project provides open weights, training details, dataset generation resources, and a playground for testing the model with custom tools. Needle is optimized for single-shot function calling rather than broad conversational ability, so its core use case is selecting the right tool and producing structured arguments. ... -
7
MiroFlow
Agent framework that enables tool-use agent tasks
...This architecture allows agents to perform advanced reasoning tasks such as deep research, future event prediction, and multi-step knowledge analysis. The framework emphasizes reliability and scalability by incorporating robust workflow execution, concurrency management, and fault-tolerant design to handle unstable APIs or network conditions. -
8
Umi-OCR
OCR software, free and offline
Umi-OCR is a free and open-source optical character recognition (OCR) tool designed to provide fast, offline text extraction from images, screenshots, PDFs, and more without requiring a network connection. It includes a highly efficient offline OCR engine with built-in multilingual recognition libraries, so users can extract text across multiple languages with high accuracy directly on their machines. The software supports flexible usage patterns including screenshot capture OCR, batch... -
9
SlowFast
Video understanding codebase from FAIR for reproducing video models
...The slow pathway encodes semantic context by sampling frames sparsely, while the fast pathway captures motion and fine temporal cues by operating on densely sampled frames with fewer channels. Together, these two pathways complement each other, allowing the network to model both appearance and motion without excessive computational cost. The architecture is modular and supports tasks like action recognition, temporal localization, and video segmentation, performing strongly on benchmarks like Kinetics and AVA. The repository provides training recipes, pretrained models, and distributed pipelines optimized for large-scale video datasets. -
Enterprise-grade ITSM, for every businessFreshservice is an intuitive, AI-powered platform that helps IT, operations, and business teams deliver exceptional service without the usual complexity. Automate repetitive tasks, resolve issues faster, and provide seamless support across the organization. From managing incidents and assets to driving smarter decisions, Freshservice makes it easy to stay efficient and scale with confidence.
-
10
HunyuanCustom
Multimodal-Driven Architecture for Customized Video Generation
HunyuanCustom is a multimodal video customization framework by Tencent Hunyuan, aimed at generating customized videos featuring particular subjects (people, characters) under flexible conditions, while maintaining subject/identity consistency. It supports conditioning via image, audio, video, and text, and can perform subject replacement in videos, generate avatars speaking given audio, or combine multiple subject images. The architecture builds on HunyuanVideo, with added modules for... -
11
Watermark-Removal
Machine learning image inpainting task that removes watermarks
Watermark-Removal repository is a machine learning project focused on removing visible watermarks from digital images using deep learning and image inpainting techniques. The system analyzes an image containing a watermark and attempts to reconstruct the underlying visual content so that the watermark is removed while preserving the original appearance of the image. The project uses neural network models inspired by research in contextual attention and gated convolution, which are methods... -
12
Agents 2.0
An Open-source Framework for Data-centric Language Agents
Agents is an open-source framework designed to build and train autonomous language agents through a data-centric and learning-oriented architecture. The project introduces a concept known as agent symbolic learning, which treats an agent pipeline similarly to a neural network computational graph. In this framework, each node in the pipeline represents a step in the reasoning or action process, while prompts and tools act as adjustable parameters analogous to neural network weights. ... -
13
deep-q-learning
Minimal Deep Q Learning (DQN & DDQN) implementations in Keras
...Because it’s self-contained and Python-based, it's well-suited for experimentation, modifications, or extension — for instance adapting to custom Gym environments, tweaking network architecture, or combining with other RL techniques. -
14
TurboQuant+
Implementation of TurboQuant (ICLR 2026)
...It is designed to be used in conjunction with modern machine learning workflows, particularly those involving large models that require optimization for deployment. TurboQuant Plus focuses on experimentation and performance tuning, allowing developers to test different configurations and evaluate trade-offs. Its architecture supports extensibility, enabling further development of quantization methods and integration with existing ML pipelines. -
15
MobileLLM
MobileLLM Optimizing Sub-billion Parameter Language Models
MobileLLM is a lightweight large language model (LLM) framework developed by Facebook Research, optimized for on-device deployment where computational and memory efficiency are critical. Introduced in the ICML 2024 paper “MobileLLM: Optimizing Sub-billion Parameter Language Models for On-Device Use Cases”, it focuses on delivering strong reasoning and generalization capabilities in models under one billion parameters. The framework integrates several architectural innovations—SwiGLU... -
16
deepjazz
Deep learning driven jazz generation using Keras & Theano
deepjazz is a deep learning project that generates jazz music using recurrent neural networks trained on MIDI files. The repository demonstrates how machine learning can learn musical structure and produce original compositions. It uses the Keras and Theano libraries to build a two-layer Long Short-Term Memory network capable of learning temporal patterns in music. The system analyzes musical sequences from an input MIDI file and then generates new musical notes that follow similar stylistic... -
17
Harbor LLM
Run a full local LLM stack with one command using Docker
Harbor is an open source, containerized toolkit designed to simplify running local large language model (LLM) environments. It combines a CLI and companion app to launch backends, frontends, and supporting services with minimal setup. With a single command, users can start preconfigured tools like Ollama and Open WebUI, enabling chat, workflows, and integrations immediately. Harbor supports multiple inference engines, including llama.cpp and vLLM, and connects them seamlessly to user... -
18
pix2pixHD
Synthesizing and manipulating 2048x1024 images with conditional GANs
pix2pixHD is a PyTorch-based implementation of a conditional generative adversarial network designed for high-resolution image-to-image translation, capable of producing photorealistic outputs at resolutions up to 2048×1024. It is widely used to convert structured inputs such as semantic label maps into realistic images, making it particularly valuable in applications like autonomous driving simulation, face synthesis, and scene generation. The model improves upon earlier GAN approaches by... -
19
PaSa
An advanced paper search agent powered by large language models
PaSa is an open-source “paper search agent” built around large language models (LLMs), designed to automate the process of academic literature retrieval with human-like decision making. Instead of simply translating a query into keywords and returning a flat list of matching papers, PaSa uses a dual-agent architecture (Crawler + Selector) that can iteratively search, read, analyze, and filter academic publications — simulating how a researcher might dig through citation networks, expand... -
20
Materials Discovery: GNoME
AI discovers 520000 stable inorganic crystal structures for research
Materials Discovery (GNoME) is a large-scale research initiative by Google DeepMind focused on applying graph neural networks to accelerate the discovery of stable inorganic crystal materials. The project centers on Graph Networks for Materials Exploration (GNoME), a message-passing neural network architecture trained on density functional theory (DFT) data to predict material stability and energy formation. Using GNoME, DeepMind identified 381,000 new stable materials, later expanding the dataset to include over 520,000 materials within 1 meV/atom of the convex hull as of August 2024. The repository provides datasets, model definitions, and interactive Colabs for exploring these materials, computing decomposition energies, and visualizing chemical families. ... -
21
Neural Network Intelligence
AutoML toolkit for automate machine learning lifecycle
Neural Network Intelligence is an open source AutoML toolkit for automate machine learning lifecycle, including feature engineering, neural architecture search, model compression and hyper-parameter tuning. NNI (Neural Network Intelligence) is a lightweight but powerful toolkit to help users automate feature engineering, neural architecture search, hyperparameter tuning and model compression. -
22
SoftVC VITS Singing Voice Conversion
SoftVC VITS Singing Voice Conversion
SoftVC VITS Singing Voice Conversion is a deep learning project focused on singing voice conversion, allowing users to transform one voice into another while preserving melody and timing. Unlike traditional text-to-speech systems, it specializes specifically in singing scenarios and does not provide general TTS functionality. The project leverages neural network architectures derived from VITS and SoftVC research to achieve high-quality voice transformation. It is commonly used in creative... -
23
ControlNet
Let us control diffusion models
ControlNet is a neural network architecture designed to add conditional control to text-to-image diffusion models. Rather than training from scratch, ControlNet “locks” the weights of a pre-trained diffusion model and introduces a parallel trainable branch that learns additional conditions—like edges, depth maps, segmentation, human pose, scribbles, or other guidance signals. -
24
picoGPT
An unnecessarily tiny implementation of GPT-2 in NumPy
picoGPT is a minimal implementation of the GPT-2 language model designed to demonstrate how transformer-based language models work at a conceptual level. The repository focuses on educational clarity rather than production performance, implementing the core components of the GPT architecture in a concise and readable way. It allows users to understand how tokenization, transformer layers, attention mechanisms, and autoregressive text generation operate in modern large language models. The project uses a small amount of code to illustrate the essential mathematical operations involved in training and running a transformer-based neural network. -
25
ConvNeXt V2
Code release for ConvNeXt V2 model
ConvNeXt V2 is an evolution of the ConvNeXt architecture that co-designs convolutional networks alongside self-supervised learning. The V2 version introduces a fully convolutional masked autoencoder (FCMAE) framework where parts of the image are masked and the network reconstructs the missing content, marrying convolutional inductive bias with powerful pretraining. A key innovation is a new Global Response Normalization (GRN) layer added to the ConvNeXt backbone, which enhances feature competition across channels. ...