Skip to main content

Your First Xircuits Workflow

The following tutorial will walk you through a basic Xircuits run. You will open an existing Xircuits workflow that trains a model, learn how to modify the parameters for your use case, and then run it.

Prerequisites

For this tutorial, ensure that:

  1. You have a Python version that is compatible with Xircuits (Python 3.8+).
  2. You already have the xai_learning related libraries installed. You can do this by pip install xircuits[learning] or by pip install -r xai_components/xai_learning/requirements.txt.
  3. Ensure that you have the examples folder in your working directory. After Xircuits is installed, you can download it by running xircuits-examples in the command line.

Opening an existing Xircuits File

New Xircuits users are recommended to start with premade Xircuits workflows created by developers. The task would be to change the parameters for their use case. In Xircuits, we have created a handful of workflows for you to try in the examples directory.

For this tutorial, open KerasTrainImageClassifier.xircuits in the example directory.

Viewing and Interpreting the Xircuits Canvas

Once the Xircuits canvas is open, you will see the Xircuits Interface.

A Xircuits workflow starts from the Start node to the Finish node. The flow sequence is indicated by the flowing blue links connecting the ▶ ports. A component node may also have data ports, which indicates how a component can receive data and parameters to configure their behavior, as well as pass data to the next nodes.

If you are unsure what a component does, you can click on the component information tooltip to display component information. For Python savvy users, you may open the component code via context menu (right-click > Open Script).

Explore the KerasTrainImageClassifier workflow. Note the component sequence.

Modifying the Xircuits Workflow

As previously mentioned, a Xircuits User will usually work on a premade workflow and adjust the parameter components for their use case. To modify Literal Components such as Literal Int or Literal String, simply double click on the component, and you will be prompted to change the value.

Video

For more advanced users, you may add more component nodes from the Xircuits component tray, or by dropping a link from a port.

Refer to the TrainImageClassifier component. Modify the number of training_epochs to 3.

Saving and Compiling a Xircuits Canvas Workflow

Once you are satisfied with your workflow, you can save and compile the Xircuits workflow. A Python script with the same filename as the .xircuits canvas will be generated.

Press Save then Compile, then verify that the Python script has been generated.

Running the Workflow

To run the workflow, press the Save & Compile button. It will prompt you to save the workflow if you haven't. Select the Python kernel to run it.

Press Save then Compile button. You should get the following output:

Details
======================================
__ __ ___ _ _
\ \ \ \/ (_)_ __ ___ _ _(_) |_ ___
\ \ \ /| | '__/ __| | | | | __/ __|
/ / / \| | | | (__| |_| | | |_\__ \
/_/ /_/\_\_|_| \___|\__,_|_|\__|___/

======================================

Xircuits is running...


Executing: ReadDataSet

Executing: TrainTestSplit
Split Parameters:
Train Split 0.8
Shuffle: True
Random State: None

Executing: Create2DInputModel
x_shape=(28, 28, 1)
y_shape=10

Executing: TrainImageClassifier
Epoch 1/3
438/438 [==============================] - 17s 14ms/step - loss: 0.4883 - accuracy: 0.8467
Epoch 2/3
438/438 [==============================] - 5s 11ms/step - loss: 0.1591 - accuracy: 0.9521
Epoch 3/3
438/438 [==============================] - 4s 10ms/step - loss: 0.1144 - accuracy: 0.9651

Executing: EvaluateAccuracy
{'loss': '0.07967730611562729', 'accuracy': '0.975803554058075'}

Finish Executing

Congratulations you have successfully ran your first Xircuits workflow! From here, you should be able to run any Xircuits workflow examples. Jump to the Examples section if you want to see the various workflows made by our engineers and open source contributors.

Commonly Faced Errors:

Details
ModuleNotFoundError

You have not installed the needed libraries for that specific component library! For this tutorial, ensure that you have run the prerequisites.

Please connect all the nodes before running.

The Xircuits workflow is not complete! Please ensure the blue links from the Start node to the Finish node is completely connected.