AI 资讯

Show HN: Fast NF4 dequantization Triton kernel (1.41x faster than bitsandbytes)

Griffith-7 2026年07月15日 21:41 1 次阅读来源：HackerNews

"I wanted to see if I could optimize the dequantization bottleneck during 4-bit LLM inference. By writing a custom kernel in Triton to optimize memory access patterns, I managed to get up to a 1.41x speedup over the standard bitsandbytes implementation. Check out the source code and benchmarks, feedback is highly appreciated!"

本文内容来源于互联网，版权归原作者所有

查看原文

# hackernews

Why did my benchmark stop at N=22? A debugging story in nine bugs

07-16

My benchmark's Python column was N/A for a year — CPython's 4300-digit limit, and eight other bugs

07-16

I spent a week trying to intercept Slack push notifications from a Chrome extension. Here's why it's impossible.

07-16

Show HN: Fast NF4 dequantization Triton kernel (1.41x faster than bitsandbytes)

相关文章

Why did my benchmark stop at N=22? A debugging story in nine bugs

My benchmark's Python column was N/A for a year — CPython's 4300-digit limit, and eight other bugs

I spent a week trying to intercept Slack push notifications from a Chrome extension. Here's why it's impossible.

I Built a Dashboard to Make Security Reports Easier to Read

Five Local-First Mac Apps I Built to Fix Everyday Workflow Problems

find command finds things