Installation & upgrade
How to install Giskard on your machine, set up your Python backend and upgrade Giskard
To install Giskard, you need a Linux or macOS machine or WSL2 in Windows with:
- Giskard uses 2 TCP ports:
19000and40051. If you don't use Giskard locally (installation in the cloud for instance), make sure that these two ports are open on the machine where Giskard is installed - sudo curl -fsSL https://get.docker.com -o get-docker.shsudo sh get-docker.sh
Run the following commands to install Giskard on your server
git clone https://github.com/Giskard-AI/giskard.git
cd giskard
docker compose pull && docker compose up -d --force-recreate --no-build
- If you have an error because of the rate limit (
toomanyrequests: Rate exceeded), please re-execute the docker-compose command line once again. - If
composeis not found, you may have an older version of docker, so you'll need to usedocker-composeinstead ofdocker compose.Alternatively, you can also upgrade your docker version
Once the docker-compose starts all the modules, you'll be able to open Giskard at http://localhost:19000/
Since the backend container may take some minutes to load, please wait a bit and refresh the webpage http://localhost:19000/
To log in, the default credentials are Login: admin / Password: admin
You're all set to try Giskard in action. Upload your first model by following the Upload your ML model & data tutorial.
In order to upgrade Giskard to the latest version, please run the following in your Giskard distribution directory
git pull
docker compose down && docker compose pull && docker compose up -d --force-recreate --no-build
- The browser may keep using an old version of Giskard UI due to caching. If there are issues after running an upgrade try to hard refresh Giskard page by pressing
Ctrl + Shift + RorCommand + Shift + R - If you installed in Giskard additional Python libraries or a new Python version, you will need to reinstall them. Please refer to configuration.
How can I connect to my local Giskard instance from Google Colab/other remote notebook or code environment?
- 1.Download the configuration file on the device hosting the Giskard instance
- 2.In that folder, run the command
ngrok start --config ngrok.yml --all --authtoken YOUR_AUTH_TOKEN - 3.You should see an output similar to this:
Sample "ngrok start" output - 4.Start your ML Worker with:
giskard worker start -h X.tcp.xx.ngrok.io -p XXXXXreplacing with the URL and port from your console. - 5.Create your GiskardClient with your address like this:
GiskardClient('https://xxxx-xx-xx-xx-xx.xx.ngrok.io')
If you encounter any other issues, join our Discord on our #support channel. Our community will help!
Last modified 21d ago