What device can I use? Hopper?

#1
by Alwin114 - opened

c the content in the title

Inference Optimization org
•
edited about 22 hours ago

Hey Alwin! MXFP4 is natively supported in Blackwell GPUs, some runtimes offer emulation but it may be inefficient. Try out vLLM's Humming Kernel https://vllm.ai/blog/2026-10-21-humming

Sign up or log in to comment