Running On <4gb Vram
#6
by Jit2024 - opened
Hi, I found out a way of running this model on <4gb Vram. And I have not used any quantization.
Repo: https://github.com/Jit-Roy/WeeLLM
that's awesome; your work is inference level routing, once you accelerate to the engine level, you will get benefit from both memory efficiency and runtime speed; diffusers is an inference and the compute kernels are all inside torch