AI Model Training

Overview

Train custom AI models directly in the Flow Editor using your own data. This guide walks you through the complete process of setting up, configuring, and training an AI model.

Prerequisites

Before you begin, make sure you have already imported your data into the Data Module.

Required Nodes

To train an AI model, you need a minimum of three nodes:

1. Select Data: Define which data you want to work with
2. Train AI Model: Configure the AI model and training parameters
3. Preview Output: Check if your training was successful

Add these nodes one by one and connect them in sequence.

Step-by-Step Training Process

1. Set Up Your Flow

Navigate to the Flow Editor and select the Node Library. Add the three required nodes and connect them.

This is how your editor should look:

2. Select Your Data

Click on the Select Data node. You can skip the configuration for now and directly open the Data tab.

3. Verify Data Selection

To check if the file was selected successfully, attach a Preview Output node to your data selector. This will visualize the file that you just selected.

Tip: You can use Preview Output nodes anytime to see what your previous node outputs.

4. Configure the Train AI Model Node

Setup Tab

Start with the setup configuration:

1. AI Category: Select the category based on your data type (e.g., Tabular, Image, Text)
2. Task Type: Choose the specific task (e.g., Classification, Regression)
3. AI Model: Select the model you want to use

If you cannot find the AI model you're looking for, try selecting a different task type or AI category.

For example, to perform classification on tabular data:

Once done, click Save Configuration and move to the Input tab.

Input Tab

The Input tab defines the model configuration and is the core of your AI model setup.

Since you selected data beforehand, you'll see automatically selected feature columns and a dropdown for the target column.

Configuration steps:

1. Remove your target column from the feature columns (e.g., click the cross next to "Stage")
2. Select it as the target column in the target dropdown
3. Configure additional settings (optional)

Recommended: Enable hyperparameter tuning and select several settings. The system will automatically find the optimum values for you, resulting in a better model.

This is how the configuration should look:

Save the configuration.

Test Tab

The Test tab shows all input requirements that are necessary or optional.

Check that you have as many inputs as there are input requirements. If not, refer to your model's documentation to add what is necessary.

If everything is configured correctly, clicking on Select Data will generate files for the AI model with "pending" status.

In this example, we have training metrics and feature importance files:

5. Run the Training

Click Run to start the training process.

Execution logs will appear showing real-time updates about your flow progress:

Training Time: For tabular models, this happens quite fast. Larger datasets or heavier models can take hours.

6. View Results

Once training is complete, the logs will close and you'll see the results:

The pre-created files will now show as "ready". Click on them to view the results:

Understanding the Outputs

Artifacts

The training creates artifacts - files that are generated to use the model later. In this example:

• label_encoder: Encodes categorical labels
• model_weights: Contains the trained model parameters

Metrics and Analysis

You can use files like training metrics or feature importance to:

• Create plots in the Plots Module dashboard
• Analyze model performance
• Understand feature contributions

Next Steps

Use Your Trained Model

Use the artifacts in a separate inference flow to make predictions with your trained model.

Deploy as API

To trigger the training from outside the aicuflow tool, deploy the training flow to get an API endpoint.

Monitor Performance

Track your model's performance over time and retrain when necessary to maintain accuracy.