Awesome GitHub RepositoriesLong-Context Sequence Processors

Qwen3-Coder is a specialized large language model designed for software development, technical reasoning, and automated code synthesis. Built on transformer-based sequence modeling, it functions as a multilingual programming assistant capable of generating, completing, and debugging source code across more than one hundred programming languages. The model distinguishes itself through its capacity to process and maintain logical coherence across massive datasets, supporting context windows of up to one million tokens. This allows for repository-scale reasoning, enabling the model to analyze co

TensorRT-LLM is a platform and toolkit designed for compiling, optimizing, and serving transformer-based models on accelerated hardware. It functions as a framework that transforms machine learning models into efficient execution graphs, providing an engine to refine these models for specific hardware to maximize throughput and minimize latency during text generation. The project distinguishes itself through advanced execution strategies that manage the entire inference pipeline. It utilizes kernel-level fusion and static graph execution to optimize mathematical operations and computational f

Airllm is a framework designed to execute and fine-tune large language models on consumer-grade hardware. By employing layer-wise model decomposition and memory-efficient loading techniques, the engine enables the operation of massive models that would otherwise exceed available system or video memory. The project distinguishes itself through a suite of optimization strategies that balance memory footprint with performance. It utilizes block-wise weight quantization and asynchronous layer prefetching to reduce resource consumption and hide data transfer latency. Additionally, the framework su

Lingbot-map is a feed-forward neural network designed for real-time 3D scene reconstruction from streaming video. It processes video frames one at a time without iterative optimization, producing dense geometry and camera poses at interactive frame rates directly from a live feed. The project distinguishes itself through its ability to maintain stable geometry and pose alignment across very long video sequences, handling thousands of frames without drift. It achieves this through a combination of coordinate grounding memory, sliding-window inference with overlapping keyframes, and a paged KV

bert4keras este o reimplementare ușoară a arhitecturii transformer BERT pentru framework-ul de deep learning Keras. Acesta servește drept set de instrumente pentru procesarea limbajului natural și bibliotecă de modele transformer utilizată pentru clasificarea textului, etichetarea secvențelor și extracția embedding-urilor semantice. Framework-ul include un sistem de model sequence-to-sequence pentru răspunsul la întrebări și generarea de text, precum și un server de inferență a modelului pentru a implementa transformere antrenate ca API-uri web pentru predicții în timp real. Capabilitățile acoperă o gamă largă de sarcini de înțelegere a limbajului natural, inclusiv înțelegerea lecturii, extracția relațiilor și procesarea textelor lungi. Biblioteca oferă instrumente pentru pre-antrenarea și fine-tuning-ul modelelor de limbaj, alături de tehnici de optimizare precum reducerea parametrilor, antrenarea adversă pentru robustețe și configurarea ratei de învățare pe niveluri. Proiectul include un loader de conversie a ponderilor pentru a transforma ponderile pre-antrenate din formate externe în structuri Keras compatibile.

Flash Linear Attention is a training framework and inference engine for sequence models that use linear attention and state space mechanisms, designed to process long contexts with reduced memory and compute overhead. It provides hardware-optimized token mixing layers and fused CUDA kernels that minimize memory bandwidth and launch overhead across different GPU architectures, and includes a causal inference engine that generates text token-by-token using cached hidden states for efficient autoregressive decoding. The project supports building hybrid sequence models that interleave standard at

DeepSeek-V2 este un model de limbaj de mari dimensiuni (LLM) conceput pentru procesarea limbajului natural și analiza secvențelor lungi de text. Utilizează o arhitectură de tip „mixture-of-experts” pentru a echilibra performanța ridicată cu eficiența inferenței. Modelul folosește un mecanism de rutare rar (sparse) și neuroni experți partajați pentru a capta cunoștințe comune, menținând în același timp specializarea. Reduce consumul de memorie și crește debitul prin tehnici precum multi-head latent attention, group-query attention și compresia tensorială de rang scăzut. Aceste capabilități permit procesarea și regăsirea informațiilor din volume mari de token-uri și susțin o implementare economică prin reducerea costurilor hardware și a blocajelor de memorie. Sistemul este compatibil cu interfețele API standard pentru integrarea cu toolchain-urile existente pentru modele de limbaj.

mini-sglang is a collection of tools for large language model inference, serving as an OpenAI-compatible inference server, a memory-efficient prefill engine, and a tensor parallelism runtime. It also functions as a local batch processing engine for offline benchmarking and ablation studies. The project focuses on acceleration and memory management through a KV cache manager that reuses precomputed caches for shared request prefixes. It handles large model workloads by distributing tasks across multiple GPUs and manages peak memory consumption by splitting long input sequences into smaller chu