Skip to content

Latest commit

 

History

History
110 lines (82 loc) · 3.9 KB

install.md

File metadata and controls

110 lines (82 loc) · 3.9 KB

Installation

Prerequisites

  • Python>=3.11.0,<3.14.0
  • Ubuntu-24.04 for Blackwell GPUs

Below are installation methods for different operating systems.

Operating System Windows App From Source Docker
Windows ✅️ Not recommended Not recommended
Linux ❌️ ✅️ ✅️
macOS ❌️ ✅️ ❌️

From Source

For Linux/WSL (GPU):

Note: If you are using DGX Spark, please refer to the Docker installation section

git clone https://github.com/hetu-project/prakasa.git
cd prakasa
pip install -e '.[gpu]'

For macOS (Apple silicon):

We recommend macOS users to create an isolated Python virtual environment before installation.

git clone https://github.com/hetu-project/prakasa.git
cd prakasa

# Enter Python virtual environment
python3 -m venv ./venv
source ./venv/bin/activate

pip install -e '.[mac]'

Next time to re-activate this virtual environment, run source ./venv/bin/activate.

Extra step for development:

pip install -e '.[dev]'

Windows Application

Click here to get latest Windows installer.

After installing .exe, right click Windows start button and click Windows Terminal(Admin) to start a Powershell console as administrator.

❗ Make sure you open your terminal with administrator privileges.

Ways to run Windows Terminal as administrator
  • Start menu: Right‑click Start and choose "Windows Terminal (Admin)", or search "Windows Terminal", right‑click the result, and select "Run as administrator".
  • Run dialog: Press Win+R → type wt → press Ctrl+Shift+Enter.
  • Task Manager: Press Ctrl+Shift+Esc → File → Run new task → enter wt → check "Create this task with administrator privileges".
  • File Explorer: Open the target folder → hold Ctrl+Shift → right‑click in the folder → select "Open in Terminal".

Start Windows dependencies installation by simply typing this command in console:

prakasa install

Installation process may take around 30 minutes.

To see a description of all Prakasa Windows configurations you can do:

prakasa --help

Docker

For Linux+GPU devices, Prakasa provides a docker environment for quick setup. Choose the docker image according to the device's GPU architechture.

GPU Architecture GPU Series Image Pull Command
Blackwell/Ampere/Hopper RTX50 series/RTX40 series/B100/B200/A100/H100... docker pull gradientservice/prakasa:latest
DGX Spark GB10 docker pull gradientservice/prakasa:latest-spark

Run a docker container as below. Please note that generally the argument --gpus all is necessary for the docker to run on GPUs.

# For Blackwell/Ampere/Hopper
docker run -it --gpus all --network host gradientservice/prakasa:latest bash
# For DGX Spark
docker run -it --gpus all --network host gradientservice/prakasa:latest-spark bash

The container starts under prakasa workspace and you should be able to run prakasa directly.

Uninstalling prakasa

For macOS or Linux, if you've installed Prakasa via pip and want to uninstall it, you can use the following command:

pip uninstall prakasa

For Docker installations, remove Prakasa images and containers using standard Docker commands:

docker ps -a               # List running containers
docker stop <container_id> # Stop running containers
docker rm <container_id>   # Remove stopped containers
docker images              # List Docker images
docker rmi <image_id>      # Remove Prakasa images

For Windows, simply go to Control Panel → Programs → Uninstall a program, find "Gradient" in the list, and uninstall it.