Optimus
ML from scratch in python and CUDA for educational purposes.
Setup Dev Env
We use docker to setup and use dev environment. Please make sure you have docker CLI, engine and compose installed. You can refer this link to install everything you need.
Build optimus dev image using
docker compose build
Then start the docker container in interactive mode with bash
docker compose run optimus bash
Building Optimus
Once you start the docker container, you will be put in /workspace/host/optimus by default.
Do the following to genrate the make files.
mkdir build && cd build
cmake ..
Now build Optimus with
make all
Running Tests
Run any python unit test individually, for example to test the matmul op
python3 /workspace/host/optimus/tests/ops/test_matmul.py