These methods allow butcher::butcher() to reduce the memory footprint of
fitted kerasnip model objects. The Keras model itself (stored as raw bytes
in $fit$keras_bytes) is always preserved so that predict() continues
to work after butchering.
The main saving comes from axe_data(), which removes the training history
object ($fit$history). For long training runs this can be several MB.
Usage
# S3 method for class 'kerasnip_model_fit'
axe_data(x, verbose = FALSE, ...)
# S3 method for class 'kerasnip_model_fit'
axe_env(x, verbose = FALSE, ...)
# S3 method for class 'kerasnip_model_fit'
axe_call(x, verbose = FALSE, ...)
# S3 method for class 'kerasnip_model_fit'
axe_ctrl(x, verbose = FALSE, ...)
# S3 method for class 'kerasnip_model_fit'
axe_fitted(x, verbose = FALSE, ...)