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 es una reimplementación ligera de la arquitectura de transformador BERT para el framework de aprendizaje profundo Keras. Sirve como un kit de herramientas de procesamiento de lenguaje natural y librería de modelos de transformadores utilizada para clasificación de texto, etiquetado de secuencias y extracción de embeddings semánticos. El framework incluye un sistema de modelos de secuencia a secuencia para responder preguntas y generar texto, así como un servidor de inferencia de modelos para desplegar transformadores entrenados como APIs web para predicciones en tiempo real. Las capacidades cubren una amplia gama de tareas de comprensión del lenguaje natural, incluyendo comprensión de lectura, extracción de relaciones y procesamiento de textos largos. La librería proporciona herramientas para el pre-entrenamiento y ajuste fino de modelos de lenguaje, junto con técnicas de optimización como reducción de parámetros, entrenamiento adversarial para robustez y configuración de tasa de aprendizaje por capa. El proyecto incluye un cargador de conversión de pesos para transformar pesos pre-entrenados de formatos externos en estructuras de Keras compatibles.

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 es un modelo de lenguaje de gran tamaño diseñado para el procesamiento de lenguaje natural y el análisis de secuencias de texto extensas. Utiliza una arquitectura de mezcla de expertos (MoE) para equilibrar un alto rendimiento con la eficiencia en la inferencia. El modelo emplea un mecanismo de enrutamiento disperso y neuronas expertas compartidas para capturar conocimiento común mientras mantiene la especialización. Además, reduce la sobrecarga de memoria y aumenta el rendimiento mediante atención latente de múltiples cabezas, atención de consulta de grupo y compresión de tensores de bajo rango. Estas capacidades permiten el procesamiento y la recuperación de información a partir de un gran número de tokens y facilitan un despliegue económico al reducir los costes de hardware y los cuellos de botella de memoria. El sistema es compatible con interfaces API estándar para su integración con toolchains de modelos de lenguaje existentes.

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