DeepInfer

NVIDIA TensorRT is an ecosystem of APIs for high-performance deep learning inference, encompassing an inference runtime and model optimizations that deliver low latency and high throughput for production applications. Built on the CUDA parallel programming model, TensorRT optimizes neural network models trained on all major frameworks, calibrating them for lower precision with high accuracy, and deploying them across hyperscale data centers, workstations, laptops, and edge devices. It employs techniques such as quantization, layer and tensor fusion, and kernel tuning on all types of NVIDIA GPUs, from edge devices to PCs to data centers. The ecosystem includes TensorRT-LLM, an open source library that accelerates and optimizes inference performance of recent large language models on the NVIDIA AI platform, enabling developers to experiment with new LLMs for high performance and quick customization through a simplified Python API.

vLLM is a high-performance library designed to facilitate efficient inference and serving of Large Language Models (LLMs). Originally developed in the Sky Computing Lab at UC Berkeley, vLLM has evolved into a community-driven project with contributions from both academia and industry. It offers state-of-the-art serving throughput by efficiently managing attention key and value memory through its PagedAttention mechanism. It supports continuous batching of incoming requests and utilizes optimized CUDA kernels, including integration with FlashAttention and FlashInfer, to enhance model execution speed. Additionally, vLLM provides quantization support for GPTQ, AWQ, INT4, INT8, and FP8, as well as speculative decoding capabilities. Users benefit from seamless integration with popular Hugging Face models, support for various decoding algorithms such as parallel sampling and beam search, and compatibility with NVIDIA GPUs, AMD CPUs and GPUs, Intel CPUs, and more.