Keras: ImageDataGenerator Epoch comprised more than `samples_per_epoch` samples

[StevenLOL]

UserWarning: Epoch comprised more than samples_per_epoch samples, which might affect learning results. Set samples_per_epoch correctly to avoid this warning.

I have around 20k training images,

        train_datagen = ImageDataGenerator(
                zoom_range=0.2,
                rotation_range=20,
            width_shift_range=0.1,
            height_shift_range=0.1,
                horizontal_flip=False)
...
...
model.fit_generator(
                train_datagen.flow(train_data,train_target,batch_size=40,shuffle=1),
                samples_per_epoch=2520,
                nb_epoch=400)

Ideally this should train with 2520 each epoch, however I gives 2507 on the 9th epoch every time. It says " UserWarning: Epoch comprised more than samples_per_epoch samples, which might affect learning results. Set samples_per_epoch correctly to avoid this warning."

[abdelrahmansaud]

I had this warning and solved it by having samples_per_epoch a number that can be divided by batch_size
e.g. batch_size 32 ... and samples_per_epoch 256

[MoyanZitto]

@abdelrahmansaud is right, this warning occurs because the number of samples in your training set cannot be divided envenly. You have 20K training images, perhaps you should set samples_per_epoch to 2500.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs, but feel free to re-open it if needed.