Wow!!! it’s a CPU killer!!! on Dwarf, a pedalboard with Mod Auto Input, Neural Amp Modeler (loaded with nano-mdl), Mod IR Loader Cabsim (Loaded with 30K of waw IR) goes to 99.4% of CPU… unusable…
Without Cabsim it goes to 72.7%… unusable!!!
it is by design, NAM developers prefer to have this quality of sound without any kind of compromise.
LSTM based models can run at around 35% on a Dwarf, so basically half of the current WaveNet nano approach, if sound ends up “degraded” is a bit questionable.
The point of entry is getting the plugin to run, which is the case now.
Optimizations can still be made, but yeah with these kind of plugins the pedalboard is always going to appear a bit empty compared to other more complex setups.
It is a case of being smarter vs brute-forcing the problem. AI tech typically goes for the brute-forcing approach, throwing huge amounts of CPU at it first, then we optimize things by comparison until a good compromise is achieved.
I think we need to do an LSTM vs WaveNet blind test, as the NAM plugin supports both so it won’t be a case of different engines (like a possible AIDA-X/RTNeural vs NAM comparison)
8 Likes
I’m thinking about to try to profiling the entire section od/dist/amp/speaker to see if the result is good, maybe it can help to have enough CPU to use some dly/rev after that profile.
I saw that with AIDA-X the better way was to profile only the amp but I never tried to profile all this section on NAM, do you have any suggestions, hints?
Hi, I have a Duo and a Dwarf… can someone please explain how to load NAM models into the NAM virtual pedal? My Duo is version 1.13.1.3312, how do I get 1.13.2, or is that only for the Dwarf? And I downloaded a free NAM pack from Tonejunkie, so how do I get it into the NAM? Sorry, this is completely new to me…
EDIT: Okay, I think I figured it out. First of all, I have to manually update the Duo to version 1.13.2 as automatic updates for Duo are no longer supported. And then I should be able to load the NAM file using the File Manager to the correct folder (which requires updating to version 1.13.2).
I’m having trouble updating, but hopefully it works shortly!
Wavenet models will not work on the old Duo, not even nano models, poor thing cannot keep up.
it can still load LSTM-16 models, but as far as I know there are none of that type made for NAM right now.
The Dwarf can load wavenet nano models, with roughly 66% cpu usage so it doesn’t leave a lot of space for much else at the moment.
We are discussing internally possible optimizations and paths to take in order to get other plugins in the chain at the same time.
A mono cabinet IR loader that can do multi-threaded processing seems to be the way to go, you can already do some experiments for this by uploading cab IRs in the reverb folder and loading them with the x42 convolution plugin. this plugin is not really suitable for cabs though, dry/wet for one doesn’t make sense on cabs…
We will have news on this soon.
3 Likes
On the topic of CPU usage, we are experimenting with that idea of multi-threaded IR processing and just pushed a plugin to beta that does this (look for “Cabinet Loader” by MOD, no modgui yet)
Instead of maxing out the Dwarf CPU with this approach the CPU usage of NAM + IR ends up around 73%
The plugin adds around 42 samples of latency, around 0.87ms. Might be a good compromise for the (apparent) CPU gains.
EDIT: The previous sentence was wrong, we didnt test properly
I do not want to go too off-topic here, it is mostly to demonstrate we are working on it.
After a bit more checks and testing we will finalize the plugin and make a thread specific to it.
13 Likes
Thanks @falkTX for this plugin! Now I have more free CPU for other plugins 
For me it will be useful to have a off switch that stops taking CPU for this cabinet loader. I tend to have the same pedalboard with headphones or with an amp cabinet, and now with the Vibro plugin I could choose a bigger model if there is no cabinet.
Note: interesting, with this plugin loading an empty .wav reduces a lot the CPU use (like a real switch off), so it is different to the other IR loader cabsim plugin.
This has been brought up a few times, but it is a tricky question… if the plugin consumes different amounts of CPU when on vs off it can cause lockups or extremely high CPU usage that takes a while to recover from if we accidentaly enable a bunch of them at once.
I am not against the idea, but it needs some planning for when and how to safely disable some plugins.
That is expected, the engine for the plugin trims zeros at the end in order to save on processing. So your file likely ends up doing no processing at all if it is all silence.
1 Like
It’s cool it indeed saves a lot of CPU but it only loads the default IR (forward-audio_AliceInBones) and none of my 24 bit 48khz WAV IRs… Any idee?
A correction to this… the plugin doesnt actually add latency, our quick testing last time was flawed due to gain differences between different IR loader plugins.
Make sure to update it, there was an issue where non-48kHz mono files were loaded incorrectly.
Also just created a topic for this plugin MOD Cabinet Loader so lets have all discussion about it in there and leave this space for NAM, thanks!
3 Likes
The Neural Amp Modeler plugin is now more CPU efficient, based on the changelog.
Is it possible to load also standard NAM profiles on the Mod Dwarf?
as mentioned in Plugin Updates (Official Store) - #58 by falkTX
Same limitations regarding nano vs standard models still apply
So we still need the nano models. The ones labeled “standard” are extremely heavy, wouldn’t count on them ever loading on a Dwarf to be quite honest.
The real question is if we actually need such heavy models. AIDA-X shows that we can get quality neural amps without needing to spend so much CPU power, and NAM supports both lighter WaveNet and LSTM models. There are some issues training LSTM models on NAM side though, making direct comparisons a bit difficult, but we are slowly making progress on that.
8 Likes
Sorry guys if I´m asking a stupid question:
HOW did you load .nam-Files into the Modeler?
Drag-n-drop downloaded .nam´s into the "Neural Amp Modeler"´s
“– choose a NAM model –”-dialouge obviously
didn´t work 
thanks
the file manager on the bottom left is your entry point.
then there is a dedicated section/folder for NAM Models
1 Like
oops, I totally missed that out ^^
…thanks for your patient answer°
Dear MOD Device,
after running the NAM - I know, this is it!!! 
… that said:
I´m looking forward to get a hi-performance edition
of the Dwarf, which I absolutely will be willing to purchase - if you ever thinking of bringing em out.
I checked the NAM performing on my DUO X (no- and it nearly can handle 2 of them…
1 NAM makes the processor going up to about 43%,
2 of them giving 89% will time after time spike the processor to 97% -
which isn´t working stable, sadly.
alternatively,
I´m hoping for more .JSON´s for the AIDA X
JFYI… I´m in need of clean, dynamic (BASS) Amp models:
Hi-Watt, Sunn
^^
2 Likes
Hi there! I recently purchased the Mod Dwarf, and I’m really excited about it! However, I’m experiencing a minor issue: when I load the Neural Amp Modeler plugin and select a .nam file, the CPU immediately jumps to 100%. The Mod disconnects and restarts, automatically removing the plugin. During those few seconds before the device restarts, the sound is completely glitched.
I’ve tried inserting the NAM in an entirely empty pedalboard, and I’ve attempted setting the Buffer Size to 256, but the issue persists.
Am I doing something wrong? Could there be an issue with my Dwarf?
It’s worth noting that both the Firmware and the Plugin are properly updated to the latest version.
Thanks in advance for any help!
1 Like
Unless the model you use is of the type “nano”, it will probably be too heavy for the Dwarf to process.
1 Like
The architecture of the NAM is very heavy and not really suited to run on devices other than a powerful PC, unless you use one of the nano models which are special NAM models of much lighter weight.
NAM is not especially optimized as it’s normally meant to run on very powerful CPUs.
For running on embedded devices, you are far better off using the AIDA-X neural modeler which is using a way better optimized architecture and can offer an equivalent performance with a much smaller footprint.
4 Likes