Easily execute your model in your Python environment
To leverage your Python environment with all the packages you've already installed, Giskard provides the ML Worker component.
- It opens a secured connection between your working Python environment and the Giskard platform you've installed
- It executes the model in your working Python environment (notebook, Python IDE, etc)
giskardpython library in the desired code environment:
pip install giskard
2. Then start an ML worker:
giskard worker start
If ML Worker manages to connect to Giskard instance, you should see the following message in the worker logs: "Connected to Giskard server." By default,
giskard worker startestablishes a connection to the Giskard instance installed on
If Giskard is not installed locally, please specify the IP address (and a port in case a custom port is used). For example,
giskard worker start -h 188.8.131.52
To start ML Worker as a daemon and let it run in the background, add
giskard worker start -d
It's possible to start multiple ML Workers, for example, to connect them to different Giskard instances. It's not possible, however, to have multiple workers that use the same python interpreter to be connected to the same Giskard instance.
If multiple workers are connected to Giskard, the latest one will be used.
To stop a particular ML Worker that runs as a daemon
stopcommand should be called with the same parameters that were used to start it.
For example, to stop a worker started with default arguments, it's enough to call
giskard worker stop
If a worker was started like
giskard worker start -h 184.108.40.206 -p 1234 -d
then it can be stopped with
giskard worker stop -h 220.127.116.11 -p 1234
To stop all ML Workers running on a given machine:
giskard worker stop -a
Admin users can find information about an ML Worker that is currently active in Giskard on a Giskard settings page:
By default, ML Worker execution logs are located in
You can access the logs by executing the following command in your notebook:
!tail -f $HOME/giskard-home/run/ml-worker.log