Sayak Paul
@RisingSayak
ML at Hugging Face ๐ค
๊ฐ์
May 2012
125 ํ๋ก์ ์ค 24.8K ํฌ
Here's a thread on how KV Caching is incorporated in QwenImage 2.1 ๐
It separates the fixed context from image positions that change at every denoising step. Cache the fixed part once, and later steps get much cheaper.
We measured a 2.55x speedup ๐ฅ
1/8
๋ ๋ณด๊ธฐ
