Open-source alternatives to FasterTransformer

DeepSpeed is a distributed deep learning optimization library and framework designed for the training and inference of massive AI models. It serves as a model parallelism orchestrator and a toolkit for scaling large language models across multiple GPUs and compute nodes. The project distinguishes itself through 3D parallelism orchestration, which combines data, pipeline, and tensor parallelism. It utilizes ZeRO-based memory partitioning to eliminate redundant storage and employs CPU-offload memory management to move weights and optimizer states to system RAM. Additionally, it provides special

FastDeploy is a high-performance deployment framework for large language models, vision models, and multimodal models. It provides the infrastructure to launch model services that process combined image, video, and text inputs, exposing these capabilities through a standardized, OpenAI-compatible API for chat and text completions. The project distinguishes itself through advanced inference pipeline engineering and GPU optimization. It employs speculative decoding, tensor parallelism, and a disaggregated execution model that separates prefill and decode phases across different hardware resourc

CTranslate2 is a C++ inference engine and runtime for Transformer models, designed to execute models on both CPU and GPU with optimizations for speed and memory efficiency. It functions as a model format converter, quantization tool, and REST API server, enabling deployment of neural machine translation, automatic speech recognition, and text generation models. The engine distinguishes itself through a suite of runtime optimizations including layer fusion, weight-matrix quantization, batch-by-length grouping, and a caching allocator that reuses GPU memory. It supports tensor-parallel model di

Intel XPU LLM Acceleration Library is a toolkit designed to accelerate large language model inference and finetuning on Intel CPUs, GPUs, and NPUs. It provides a distributed inference engine for scaling models across multiple accelerators, a multimodal model runtime for vision and speech tasks, and a low-bit model quantization tool for converting weights into INT4, FP8, and GGUF formats. The project features a parameter-efficient finetuning framework that enables model adaptation using QLoRA and DPO on Intel hardware. It distinguishes itself by providing specialized optimizations for Intel XP

gpt-fast is a PyTorch transformer inference engine designed for text generation using a native tensor library implementation. It provides a runtime for executing large language models without the need for external C++ extensions. The project implements speculative decoding to accelerate generation by using a small draft model for token prediction and a larger model for verification. It further optimizes performance through a compiled prefill stage and a multi-GPU tensor parallelism library that shards linear layers across multiple graphics processing units. Memory efficiency is managed throu

fastllm is a set of specialized software components for model weight conversion, Mixture-of-Experts runtimes, and tensor parallelism. It provides an OpenAI compatible API server to expose large language model capabilities through a standardized request format. The project features a tensor parallelism framework that splits computational workloads across multiple GPUs to accelerate execution. It includes a dedicated runtime optimized for Mixture-of-Experts architectures and a quantization tool to convert model weights into lower precision formats to reduce memory usage and increase throughput.

Ktransformers is a comprehensive framework designed for the operation, fine-tuning, and serving of large language models. It functions as a heterogeneous inference engine and quantized execution runtime, enabling the deployment of massive models by distributing computational workloads across both CPU and GPU resources. This architecture allows users to bypass local memory constraints, making it possible to run and train models that exceed the capacity of a single device. The project distinguishes itself through specialized support for sparse architectures, particularly mixture-of-experts mode

OpenNMT-py is a PyTorch neural machine translation framework used for training and deploying neural machine translation and large language models. It functions as a distributed model training system, an inference engine, and a toolkit for fine-tuning large language models. The framework distinguishes itself with a dedicated toolkit for adapting large language models through low-rank adaptation, quantization, and instruction tuning. It also includes a neural machine translation server that allows trained models to be hosted and exposed via REST API endpoints. The project covers a broad range

LightLLM is a high-performance serving framework for deploying and executing large language models. It functions as a multi-GPU inference engine and server capable of handling dense architectures, mixture-of-experts designs, and multimodal models that process both text and images. The system is distinguished by its specialized support for Mixture-of-Experts models using expert parallelism and fused kernels. It implements structured text generation through deterministic state machines and pushdown automata to enforce precise output formats. To optimize throughput, the framework employs specula

tiny-llm is a large language model inference engine and transformer model implementation. It serves as a quantized model runtime and paged key-value cache manager, providing a specialized inference stack optimized for Apple Silicon. The system distinguishes itself through high-throughput execution techniques, including continuous batching and paged attention. It utilizes a paged memory system to eliminate fragmentation during token generation and employs on-the-fly dequantization of compressed weights to reduce the memory footprint during matrix multiplication. The project covers a broad ran

lmdeploy is a high-performance inference engine and deployment framework for large language models and vision models. It functions as a multi-modal model server and compression toolkit designed to serve models with high throughput and low latency. The system enables the distribution of model services across multiple machines using request-based load balancing and tensor parallelism. It includes specialized tools for model quantization and compression to reduce the memory footprint of weights and caches. The framework covers broad capability areas including production deployment, distributed

Sglang is a high-performance inference engine and serving system designed for large language and multimodal models. It provides a programmable interface for orchestrating complex generation workflows, enabling developers to coordinate multi-turn dialogues, tool invocations, and reasoning chains through a domain-specific language. The platform is built to support production-scale deployments, offering an OpenAI-compatible API that allows for integration with existing application ecosystems. The system distinguishes itself through a disaggregated architecture that separates compute-intensive pr

NeMo is a multimodal AI framework and toolkit designed for the development, training, and scaling of large language models, generative AI systems, and speech-based models. It functions as an automatic speech recognition toolkit, a text-to-speech engine, and a framework for building models that process and generate combinations of text, image, and audio data. The project serves as a conversational AI orchestrator capable of managing real-time, interruptible voice interactions. It provides specialized workflows for speech translation, converting spoken audio from one language into text or speec

llama.cpp is a high-performance C++ inference engine and runtime for executing large language models locally across various hardware architectures. It provides the core components for local model execution, including a dedicated model quantizer for compressing weights into the GGUF format and a system for generating text embeddings for semantic search. The project distinguishes itself through specialized memory and execution optimizations, such as block-wise weight quantization to reduce memory footprints and memory-mapped model loading. It supports structured text generation by using formal

Composer is a PyTorch distributed training framework designed for scaling large-scale models across multi-node GPU clusters. It functions as a large language model trainer, a distributed model optimizer, and a training lifecycle manager. The project differentiates itself as a deep learning regularization library, providing specialized optimization techniques such as Sharpness Aware Minimization, MixUp, and CutMix to improve model generalization. It further distinguishes its training flow through the use of sequence length warmup, progressive layer freezing, and sharded-state checkpointing for

Torchtitan is a reference implementation for distributed deep learning built within the PyTorch ecosystem. It provides a framework for training large neural network models across multiple GPUs and nodes by combining several parallelism techniques, including fully sharded data parallelism (FSDP), tensor parallelism, and pipeline parallelism, making it possible to train models that exceed the memory capacity of a single device. The system distinguishes itself through asynchronous checkpointing, which saves model and optimizer state to persistent storage without pausing the training loop, enabli

Text Generation Inference is a production-ready engine designed for the deployment and serving of large language models. It functions as a containerized runtime environment that manages model execution, scales across distributed hardware, and provides high-performance inference capabilities for demanding production environments. The project distinguishes itself through advanced optimization techniques, including continuous batching to maximize hardware utilization and tensor parallelism to shard large models across multiple accelerator cards. It supports efficient inference through custom com

FlexGen is an inference engine for large language models designed for high-throughput execution on single or multiple GPUs. It functions as a framework for managing model execution through a combination of memory offloading, weight compression, and pipeline orchestration. The system enables the execution of models that exceed available GPU memory by moving tensors and caches between GPU memory, system RAM, and disk storage. It utilizes 4-bit weight quantization to reduce the memory footprint of model parameters, allowing for increased batch processing capacity. The project covers distributed

faster-whisper is an automatic speech recognition framework and an optimized implementation of the Whisper speech-to-text engine. It functions as a CTranslate2 inference engine designed to convert spoken audio into written text. The project serves as a model quantization tool that transforms large audio model weights into lower precision formats. This process reduces memory usage and increases execution speed on hardware by utilizing integer quantized weights. The framework covers a broad range of capabilities including batch audio transcription for parallel processing and voice activity det

AISystem is a comprehensive AI full-stack infrastructure project covering the entire pipeline from AI chip architecture to high-level training frameworks. It encompasses the development of AI compiler frameworks, inference engines, and distributed training orchestrators designed to coordinate workloads across a heterogeneous compute stack of CPUs, GPUs, and NPUs. The project focuses on the deep integration of software and hardware, employing software-hardware co-design to align tensor layouts with physical memory structures. It provides specialized capabilities for accelerating Transformer mo

Swift is a toolkit for the full-parameter and parameter-efficient fine-tuning of large language and multimodal models. It functions as a multimodal model trainer for text, image, video, and audio data, and includes specialized tools for model compression and reinforcement learning from human feedback. The framework provides an alignment toolkit for optimizing model behavior using preference learning algorithms and reinforcement learning. It integrates parameter-efficient fine-tuning methods to adapt models with minimal memory and compute requirements, alongside utilities for reducing hardware

AutoGPTQ is a model compression toolkit and post-training quantization framework designed to reduce the memory footprint of large language models. It utilizes the GPTQ algorithm to compress neural network weights, lowering hardware requirements and reducing VRAM usage. The project serves as an inference accelerator by providing optimized kernels that increase token generation speed. It features model architecture extensibility, allowing quantization capabilities to be added to new model structures through configurable patterns. The framework covers a comprehensive quantization pipeline, incl

ipex-llm is an acceleration library and inference engine designed to optimize the execution and finetuning of large language models on Intel GPUs and NPUs. It provides a HuggingFace compatible model backend and a dedicated quantization toolkit for converting model weights into low-bit precision formats. The project facilitates distributed inference by splitting large model workloads across multiple accelerators using pipeline and tensor parallelism. It enables the deployment of models on Intel Arc, Flex, and Max GPUs to increase throughput and reduce latency. The library covers a broad range

MiniCPM is a collection of small language models designed for local, on-device deployment in resource-constrained environments. The project focuses on running dense Transformer models on consumer hardware, including GPUs, CPUs, and Apple Silicon, without requiring custom code forks. The project distinguishes itself through heavy optimization for edge hardware, utilizing quantized weight compression in GGUF and MLX formats to reduce memory overhead. It implements advanced inference techniques such as speculative sampling and radix-tree prefix caching to accelerate generation speed and throughp

This project is a large language model inference library and framework designed to run models for text generation, problem solving, and coding assistance. It includes a multimodal framework for processing combined image and text inputs and a tool-use implementation that enables the execution of external functions based on model reasoning. The system features a distributed GPU inference engine that spreads large model workloads across multiple graphics processors to increase processing speed and meet memory requirements. It also provides containerized model deployment through pre-packaged imag

Fairseq is a PyTorch toolkit for sequence-to-sequence modeling, specializing in neural machine translation, automatic speech recognition, and large-scale language model training. It provides a framework for processing and aligning diverse data sources, including text, audio, and video, to support tasks such as speech-to-text conversion and multimodal sequence learning. The project is distinguished by its distributed training capabilities, which utilize parameter sharding, mixed-precision training, and CPU offloading to handle models that exceed single-device memory. It also includes specializ

PaddleNLP is a development library and toolkit for training, fine-tuning, and deploying large and small language models using the PaddlePaddle framework. It provides a comprehensive suite for the entire natural language processing lifecycle, from model development to high-performance inference. The project features a standardized model zoo for loading and managing pre-trained models and tokenizers through a unified interface. It distinguishes itself with a specialized model compression framework that reduces memory footprints via weight precision conversion and lossless size optimization, alo