CODEWITHSUNDEEP
tensorflowkeras
2Obe
2Obe

Reputation: 3710

validation_split argument differences

Due to the Train a Keras model documentation there is a validation_split argument in the fit method to set apart validation data. Now one can use the ImageDataGenerator object in the fit method which itself holds a validation_split argument. Now I am confused which parameter to set if I want to have say 20% of my data as validation data. Additionally: What happens if I only set any one of them solely or both together?

Upvotes: 0

Views: 136

Answers (1)

Gerry P
Gerry P

Reputation: 8092

the documentation for model.fit is here. The documentation clearly states the parameter validation_split is not supported when x is a dataset, generator or keras.utils.Sequence instance. So if you are using the ImageDataGenerator to provide the data you can not specify validation_split in model.fit.

Upvotes: 1

Related Questions