NeuralAmpModelerCore

NeuralAmpModelerCore copied to clipboard

Improve activations' implementations using Eigen better

3

Tanh uses a lot of compute (hence "fast tanh"). But this e.g.: https://github.com/sdatkinson/NeuralAmpModelerCore/blob/846968710a670d662b15e449edba852d747d748e/NAM/activations.h#L75-L81 should be able to be implemented more idiomatically with Eigen. It would also be nice for any...

sdatkinson

TPU and GPU inference support

2

Are there any plans to support inference on some accelerators? Let's say use ONNXRuntime or TensorRT to free up CPU resources.

novikov-alexander

Following on from #125, this issue is to track improvements to the code for [gating activations in WaveNet layers](https://github.com/sdatkinson/NeuralAmpModelerCore/blob/e181f61efb8d05d34add45b5eecb3893ff21177c/NAM/wavenet.cpp#L38-L48).

sdatkinson

25% performance improvement inlarge wavenet models.

1

I'd like to offer you a piece of work that I have recently been doing for the ToobAmp project. The net benefit: a 25% performance improvement for the most commonly...

rerdavies

I'm just putting this out here in case it's something you might consider merging @sdatkinson. Although i expect it would be a rather controversial change. I find the code style...

olilarkin

And const_iterator rather than iterator

olilarkin

olilarkin

I haven't properly checked if this is the case here, but using Eigen types as function parameters can result in unwanted temporary copies in some cases. The Eigen::Ref class exists...

olilarkin

See http://eigen.tuxfamily.org/dox-3.2/group__TopicStructHavingEigenMembers.html and act accordingly.

sdatkinson