<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Triton on Vectors &amp; Verbs</title><link>https://vectorsandverbs.com/tags/triton/</link><description>Recent content in Triton on Vectors &amp; Verbs</description><generator>Hugo</generator><language>en-us</language><lastBuildDate>Sun, 05 Apr 2026 10:14:00 +0530</lastBuildDate><atom:link href="https://vectorsandverbs.com/tags/triton/index.xml" rel="self" type="application/rss+xml"/><item><title>Speed of Thought: Optimizing Gemma 27B with Custom Triton Kernels</title><link>https://vectorsandverbs.com/posts/gemma-triton-optimization/</link><pubDate>Sun, 05 Apr 2026 10:14:00 +0530</pubDate><guid>https://vectorsandverbs.com/posts/gemma-triton-optimization/</guid><description>Profiling Gemma 27B on dual RTX 5090 Blackwell GPUs, discovering a legacy CUDA path bottleneck, and building a custom Triton kernel that shaves 1.1 seconds off every inference batch.</description></item></channel></rss>