Tensor parallelism is all you need. Run LLMs on weak devices or make powerful devices even more powerful by distributing the workload and dividing the RAM usage.

A tool for cell instance aware segmentation in densely packed 3D volumetric images

A library capable of handling complexed valued neural network alternatives built with Rust.

Project repository for the discipline "Social Network Analysis" (HSE 2024)

Tensors and Dynamic neural networks in Python with strong GPU acceleration

Low-code framework for building custom LLMs, neural networks, and other AI models

Open standard for machine learning interoperability

NeuroMelNet - End-to-End binary classification model for AI speech detection with PyTorch implementation. Python API & Tg-bot API

Neural Turing Machine for a Multi-Processor System on Chip verified with UVM/OSVVM/FV

Neural Turing Machine for a System on Chip verified with UVM/OSVVM/FV

Neural Turing Machine for a Processing Unit verified with UVM/OSVVM/FV

🦋 A personal research and development (R&D) lab that facilitates the sharing of knowledge.

Sandbox for graphics paper implementation

Qdrant - High-performance, massive-scale Vector Database for the next generation of AI. Also available in the cloud https://cloud.qdrant.io/

A repo designed to convert audio-based "weak" labels to "strong" intraclip labels. Provides a pipeline to compare automated moment-to-moment labels to human labels. Methods range from DSP based foreground-background separation, cross-correlation based template matching, as well as bird presence sound event detection deep learning models!

Awesome Time-Series Imputation Papers, including a must-read paper list about using deep learning neural networks to impute incomplete time series containing NaN missing values/data

neural network framework in c

The Credit Card Fraud Detection Problem includes modeling past credit card transactions with the knowledge of the ones that turned out to be a fraud. This model is then used to identify whether a new transaction is fraudulent or not. Our aim here is to detect 100% of the fraudulent transactions while minimizing the incorrect fraud classifications.

An FPGA design for simulating biological neurons

A Python and Rust chess engine that starts from Zero.