0
$\begingroup$

I am re-training a pretrained model VGG16.

In the last layers, im using two FC layers of size 2048 each, with dropout=0.5.

When i saved the model, The size of the model was found to be 2 GB (which is so huge). I have trained this model for 9 classes.(900 images)

Is there any way, where i could reduce the size of the model but the performance shouldn't be affected ?

The reason is i have to use this model in AWS cloud, where the systems are having 8GB ram. I am worried if their performance would be affected by such models.

$\endgroup$
0
$\begingroup$

You can change the type of your weight, you will loose precision but reduce the size of the weights. You can also do some regularisation during training to learn sparse weight matrix

You can read these two articles :

|improve this answer|||||
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.