How it started:
I setup #ollama on a $250 old mini pc, downloaded a few models, installed openweb-ui, gave a few friends access to it.
How it's going:
The PC is running 100% like 14 hours a day with friends and family of friends using it. Some have cancelled their paid accounts to use this instead.
I maxed out the ram at 64 GB to keep more models in ram based in usage patterns.
I'm now hosting ollama as a service. People pitched in to pay for the ram and power without me asking.
Responses aren't instant because no usable GPU (AMD 680M) but no one seems to care.
#Privacy > speed.
Login to reply
Replies (2)
I found ollama too awkward to set up for models not pre-provisioned by the ollama team. I turned to #koboldcpp to run any model I want with single-binary installation (text gen, image gen, TTS) and super simple configuration. And it comes with webuis for text and stable diffusion; or use any ui that supports koboldcpp-api or openai-api or automatc1111-api. CUDA, ROCM, Vulkan, CPU-only all supported.
I'll check it out. Thanks!