14 citations · 22 across the 11 of their papers we have counts for
8 papers · 1 filter
Search Your Block Floating Point Scales!
Tanmaey Gupta, Hayden Prairie, Xiaoxia Wu +10
Quantization has emerged as a standard technique for accelerating inference for generative models by enabling faster low-precision computations and reduced memory transfers. Recent…
HipKittens: Fast and Furious AMD Kernels
William Hu, Drew Wadsworth, Sean Siddens +6
AMD GPUs offer state-of-the-art compute and memory bandwidth; however, peak performance AMD kernels are written in raw assembly. To address the difficulty of mapping AI algorithms…
Exploring Diffusion Transformer Designs via Grafting
Keshigeyan Chandrasegaran, Michael Poli, Daniel Y. Fu +9
Designing model architectures requires decisions such as selecting operators (e.g., attention, convolution) and configurations (e.g., depth, width). However, evaluating the impact…
ThunderKittens: Simple, Fast, and Adorable AI Kernels
Benjamin F. Spector, Simran Arora, Aaryan Singhal +2
The challenge of mapping AI architectures to GPU hardware is creating a critical bottleneck in AI progress. Despite substantial efforts, hand-written custom kernels fail to meet th…
Hydragen: High-Throughput LLM Inference with Shared Prefixes
Jordan Juravsky, Bradley Brown, Ryan Ehrlich +3
Transformer-based large language models (LLMs) are now deployed to hundreds of millions of users. LLM inference is commonly performed on batches of sequences that share a prefix, s…
FlashFFTConv: Efficient Convolutions for Long Sequences with Tensor Cores
Daniel Y. Fu, Hermann Kumbong, Eric Nguyen +1
Convolution models with long filters have demonstrated state-of-the-art reasoning abilities in many long-sequence tasks but lag behind the most optimized Transformers in wall-clock…
Laughing Hyena Distillery: Extracting Compact Recurrences From Convolutions
Stefano Massaroli, Michael Poli, Daniel Y. Fu +11
Recent advances in attention-free sequence models rely on convolutions as alternatives to the attention operator at the core of Transformers. In particular, long convolution sequen…
Monarch Mixer: A Simple Sub-Quadratic GEMM-Based Architecture
Daniel Y. Fu, Simran Arora, Jessica Grogan +7
Machine learning models are increasingly being scaled in both sequence length and model dimension to reach longer contexts and better performance. However, existing architectures s…