GCP#

Installing Giskard in GCP enables you to inspect & test models that you created in the GCP environment (Workbench, Vertex AI, etc.). Here are the 3 steps to install Giskard in a new VM instance in GCP:

1. Create a Giskard VM Instance in GCP#

  1. Go to VM instances in Compute Engine and create a VM instance

  2. In the configuration of your VM :

    1. We recommend you choose at least an e2-standard-2 (2vCPU, 8GB memory)

    2. Choose Allow full access to all Cloud APIs

    3. In the firewall section, allow HTTP and HTTPS traffic

  3. Connect to your VM in SSH by opening a browser window

  4. Create a firewall rule to open the 19000 port of the Giskard instance. Here is the command line that you can execute in the terminal opened by your SSH connection:

gcloud compute firewall-rules create giskard-main --allow tcp:19000

Warning

Make sure you have the proper rights to open a port. If not contact your GCP administrator.

Note

Creating the firewall rules can also be done through UI in the VPC Networksection:

  • Go to the firewall in VPC Network section of GCP

  • Click on create a firewall rule

  • In Targets, select All instances in the network

  • In Source filter, choose IPv4 ranges

  • In source IPv4 ranges, select 0.0.0.0/0

  • In Protocols and ports, select Specified protocols and ports

  • Then select TCP, and type 19000

2. Install Giskard in the GCP VM#

  • Installation of the Giskard requirements (git and docker)

 curl -fsSL https://get.docker.com -o get-docker.sh
 sudo sh get-docker.sh

  • Installation of Giskard

giskard hub start

3. Connect to your instance#

  • Get the external IP address of your Giskard VM in the VM instances section of the Compute Engine

  • Go to http://<your IP address>:19000 in your web browser

Hint

You can stop the instance and restart it when you need to save your GCP compute costs. However, note that

  • the IP address will not necessarily be the same. So make sure you copy it again when it’s launched.

  • you will need to restart the giskard hub, by executing in the Giskard folder:

giskard hub start

4. Start the ML worker#

Giskard executes your model using a worker that runs the model directly in your Python environment, with all the dependencies required by your model. You can either execute the ML worker:

  • From your local notebook within the kernel that contains all the dependencies of your model

  • From Google Colab within the kernel that contains all the dependencies of your model

  • Or from your terminal within the Python environment that contains all the dependencies of your model

Note

If you plan to use LLM-assisted tests or transformations, don’t forget to set the OPENAI_API_KEY environment variable before starting the Giskard worker.

To start the ML worker from your notebook, run the following code in your notebook:

!giskard worker start -d -k YOUR_KEY -u http://<your IP address>:19000/

The API Access Key (YOUR_KEY) can be found in the Settings tab of the Giskard Hub.

⚠️ Warning

To see the available commands of the worker, you can execute:

!giskard worker --help

You’re all set to try Giskard in action. Upload your first model, dataset or test suite by following the upload an object page.

To start the ML worker from your Colab notebook, run in your Colab cell:

!giskard worker start -d -k YOUR_KEY -u http://<your IP address>:19000/

The API Access Key (YOUR_KEY) can be found in the Settings tab of the Giskard Hub.

⚠️ Warning

To see the available commands of the worker, you can execute:

!giskard worker --help

You’re all set to try Giskard in action. Upload your first model, dataset or test suite by following the upload an object page.

  • Run the following command within the Python environment that contains all the dependencies of your model:

giskard worker start -k YOUR_KEY -u http://<your IP address>:19000/

The API Access Key (YOUR_KEY) can be found in the Settings tab of the Giskard Hub.

⚠️ Warning

To see the available commands of the worker, you can execute:

!giskard worker --help

You’re all set to try Giskard in action. Upload your first model, dataset or test suite by following the upload an object page.