NAM: Neural Amp Modeler

[2dor]

So how many epochs do you usually go for?

It depends on what I'm training: if it's hi-gain stuff, I go with 1000 epochs at a minimum combined with a lower learning rate & decay (more recently, leveraging my own take on DLC86's TTS sequence). For stuff that converges easier maybe less but leverage the early-stop feature where training halts when a certain ESR is reached (I might go 0.005 or lower depending on how patient I am & how much stuff I have in the pipeline).
I usually do a quick run through the tones I'm training to gauge how they fare & take it from there.

 

[Progressive_disclosure]

Just did a new batch of captures of my ENGL Fireball 25, this time using the internal (resistive) load.

One capture also has the clean channel with a Wampler Euphoria in front. Any tips to get even better captures are welcome!

Link: https://www.tone3000.com/tones/engl-fireball-25-internal-load-37275

Now I’ve added new captures with a 1x12 V30 as well, they’re in the same pack. They sound much better then with the resistive load!

Also did a hi gain one with a Precision Drive in front.

 

[[Nathan]]

Is xSTD still the somewhat "secondary" standard that people are using? I can't find the values but I vaguely recalling pumping something into Tone3000 to alter the standard training values. If xSTD is still the one can someone shoot me those values, and if there's something "better" then let me in on it!

I'm doing a bunch of capturing now and am going to look at the latest ToneX / NAM but I figure if one of these xSTD type ones is a notch better then its worth throwing it into the mix.

*EDIT*
I found the xSTD stuff. I was searching for DLC but I think he's gone and thats why it wasnt coming up