Transformer

A comprehensive Chinese-language course for building large language models from scratch, covering pre-training data engineering, tokenizers,...

vLLM Ascend Plugin is a community-maintained hardware acceleration plugin for vLLM that enables large language model serving on Huawei Ascend...

vLLM-Omni is a framework for efficient serving of multimodal and omni-modality models (supporting text, audio, image, video, and world models) built...

AlphaGPT is an open-source automated factor factory using deep reinforcement learning for quantitative finance applications. It is specialized for...

VibeThinker is a family of compact dense reasoning language models (1.5B and 3B parameters) designed to achieve frontier-level performance on...

This is a unified evaluation framework for testing generative language models against 60+ standard academic benchmarks with hundreds of subtasks. It...

whisper.cpp is a high-performance C/C++ port of OpenAI's Whisper automatic speech recognition (ASR) model, designed to run efficiently on a wide...

SGLang is a high-performance inference serving framework optimized for large language models and multimodal models, with support for speculative...

faster-whisper is a high-performance reimplementation of OpenAI's Whisper speech-to-text model using CTranslate2, delivering up to 4× faster...

Heretic is a specialized tool for removing safety alignment (censorship) from transformer-based language models through automatic directional...

Cactus is a low-latency AI inference engine purpose-built for mobile devices and wearables, enabling on-device LLM, vision, and speech processing via...

Nano-vLLM is a lightweight, from-scratch implementation of vLLM optimized for fast LLM inference on consumer GPUs, achieving comparable or better...

A curated research resource listing papers and tooling for fraud, anomaly, and outlier detection using graph neural networks and transformer-based...

vLLM is a high-throughput, memory-efficient inference and serving engine for large language models, using PagedAttention and continuous batching to...

FireRedASR is an open-source industrial-grade automatic speech recognition system supporting Mandarin, Chinese dialects, and English, with SOTA...

A structured educational repository providing hands-on Jupyter notebooks and code for learning AI fundamentals, from math and PyTorch through...

nndeploy is a C++ AI deployment framework designed to optimize inference for models (LLMs, diffusion models, transformers) across heterogeneous...

BeatAI is an educational content platform and learning resource hub focused on making AI fundamentals accessible to learners at all levels, from...

NeuralForecast is a Python library providing scalable neural forecasting models (NBEATS, Transformers, DeepAR, etc.) for time-series prediction...

Manga Image Translator is a specialized tool for translating text within manga, comics, and images using deep learning-based OCR, text detection, and...

Parallax is a distributed inference framework that enables decentralized LLM serving across heterogeneous devices using P2P communication and...

This is a curated collection repository (awesome-list) aggregating research papers, blogs, and technical resources on long-context language modeling...

TorchAO is a PyTorch native library for quantization and sparsity optimizations spanning training and inference, with demonstrated speedups (1.5x...

MNN is a lightweight, high-performance inference engine optimized for on-device execution of LLMs and edge AI models, specifically battle-tested by...

Hugging Face Transformers is the de facto standard Python library for accessing, fine-tuning, and deploying state-of-the-art pretrained models across...

RWKV is a specialized linear-time RNN architecture that achieves transformer-level LLM performance while maintaining constant memory usage and no...

torchdistill is a configuration-driven framework for knowledge distillation and reproducible deep learning experiments built on PyTorch. It enables...

A React Native transformer that enables importing SVG files as React components, mirroring the SVGR workflow from web development. This specialized...

PaddleSpeech is a specialized speech and audio toolkit built on PaddlePaddle that provides state-of-the-art models for ASR, TTS, speaker...

tsai is a specialized deep learning library for time series analysis built on PyTorch and fastai, offering state-of-the-art models for...