CODEWITHSUNDEEP
rkeras
David Díaz
David Díaz

Reputation: 171

How to properly save a keras model in R?

I haven't been able to save a keras model, I have been using this code:

save_model_hdf5(model, "F:/Models/1")

And I get this error message:

#Error in py_call_impl(callable, dots$args, dots$keywords):
#OSError: Unable to create file(unable to open file: name = 'F:\Modelos\1', error = 13,error message = 'Permission denied', flags = 13,o_flags = 302)

Upvotes: 0

Views: 1653

Answers (1)

Peter
Peter

Reputation: 2370

Python:

I use model.save(filename.hdf5) to save my models. Note that model is an object, e.g. created by model.compile(...). Find a full example here:

# Set up model
model = models.Sequential()
...
# Compile model
model.compile(...)
...
# Save model
model.save('C:/savepath/savename.hdf5')

It is possible to load the model again, e.g. to make predictions as outlined here.

# Load model
model = load_model('C:/savepath/savename.hdf5')

R:

In R, saving models works a little different:

# Set up a model
model <- create_model()
...
# Fit the model
model %>% fit()
...
# Save the model
model %>% save_model_hdf5("C:/savepath/savename.hdf5")

Or alternatively (without infix operator %>% which is not part of base R):

save_model_hdf5(model, "C:/savepath/savename.hdf5")

Upvotes: 2

Related Questions