-
Notifications
You must be signed in to change notification settings - Fork 0
Description
What part of the documentation is this about?
The vignettes section of the documentation.
What is the issue?
The documentation lacks a clear, comprehensive example of how to tune important training parameters like epochs, batch_size, optimizer, loss, and learn_rate. This may make it difficult for users to understand how to optimize these crucial aspects of the Keras training process within the tidymodels framework.
Suggested improvement
A new vignette should be created to provide a detailed, step-by-step tutorial on how to tune these fit_* and compile_* arguments. This guide should demonstrate how to:
- Define a
kerasnipmodel specification with tunablefit_*andcompile_*arguments. - Use
dialsandkerasnip's special parameter functions (optimizer_function(),loss_function_keras()) to create a hyperparameter grid for these arguments. - Run a standard
tidymodelstuning workflow to find the optimal training parameters. - Finalize the workflow and fit the model with the best parameters.
Additional context
Adding this guide would fill a significant gap in the documentation and provide users with a much-needed, reproducible example for a common and important task.