whoops, meant to open this on my own fork 😉

@philipperemy I've forked your project and moved to python 3.12 because I want to explore the performance of the TCN in one of my own projects where I'm already using 3.12. I chose simply dropping the dependency on tensorflow-addons, which involved removing the option for WeightNormalization (I think for my project the other two normalization options are definitely sufficient).

If you want to support python 3.12 (either without weightnormalization or by annexing it from tensorflow-addons in some way) or if you want to stay at 3.10 for now but still introduce a parameter for residual depth (like I'm doing here), I'm happy to contribute.

Hey, I guess we can drop the dependency on tensorflow-addons. That's a good idea. We don't need to support weightnormalization anymore. Or if there's any other easy way.. Can you make a PR? It'd be appreciated!

I opened a PR: #266 if you have any remarks, let me know!