trends.barnas.net

tool · first seen 20 Oct 2026

FlexAttention

A PyTorch API for expressing custom attention variants such as score modification and masking compiled into fused kernels.

Mentions per week

0119 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

graph generationcausal attentioncache blocking