A test with static neural network could be really useful

jatinchowdhury18 / RTNeural-NAM

Experimenting with Neural Amp Modeler models implemented with RTNeural

BSD 3-Clause "New" or "Revised" License

9 stars 1 forks source link

A test with static neural network could be really useful #5

Open aexposit opened 3 days ago

aexposit commented 3 days ago

Hi while creating the neural network dynamically gives flexibility, at the end probably more than 95% of the NAM models are NAM "STANDARD" (the network with up to 512 dilation rates and 2 layers). Is there any better performance if just NAM "standard" is used in your code and the network model is fixed ? Just out of my curiosity...

Thx a lot

rossbalch commented 3 days ago

Personally I would favour flexibility. There are situations where you might want to save on CPU with lighter models where you don't need as many parameters. Or maybe you need a heavier model is some specific circumstances.