Hi there! I’m running the training inthe collab notebook, and up until and including step 2 everything works fine, the training completes.
When I get to the step 3 and try to run model evaluation, I’m getting the following error
FileNotFoundError: [Errno 2] No such file or directory: '/content/Automated-GuitarAmpModelling/Results/Preamp_LSTM-16/training_stats.json’
Although all the files are there, in the path indicated!
Any ideas on what I am doing wrong?
that happened to me, too. I just renamed the drive folder to what the script expected and re-ran the cell. I have other issues, though, that may be related.
That’s the problem: there’s no discrepncy between the real and the expected file names!!!
UPD: Oh wait, there’s an extra ‘0’ at the end of the file that I didn’t notice! Thanks for the hint!
I saw the exact same issue. My result was an ESR of around 0.9 or 1.0, which ended up in silence in AIDA. I’m doing something wrong 
Well, I’ve got results - thanks to your tip to rename the folder! - and keep experimenting.
What is ESR anyway?
Google gives me:
I have no idea. I couldn’t find anything different than what you found. I’m guessing it’s a value specific to AIDA’s algorithm.
its simply the error signal ratio.
If you want to deep dive: Applied Sciences | Free Full-Text | Real-Time Guitar Amplifier Emulation with Deep Learning
Thanks. So, the less the better. Good, then I do it right.
Hi everyone ! sorry i’m a nood concerning all this technicel stuff!
I am really looking forward to this new tech !
Followed all the instructions on the video and the website but i keep getting this
can anyone help me please
step 1 result
Input file name: /content/drive/MyDrive/aida/input.wav
Target file name: /content/drive/MyDrive/aida/target.wav
Input rate: 48000 length: 8653368 [samples]
Target rate: 48000 length: 8653368 [samples]
Preprocessing the training data…
step 2 result
args.device = MOD-DWARF
args.file_name = aida
args.input_size = 1
args.hidden_size = 16
args.unit_type = LSTM
args.skip_con = 0
existing model file found, loading network… continuing training…
16% 63/400 [01:40<07:26, 1.32s/it]validation patience limit reached at epoch 64
16% 63/400 [01:44<09:16, 1.65s/it]
done training
testing the final model
testing the best model
finished training: aida_LSTM-16-0
Training done!
ESR after training: 0.9973483085632324
step 3 result
FileNotFoundError Traceback (most recent call last)
in <cell line: 13>()
11
12 model_dir = f"/content/Automated-GuitarAmpModelling/Results/{file_name}_{config_file}"
—> 13 model_path, esr = extract_best_esr_model(model_dir)
14 model_data = miscfuncs.json_load(model_path)
15 model = load_model(model_data).to(device)
/content/Automated-GuitarAmpModelling/colab_functions.py in extract_best_esr_model(dirpath)
91 def extract_best_esr_model(dirpath):
92 stats_file = dirpath + “/training_stats.json”
—> 93 with open(stats_file) as json_file:
94 stats_data = json.load(json_file)
95 test_lossESR_final = stats_data[‘test_lossESR_final’]
FileNotFoundError: [Errno 2] No such file or directory: ‘/content/Automated-GuitarAmpModelling/Results/aida_LSTM-16/training_stats.json’
step 4
FileNotFoundError Traceback (most recent call last)
in <cell line: 60>()
58
59
—> 60 shutil.copyfile(os.path.join(model_dir, ‘model_keras.json’), os.path.join(‘/content’, model_filename))
61 files.download(os.path.join(‘/content’, model_filename))
62
/usr/lib/python3.10/shutil.py in copyfile(src, dst, follow_symlinks)
252 os.symlink(os.readlink(src), dst)
253 else:
→ 254 with open(src, ‘rb’) as fsrc:
255 try:
256 with open(dst, ‘wb’) as fdst:
FileNotFoundError: [Errno 2] No such file or directory: ‘/content/Automated-GuitarAmpModelling/Results/aida_LSTM-16/model_keras.json’
I would love sole help please!!!
It seems that for some strange reason the actual name of the Google drive folder where the training data are stored is
Automated-GuitarAmpModelling-0
While the script is looking for
Automated-GuitarAmpModelling
Simply rename it (removing the “-0” suffix and should work.
By the Way, your “ESR” (a measure of how the model is distant from the target) is almost 1, which is too hight (should be close to 0.05, I believe).
If this is not because of the folder name you could have some time misalignment between input.wav and target.wav?
Thank you for your help! the folder name was the problem!
but i cant’ seem to get a great esr! will try and experiment more
Thank you again!
@spunktsch neural dsp plugins!! which i already cloned on my quad cortex!! it’s a good reference for my ears!!
make sure you have gate, all effects and the cab sim disabled. Try to match the target to (roughly) the same peak db as the input. The super high gain stuff of some plugins is difficult to model.
@spunktsch okok ! so i cant capture a rig containing amp and cab? i found a capture here that sound’s really great so i’m sure this little pluging will have a lot to offer in the coming months!!
do you know why this keeps showing by any chance
35% 139/400 [28:05<51:25, 11.82s/it]validation patience limit reached at epoch 140
35% 139/400 [28:18<53:09, 12.22s/it]
thank you for your time
this is totally normal. It stops when the model is already the best version.
(should be a easier to understand prompt on our side of the training) @itskais .
And you definitely can capture amp+cab. This needs another dataset and a custom setup of the train/val/test ratio. This is likely implemented in the next major update.
@spunktsch ok great thank you !! i will wait for the next major update ! it will probably be better and more accurate and ease of use!! thank you for your time!! for the moment i cant get any good results with the amp and cab setup! trying only amp !
@spunktsch . Just an update ! Did everything! Wave marching . Around -6 db for the input and target ! Did only the amp capture and not the amp and can. Better result but still nowhere close !! Anyway i think that for the moment it has trouble with high gain amps . ! 0 late’cy between the input and the target ! No worries though i’m sure that this will be something killer soon gives me time to buy the mod dwarf;)!.
Just wanted to keep you up to date thank you for your help!
yeah - it definitely needs work on the high gain stuff when using software. Had less issues on the JCM900 SL-X at my studio. Do you mind to share what you used?
I’d like to look into to it and see whats causing the issue.
