Package index • kerasnip

Core Functions

Functions for generating and managing custom model specifications.

create_keras_sequential_spec(): Create a Custom Keras Sequential Model Specification for Tidymodels
create_keras_functional_spec(): Create a Custom Keras Functional API Model Specification for Tidymodels
inp_spec(): Remap Layer Block Arguments for Model Specification
remove_keras_spec(): Remove a Keras Model Specification and its Registrations

Registration Helpers

Functions for registering custom Keras components like loss functions, optimizers, and metrics.

register_keras_loss(): Register a Custom Keras Loss
register_keras_metric(): Register a Custom Keras Metric
register_keras_optimizer(): Register a Custom Keras Optimizer
keras_optimizers keras_losses keras_metrics: Dynamically Discovered Keras Objects

Model Inspection and Evaluation

Functions for summarizing, evaluating, and extracting information from trained Keras models.

compile_keras_grid(): Compile Keras Models Over a Grid of Hyperparameters
extract_valid_grid(): Extract Valid Grid from Compilation Results
inform_errors(): Inform About Compilation Errors
extract_keras_history(): Extract Keras Training History
extract_keras_model(): Extract Keras Model from a Fitted Kerasnip Object
keras_evaluate(): Evaluate a Kerasnip Model
axe_data(<kerasnip_model_fit>) axe_env(<kerasnip_model_fit>) axe_call(<kerasnip_model_fit>) axe_ctrl(<kerasnip_model_fit>) axe_fitted(<kerasnip_model_fit>): Butcher axe methods for kerasnip_model_fit

Custom recipe steps

Custom stpes for recipe which uses kerasnip models specifications

step_collapse(): Collapse Predictors into a single list-column
step_sequence(): Build a Sliding Window of Predictors for Sequence Models
step_lead(): Create a Lead Predictor

Post-Processing multi-output/multistep models

Adapters that present one output (or forecast step) of a multi-output/multistep fit as an ordinary single-output fit, so it can be post-processed with tailor/probably.

kerasnip_output_view(): View a Single Output of a Multi-Output kerasnip Fit
kerasnip_step_view(): View a Single Forecast Step of a Multistep kerasnip Fit
kerasnip_step_truth(): Recover Truth Values for a Multistep Forecast Step
kerasnip_add_tailor(): Attach a tailor Post-Processor to One Output or Step of a Multi-Output or Multistep Workflow