Exported notebook has internal dependency

Report bugs
#1

Hello all,

I’m trying to use exported notebook with Colab for testing multiple models at once before picking an official model for deployment. I found out a weird issue as follows:
The exported notebooks are sometimes packed with internal dependencies (ei_tensorflow.training) from EI causing trouble to run it on local or other platforms other than on EI.

What is the purpose of this dependency? Could you guys at EI fix this inconsistency? Because I don’t want to do a workaround removing it every time.

Thank You!

#2

Hi @duyanh-y4n, thanks for your feedback!

To help me track down this bug, what type of model are you seeing this for?

Warmly,
Dan

#3

Hi @dansitu,
I’m doing an audio recognition model with 1D and 2D CNN.
Thanks!

#4

So I think (from the top of my head) that these are introduced when you enable data augmentation. If you set it to off the notebook should be clean.

#5

Hi @janjongboom,
I’ve reactivated data augmentation but there is no internal dependency. So my experience so far is it appears just occasionally.