Working with Generators
Once you have created a synthetic or semi-synthetic generator, you can perform various actions to extend its functionality or create variations. This page covers the two main ways to work with existing generators: copying and continuing training.
Note: Continue training is only available for synthetic and semi-synthetic generators.
Overview
Both copy generator and continue training allow you to build upon existing generators, but they serve different purposes:
| Feature | Copy Generator | Continue Training (*) |
|---|---|---|
| Purpose | Create new generator with a different configuration | Extend training of existing generator |
| Source Dataset | Same as original | Same as original |
| Model Weights | Starts fresh | Starts from existing checkpoint |
| Settings Flexibility | Full flexibility - all settings can be changed | Limited - only training parameters |
| Training Time | Full training required | Usually shorter |
| Use Case | Configuration experimentation | Model improvement |
* = available only for synthetic and semi-synthetic generators
Copy Generator
Creates a completely new generator that copies the configuration from an existing generator but starts training from scratch.
How to copy a generator
Navigate to the generator view page and open the breadcrumb menu at the top.
Click the “copy generator” option.
You will now be able to change all generator settings and create a new one.
Continue Training
Creates a new generator that resumes training from an existing generator’s checkpoint, preserving the learned model weights and patterns.
How to continue training
Navigate to the generator view page and open the breadcrumb menu at the top.
Click the “continue training” option.
You will now be able to change some of the settings of the previous generator and create a new one preserving the learned model weights and patterns.
Which option to choose
Choose Copy Generator when:
- You want to test different model architectures
- You need to change table/column selections
- You want to experiment with different data treatment options
- You’re creating multiple independent variations for comparison
- The original configuration is good but needs significant changes
Choose Continue Training when:
- Your generator shows promise but could benefit from more training
- You want to fine-tune existing parameters slightly
- You want to preserve existing learned patterns while making minor improvements