Microsoft

LeanAttention: Hardware-Aware Scalable Attention Mechanism for the Decode-Phase of Transformers

Transformer-based models have emerged as one of the most widely used architectures for natural language processing, natural language generation, and image generation. The size of the state-of-the-art ...
IEEE

A Sub-0.8-pJ/bit Universal Soft-Detection Decoder Using ORBGRAND

Abstract: Guessing random additive noise decoding (GRAND) has enabled the practical implementation of maximum likelihood (ML) or near-ML decoding, shifting the paradigm of code-specific decoder design ...
GitHub

Kernel Library for LLM Serving

FlashInfer is a library and kernel generator for Large Language Models that provides high-performance implementation of LLM GPU kernels such as FlashAttention, SparseAttention, PageAttention, Sampling ...
IEEE

FPGA Design and Implementation of BCH Encoders and Decoders Using Pipelined Architecture for High Throughput Applications

This paper presents the design and FPGA implementation of a high-throughput BCH (n,k) encoder and decoder using a fully pipelined architecture. Unlike conventional designs based on finite state ...