Hysen Labsmirage-project
mirage
Mirage Persistent Kernel: Compiling LLMs into a MegaKernel. Quickstart Mirage allows you to compile LLMs from the Hugging Face model zoo into a megakernel using just a few dozen lines of Python-mainly to define the kernel s inputs and outputs.
2,394 stars235 forksCudaApache-2.0
01
EDITORIAL BRIEF
What it solves
Mirage Persistent Kernel: Compiling LLMs into a MegaKernel. Quickstart Mirage allows you to compile LLMs from the Hugging Face model zoo into a megakernel using just a few dozen lines of Python-mainly to define the kernel s inputs and outputs.
The project context on this page is free to read. The original GitHub repository remains the source of truth; sign in only when you want to save or join the discussion.
Community notes