KotlinDL 0.4 Is Here! NoTop models in the ModelHub...

# kotlindl

[roman.belov]

KotlinDL 0.4 Is Here! NoTop models in the ModelHub, new layers and activation functions, better Image Preprocessing DSL and just see the GIF below! Read all in the blogpost: https://blog.jetbrains.com/kotlin/2022/06/kotlindl-0-4-is-out-with-pose-detection-api-[…]or-object-detection-and-efficientnet-for-image-recognition/

[zaleslaw]

Cool gif!

[Hampus Londögård]

Is NLP, tabular etc coming to KotlinDL?

[zaleslaw]

Probably, NLP support is the next step, Classification task could be solved with Dense layer aka Logisitic Regression, see the example in exsmples/ml. For classic ML there will be a separate library

[Hampus Londögård]

I know, but I’m thinking Transformers & RNNs.

For now I’m using DeepJavaLibrary which is pretty nice, using simple Kotlin wrappers it’s pretty ergonomical too.

[zaleslaw]

Djl is quite good library!

[Hampus Londögård]

Yes, but it’s much more heavily focused on simply inference. Compared to KotlinDL that is! 🙂

[zaleslaw]

Yes, Djl has it's known benefits and limitations and of course purposes to be integrated well with AWS and SageMaker

[Hampus Londögård]

I’m excited for KotlinDLs future as it’s a much more “wide” focus (training+inference+some preprocessing).

Is there any idea how big effort it’ll be to move to TF 2.0 as that API stabilize?