sauloaalmeida/fsp-python-gpu

Feature Space Partition implementation in Python using GPU

FSP (Feature Space Partition) implementation in Python using GPU

Licença

THis project is licenced uder Apache 2.0 - details in LICENSE.

Abstract

Implementation of FSP in python running over GPU hardware.

How Execute

Software requirements:

Docker container or Conda/Manba/Miniforge3 environmet!

Tested environments:

Env	Result
Windows 11 + Docker V19.03.8	OK
Ubuntu 22.04 + Docker V20.10.14	OK
Ubuntu 22.04 + Miniforge V20.10.14	OK

Fast Execution (accessing Docker Hub directly)

1 - Run image automatically downloading the last version published in docker hub

$ docker run -p 8888:8888 fsp/fsp-python-gpu

2 - Open your browser in the same host machine and access jupyter notbook URL

http://127.0.0.1:8888

Running locally from conda environment

1 - With a Conda/Mamba/Miniforge enviroment installed and initilized, clone the original repository in GitHub

$ git clone https://github.com/sauloaalmeida/fsp-python-gpu --branch=v1.0

2 - Go to the cloned directory

$ cd fps-python-gpu

3 - Create conda project environment

$ conda env create -f fsp-gpu.yml

4 - Activate project environment

$ conda activate fsp-gpu

Running locally unit tests

1 - After activate conda project environment:

$ conda activate fsp-gpu

2 - For a detailed report, in root's directory execute:

$ pytest -v

Running on docker locally from scratch

1 - Clone the original repository in github

$ git clone https://github.com/sauloaalmeida/fsp-python-gpu --branch=v1.0