Mia
@MiaAI_lab
Building with AI & LLMs | Insights, recipes, tools & honest experiments
๊ฐ์
July 2022
364 ํ๋ก์ ์ค 12.3K ํฌ
FYI the best Qwen 3.6 35b nvfp4 to run is the @NVIDIAAI nvfp4. Do not use unsloth nvfp4, it performs worse.
Here's how it compares to bf16 ๐
