minimal workflow engine for data processing (POC)

Method for searching relevant podcast segments from transcripts using transformer models

Artificial Intelligence (AI) chip with his own hardware design and software. TruthGPT source tree

Code repository for the research paper "Space Efficient Transformer Neural Network"

📦 Optimize tokenization in C++ for HuggingFace models with a fast, production-ready library supporting BPE, WordPiece, and Unigram methods.

Transformers Games by High Moon Studios. Transformers ReEnergized Server Project is looking to Reverse Engineer and bring back the Transformers Multiplayer Servers that were shut down by Demonware Services and revive them for fans.

Optimization of Attention layers for efficient inferencing on the CPU and GPU. It covers optimizations for AVX and CUDA also efficient memory processing techniques.

Lightweight, header-only Byte Pair Encoding (BPE) trainer in modern C++17. Produces HuggingFace-compatible vocabularies for transformers and integrates with Modern Text Tokenizer.

GGML-based C++ inference for Mel-Band-Roformer vocal separation | 纯 C++ 实现的基于 GGML 的 Mel-Band-Roformer 人声分离推理

High-performance distributed matrix computation for AI workloads. Supports CPUs, Vulkan/Metal GPUs, PyTorch CUDA nodes, and LLaMA/ggml backends. Uses shard-based distribution with ZeroMQ networking, RAM/disk storage, and flexible environment-based configuration for multi-node clusters.

🌐 Build dynamic breadcrumbs that update in real-time with TanStack Start, showcasing server-side rendering and file-based routing for smooth navigation.

Arduino coil and transformer tester and ring detector

Poplar implementation of FlashAttention for IPU

Experimental Implementation of Transformer (With Smolgen) on NVIDIA GPU with the help of LibTorch

NanoOWL Detection System enables real-time open-vocabulary object detection in ROS 2 using a TensorRT-optimized OWL-ViT model. Describe objects in natural language and detect them instantly on panoramic images. Optimized for NVIDIA GPUs with .engine acceleration.

🚀 Enhance vllm with exvllm to utilize MOE mixed inference, enabling efficient CPU + GPU processing of large models effortlessly.

C++ Library for tokenizing MIDI files, designed to be compatible with the MIDITok python library

This is the official implementation for our paper;"LAR:Look Around and Refer".

Global vision positioning system for UAVs in outdoor GNSS-denied environments

Whisper in TensorRT-LLM