GenSynth Documentation

Fine-tuning a Model

During fine-tuning, a model cycle produced by Generative Synthesis is trained further to improve performance.

To fine-tune a model, select Fine-Tune from the View Details image11.png drop-down for your model. This will give you the option to create a new and subsequent job from your optimized model.

image43.png

You may optionally enter the number of iterations you want and/or select a learning rate type from the drop-down and click Start. If the number of iterations is left blank, GenSynth will automatically learn the best settings.

Additionally, when fine-tuning, you can select whether you want to fine-tune by continuing from the "best" iteration of a cycle (as judged by performance on the validation dataset), or the "last" iteration of a cycle. Training from the best iteration may help avoid overfitting the data. GenSynth uses the best iteration by default.

The new job will show in the History tab and in the image15.png menu for the model and on the Home screen while it is running.

image45.png