How to load two networks in a gpu and to get predictions

General Discussion
,
#1

Hello.
Thank you for all your answers.

I trained two different networks for one problem, and saved them one by one (here I did not calculate and save the predictions because the predictions are too large).
I want to load these networks on a gpu, calculate the predictions simultaneously for the same input, and compare with each other…

I wrote the program something like

model2=tf.keras.models.load_model(path2)
model1=tf.keras.models.load_model(path1)
pred1=model1.predict(input1)
pred2=model2.predict(input1)

I can get the pred1 and pred2, however the results look like the same…
I hope you know the resolution

Best wishes
Yu

#3

@Yuichiro_Morishita,

If the predictions from both models are the same, it is most likely because they have learned similar patterns and produce similar outputs for the same input.

Thank you!

#4

Thank you
I checked the networks gave different predictions when separately executed.

#6

Hello
The above problem is due to my misunderstanding…
The program above loads the networks separately, and I can get each of predictions