compile problem

[Violet]

Hi, I am new in applying the neural network models, I am using linux Ubuntu , and the versions that are installed are:

tf_config()
TensorFlow v2.16.1 (~/miniconda3/lib/python3.12/site-packages/tensorflow)
Python v3.12 (~/miniconda3/bin/python3.12)

so I am doing the examples that are represented in the reference "Deep Learning with R, second edition" in chapter 10, all the steps were good, so when I get to the code of :

model %>%

• compile(optimizer = "rmsprop",
• loss = "mse",
• metrics = "mae")

it gives me the follwoing error:

Error in UseMethod("compile") : 
  método não aplicável para 'compile' aplicado a um objeto de classe "c('keras.src.models.functional.Functional', 'keras.models.functional.Functional', 'keras.src.ops.function.Function', 'keras.ops.function.Function', 'keras.src.models.model.Model', 'keras.models.model.Model', 'keras.src.backend.tensorflow.trainer.TensorFlowTrainer', 'keras.backend.tensorflow.trainer.TensorFlowTrainer', 'keras.src.trainers.trainer.Trainer', 'keras.trainers.trainer.Trainer', 'keras.src.layers.layer.Layer', 'keras.layers.layer.Layer', 'keras.src.backend.tensorflow.layer.TFLayer', 'keras.backend.tensorflow.layer.TFLayer', 'keras.src.backend.tensorflow.trackable.KerasAutoTrackable', 'keras.backend.tensorflow.trackable.KerasAutoTrackable', 'tensorflow.python.trackable.autotrackable.AutoTrackable', 'tensorflow.python.trackable.base.Trackable', 'keras.src.ops.operation.Operation', 'keras.ops.operation.Operation', 'keras.src.saving.keras_saveable.KerasSaveable', 'keras.saving.keras_saveable.KerasSavea

Can someone help to detect where might be the error to solve the problem, thx.

[t-kalinowski]

This was an error similar to this in an earlier version of keras, which has been fixed now. Can you check if you're using the latest R package release?

Also, you may want to install keras3, the next version of Keras. Many of the examples from the book should continue to work unmodified.

Finally, consider using a virtualenv instead of miniconda - it lets you side-step any incompatibilities between conda and non-conda binaries. This should be all that's needed:

install.packages("keras3")
reticulate::install_python()
keras3::install_keras()