Little trouble with nested includes

I'm trying to get Frugally Deep working as a node, which depends on:

FunctionalPlus (single .h file)
Eigen (hundreds of .h files)
JSON (already in HISE but I think I still need to add it manually)

These are all header-only libraries, so they should work as src for third party nodes, right?

The current problem I have, is Fdeep is including

#include "fdeep/common.hpp"

And that file is including

#include <Eigen/Core>

Which returns no such file or directory, I assume because it's trying to include the whole folder.

Would I have to write an include.h file and include every single header file in that folder? If there's a better way, I'd love to hear it because I reaaaaaaally don't want to do that...

I've got the libraries in both the src folder:

And in the fdeep subfolder:

So I don't think it can't find the files, any help is much appreciated

Christoph Hart

@iamlamprey Not 100% if it helps, but try adding the ThirdParty/src folder to the HeaderSearch Paths in the projucer project (check the DspNetwork/Binaries/AutogeneratedProject.jucer first).

This way it will be able to resolve all those #include <xxx> statements.

Ok got a little bit further by adding the src path as an Additional Include Directory in VS but then ran into about 10000 syntax errors so I think I'll look for a different library...

d.healey

@iamlamprey You may only need to solve one of those errors to fix them all ;)

Christoph Hart

@d-healey yes but Eigen is a very complex and heavyweight library so it‘s not the easiest task to include it.

Yeah there doesn't really seem to be an agreed upon convention for getting TF working in C++, some people are using numpy/eigen, some are making their own wrappers like Frugally Deep and cppflow...

@Christoph-Hart Can third party nodes support prebuilt static libraries? Or does it have to be the source code? I might just have to bite the bullet and remake my autoencoder in pytorch since using TF in C++ is just a mess

KimiA

@iamlamprey Using Machine Learning models in C++ can be quite tricky but here's an open source project by CHOW-DSP that uses machine learning and juce to model a Klon Centaur pedal.

I believe it uses RT Neural which is meant for real-time audio, however, the process should be somewhat similar.

https://github.com/jatinchowdhury18/KlonCentaur
https://github.com/jatinchowdhury18/RTNeural

Hopefully, this helps with getting the implementation working.

All the best!

@KimiA Legend! I'll take a look, thank you :)

Okay think I have Tensorflow and CPPFlow building thanks for your help everyone!

Once I figure everything out and make sure it works properly I'll write up a quick post about it, in case anyone else here wants to mess around with Autoencoders (or god forbid try getting a GAN training)

Now onto the real endgame... how does one print to the HISE console from inside a Node?

why are the simplest things so difficult for me

// Parameter Functions -------------------------------------------------------
	
	template <int P> void setParameter(double v)
	{
		if (P == 0)
		{
			// This will be executed for MyParameter (see below)
			String s;
                        s = "test";
			jassertfalse;
			debugToConsole(hise::Processor::getProcessor(), s);
		}
	}

Christoph Hart

If you‘re creating a third party node you‘ll most likely end up debugging jn VS / Xcode directly so you can print and breakpoint directly in the IDE.

Just open the IDE project in the DSPNetwork/Binaries subfolder, it should have the target set to the HISE Standalone app that will launch when you start debugging the project.

@Christoph-Hart do I have to manually target HISE if this happens?

or just debug the HISE project instead?

Christoph Hart

@iamlamprey Yes, go to the Project properties and select the HISE x64 Debug.exe as Command:

If you then click on the Debug play button in VS it will launch HISE and as soon as you load a node from your DLL it will jump into the code and you can step through it and print stuff. DBG("Funky"); is your friend here.

@Christoph-Hart Got it, thank you :)

@Christoph-Hart What would be the best way to add JSON-loadability to a node? Is that a realistic ask or would I be better off trying to just make this a base processor instead of a node?

Christoph Hart

@iamlamprey there is no API for communication with Strings so that will be a bit tricky. What data do you want to communicate?

@Christoph-Hart Yeh that would explain the error.

I'm trying to load a tensorflow model, so the data would be both the model itself (layers & their hyperparameters) and the actual weights & bias of the network (the "trained" part). They're mostly doubles I think but there's some strings in there as well for the type of layer & activation functions

The goal is

audio sample -> [network] -> output, where [network] is a big ol bunch of matrix math that does all the blackbox stuff automatically

Not gonna lie I may have bit off more than i can chew here, I've been watching Cherno's C++ series and I'm starting to grasp some of it but man my brain hurts lol.

I think I'll have to just get familiar with the tensorflow C api, all these other libraries and dependencies seem to pile up errors, and I really only need a few pieces of tensorflow (conv1d, leakyrelu, layernorm etc)

Christoph Hart

@iamlamprey Do you need to load different JSON files on runtime or can you include them in the C++ source?

@Christoph-Hart Ideally different ones at runtime (loading different models), but I could probably get away with including all of the models, then just use a node Parameter to switch between them and hide that parameter from the end user

worst case scenario I can copy the weight values manually into a struct or something and write my own activations (those are pretty simple), then the only part left really is re-creating the convolution layer, which is probably the hardest part

Here's an example of a "trained" model (it's not actually trained, the weights just get randomized on initialization) if it helps explain it better

encoder_weights.json

Little trouble with nested includes

21

1.8k

10.3k

90.3k