Add Tanh shrink activation function

[zaleslaw]

The TanhShrinkactivation function is missed in KotlinDL.

The desired PR addressing this issue should include:

• Implementation of activation class named as TanhShrinkActivation (you can take inspiration from the implementation of HardSigmoid as reference) added to the Activations.kt file
• Documentation of activation function
• JUnit tests in api module
• Support for export of activation function to JSON (see ModelSaver.kt)
• Support for import of activation function from JSON (see ModelLoader.kt)

P.S. If you want to take this ticket, please leave the comment below
P.P.S Read the Contributing Guidelines.

The reference implementation could be taken from tensorflow-addons

[femialaka]

I would like to work on this one, can you please send me a slack invite so I can discuss a few things, thanks in advance.

[zaleslaw]

No need special invites to the kotlin slack, just register with your e-mail. Also, you could free ask me here, in this issue

[femialaka]

When I was trying to run irisClassification in the examples I got the following error:

"No core dump will be written. Core dumps have been disabled. To enable core dumping, try "ulimit -c unlimited" before starting Java again"

Process finished with exit code 134 (interrupted by signal 6: SIGABRT)

I ran it on a Mac with an M1 chip I suspect this may be the cause?

[zaleslaw]

Hi @femialaka please see the following bug

The KotlinDL internally uses TF 1.15 runtime, which means you can use it only with the previous Mac without M1.

But I got different errors for the different combinations of OS and architectures. Could you please share with me the OS X version?

[femialaka]

@zaleslaw sure, its Big Sur version 11.5.2 (M1 2020), I'll try another machine then and get back to you.