CODEWITHSUNDEEP
pythonmachine-learningneural-networkdeep-learningkeras
Oblomov
Oblomov

Reputation: 9635

Problems understanding keras logging output

Training my model model with 

 model.fit_generator(
    ...,
    verbose=1,
    ...
)

generates the following output: 

Epoch 1/5
64 128
128 192

  64/6400 [..............................] - ETA: 466s - loss: 0.0189 - mean_squared_error: 0.0189192 256

 128/6400 [..............................] - ETA: 253s - loss: 0.0163 - mean_squared_error: 0.0163

what does 

 64 128
128 192

mean? It certainly has to do with the batch_size of 64, but what do these specific output lines represent and why are they printed?

Upvotes: 2

Views: 588

Answers (1)

Marcin Możejko
Marcin Możejko

Reputation: 40516

In case when you set the verbose=1 option - Keras is printing the information about loss, estimated epoch time remaining and a metrics provided during model compilation after each batch. Usually it's printed dynamically - so you could only see a one bar for each epoch - but for some reason your Python is printing it new line batch after batch.

Upvotes: 2

Related Questions