Slash Compute

Inspiration

While tutoring students, I come across mainly 2 types of people. The ones that have the capability(computer), and have good ideas, or the ones that lack the capability, but have great ideas. I I found that the step between knowing and understanding is technical experience. I remember my first time running a train.py, and watching the interesting hallucinations. I was acutely aware of the barrier of entry for most people to start learning AI.

What it does

It takes a percentage of computing power from each computer and combines it into a group supercomputer. By contributing to the computer, you get usage credits, measure in flops for the computer. Additionally, you can grant research funds by giving out flops.

How we built it

We built the central engine of distributed computing and worked from there. We implemented functionality with inference and fine tuning, and then added flops to track usage.

Challenges we ran into

Issues with the main functionality.

Accomplishments that we're proud of

Getting the site up, connection our macbooks.

What we learned

How important security is with distributed computing.

What's next for Slash Compute

Expanding the reach for more users. Giving grants for research.

Built With

• 3.11+
• fastapi
• hatchling
• html
• httpx
• llama.cpp
• mlx
• mlx-lm
• msgpack
• numpy
• psutil
• pydantic
• pytest
• pytest-asyncio
• python
• python-multipart
• pywebview
• sqlalchemy
• sqlite
• sqlmodel
• typer
• uv
• uvicorn
• websockets
• zeroconf