FileNotFoundError while training the model

Neural Modelling
#22

@spunktsch i dont mind at all just not home at the moment ? Would you like the input target and json file?

#23

sure, that also works.

#24

@spunktsch https://drive.google.com/drive/folders/1c2H9M7zCwadeyd3zFEGG3ozm4FQhCTkk?usp=sharing

hope this works!

#25

There is a lot of help requests for this “FileNotFoundError”. Could it be possible to update the ypnb to assign the correct name to the folder?

3 Likes
#26

I must say that at this point it is to dificult for me to use Aida X Model Trainer, it is made for programmer, not a average person. I’v spent small amount of time reamping and most time trying to go to the third step - still" [FileNotFoundError while training the model]

1 Like
#27

my solution to this was to record a stereo track. Essentially re-recording input.wav along with the resultant audio. I split them afterward in Audacity (takes a few seconds). But this resolved my alignment issues and my ESR went from ~0.99 down to reasonable ranges (~0.001)

There’s a pedalboard link for how I recorded here: Modelling: best practices & conventions - #20 by malfunction54

#28

@malfunction54 what kind of sound were you capturing ? clean , dirty , high gain? because the best result i could get was 0.2 with a high gain amp ! had the dame same volume and no latency. looking forward to the future of this though ! just hope it doesn’t eat up a montrous amount of dsp on the dwarf!

1 Like
#29

So far, I’ve just done clean amp capture. I’ll be doing lightly driven and higher gain next. I’ll see if my technique holds up.

1 Like
#30

@malfunction54 great !please if you don’t mind keepin us updated!

1 Like
#31

Even then you have to use the “bugged” pynb to actually train a model, which is frustrating for non-tech persons to mess with the “-0” in the folder name, who is in charge of fixing that?

#32

I am unsure of who is ultimately responsible for the cloud-based training app, but I think @spunktsch is somehow involved with the project. Is that correct?

#33

yeah. It’s @madmaxwell and me. See here Aida DSP - AIDA-X

I’d also like to have a more accessible training with a dedicated frontend guiding you through the process.
But resources are pretty limited at the moment. We have our dayjobs and this became more of a passion project.

2 Likes
#34

I am not fluent in colab stuff, but I noticed that in step 2

model_dir = f"/content/Automated-GuitarAmpModelling/Results/{file_name}_{config_file}-{skip_con}"

Where skip_con can be 0 or 1

While in step 3

model_dir = f"/content/Automated-GuitarAmpModelling/Results/{file_name}_{config_file}"

1 Like
#35

thanks @Zavorra do you mind posting the issues to github?
So @itskais and @madmaxwell can track the bugs down.

#36

Done

1 Like
#37

@spunktsch by any chance did you try the files i sent you?XD

#38

i trained a clean sound from a Hot rod deville , that sounds exactly the same , woooww :exploding_head:

BUT , the problem now is that the trainer does not works FILE NOT FOUND ERROR ,
i don’t know much about code to fix this,
can you help us please

now im trying a dirty “clean” sound of the same amp, but i have a couple of weeks that i stoped to work on it because nobody of the Aida X team answers :man_shrugging:t2::man_shrugging:t2:

1 Like