Unable to use AIDA-X on Google colab anymore

[danimourinho]

I used Google Colab to train models for AIDA-X in the past, but today I encountered errors in the 0.DEPS section. After displaying the errors, Colab prompts me to restart the runtime. However, even after restarting, the same issue persists. If I skip this section, all subsequent steps also result in errors. Do you know if this is a temporary issue? I’ll share the 0.DEPS messages below

"WARNING: Your environment has PyTorch 2.0.1+cu117 and CUDA 11.7. This environment is not supported.
Proceeding to install required dependencies…
Found existing installation: torch 2.0.1+cu117
Uninstalling torch-2.0.1+cu117:
Successfully uninstalled torch-2.0.1+cu117
Found existing installation: torchvision 0.15.2+cu117
Uninstalling torchvision-0.15.2+cu117:
Successfully uninstalled torchvision-0.15.2+cu117
Found existing installation: torchaudio 2.0.2+cu117
Uninstalling torchaudio-2.0.2+cu117:
Successfully uninstalled torchaudio-2.0.2+cu117
WARNING: Skipping tensorflow as it is not installed.
WARNING: Skipping tensorboard as it is not installed.
Looking in links: https://download.pytorch.org/whl/torch_stable.html
Collecting torch==2.0.1+cu117
Downloading https://download.pytorch.org/whl/cu117/torch-2.0.1%2Bcu117-cp311-cp311-linux_x86_64.whl (1843.9 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.8/1.8 GB 223.4 MB/s eta 0:00:00
Collecting torchvision==0.15.2+cu117
Downloading https://download.pytorch.org/whl/cu117/torchvision-0.15.2%2Bcu117-cp311-cp311-linux_x86_64.whl (6.1 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 6.1/6.1 MB 159.4 MB/s eta 0:00:00
Collecting torchaudio==2.0.2+cu117
Downloading https://download.pytorch.org/whl/cu117/torchaudio-2.0.2%2Bcu117-cp311-cp311-linux_x86_64.whl (4.4 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 4.4/4.4 MB 164.4 MB/s eta 0:00:00
Requirement already satisfied: filelock in /usr/local/lib/python3.11/dist-packages (from torch==2.0.1+cu117) (3.18.0)
Requirement already satisfied: typing-extensions in /usr/local/lib/python3.11/dist-packages (from torch==2.0.1+cu117) (4.12.2)
Requirement already satisfied: sympy in /usr/local/lib/python3.11/dist-packages (from torch==2.0.1+cu117) (1.13.1)
Requirement already satisfied: networkx in /usr/local/lib/python3.11/dist-packages (from torch==2.0.1+cu117) (3.4.2)
Requirement already satisfied: jinja2 in /usr/local/lib/python3.11/dist-packages (from torch==2.0.1+cu117) (3.1.6)
Requirement already satisfied: triton==2.0.0 in /usr/local/lib/python3.11/dist-packages (from torch==2.0.1+cu117) (2.0.0)
Requirement already satisfied: numpy in /usr/local/lib/python3.11/dist-packages (from torchvision==0.15.2+cu117) (2.0.2)
Requirement already satisfied: requests in /usr/local/lib/python3.11/dist-packages (from torchvision==0.15.2+cu117) (2.32.3)
Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in /usr/local/lib/python3.11/dist-packages (from torchvision==0.15.2+cu117) (11.1.0)
Requirement already satisfied: cmake in /usr/local/lib/python3.11/dist-packages (from triton==2.0.0->torch==2.0.1+cu117) (3.31.6)
Requirement already satisfied: lit in /usr/local/lib/python3.11/dist-packages (from triton==2.0.0->torch==2.0.1+cu117) (18.1.8)
Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.11/dist-packages (from jinja2->torch==2.0.1+cu117) (3.0.2)
Requirement already satisfied: charset-normalizer<4,>=2 in /usr/local/lib/python3.11/dist-packages (from requests->torchvision==0.15.2+cu117) (3.4.1)
Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.11/dist-packages (from requests->torchvision==0.15.2+cu117) (3.10)
Requirement already satisfied: urllib3<3,>=1.21.1 in /usr/local/lib/python3.11/dist-packages (from requests->torchvision==0.15.2+cu117) (2.3.0)
Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.11/dist-packages (from requests->torchvision==0.15.2+cu117) (2025.1.31)
Requirement already satisfied: mpmath<1.4,>=1.1.0 in /usr/local/lib/python3.11/dist-packages (from sympy->torch==2.0.1+cu117) (1.3.0)
Installing collected packages: torch, torchvision, torchaudio
Successfully installed torch-2.0.1+cu117 torchaudio-2.0.2+cu117 torchvision-0.15.2+cu117

WARNING: The following packages were previously imported in this runtime:
[nvfuser,torch]
You must restart the runtime in order to use newly installed versions.

ERROR: Could not find a version that satisfies the requirement tensorflow==2.11.0 (from versions: 2.12.0rc0, 2.12.0rc1, 2.12.0, 2.12.1, 2.13.0rc0, 2.13.0rc1, 2.13.0rc2, 2.13.0, 2.13.1, 2.14.0rc0, 2.14.0rc1, 2.14.0, 2.14.1, 2.15.0rc0, 2.15.0rc1, 2.15.0, 2.15.0.post1, 2.15.1, 2.16.0rc0, 2.16.1, 2.16.2, 2.17.0rc0, 2.17.0rc1, 2.17.0, 2.17.1, 2.18.0rc0, 2.18.0rc1, 2.18.0rc2, 2.18.0, 2.18.1, 2.19.0rc0, 2.19.0)
ERROR: No matching distribution found for tensorflow==2.11.0"

[danimourinho]

@spunktsch are you aware of this issue? Any chance to look and find a fix? I forgot to mention that i’ve used the latest versions of Firefox and Chrome. Thanks in advance!

[spunktsch]

yup, old issue that’s popping up: just change the version tensorflow==2.11.0 to 2.12.0 or the one that’s available.

[danimourinho]

Thanks! This fixed the error. However, I’m still encountering some strange issues. I’m trying to capture new presets for the Neural DSP Darkglass Ultra, which I’ve successfully done in the past. For example, when I use old training files from 2023 (Halogram Darkglass - Google Drive), the training works perfectly without any issues. But when I use any new training files recorded today (exactly the same way I did previously), I’m getting terrible results. I’ve noticed that the DRY and PREDICTED signals sound identical, and with TARGET and DIFFERENCE SIGNALs, I get a mixture of the Dry and Target. When I upload a DI sample to test the capture, the DRY and PREDICTED SIGNALS sound identical. Here’s an example of the training files where I’m getting these strange results (Parallax_NoCab - Google Drive). What am I doing wrong?

[pilal]

@danimourinho

(Parallax_NoCab - Google Drive ).

I did a training with this set of files. I got a terrible score (ESR after training: 0.9643202424049377), a very clean sound with a very low output signal. I guess your target file signal is to low. Did you already make good captures of that type of sound before ? I never succeeded at capturing my weirdest pedals unfortunately.

[danimourinho]

@pilal Thanks for your reply! Yes, I’ve captured this plugin before without adjusting the volume levels and got acceptable results. However, for some reason, I’m now getting poor results. I just checked and compared the waveforms—they look quite similar, and the target even appears louder than the input. I honestly have no idea what I’m doing wrong!

[pilal]

I did another try using GRU instead of LSTM. I got something better, but far from faithful to the target file. I gave a try to NAM and got a very good result. Nam is using wavenet by default and I think that’s the key.
Anyway you can try GRU training easily with my fork of the Aida trainer Colab. AIda-x model trainer fork

[danimourinho]

Thank you very much, i will try Nam and see if i can get a good result with the nano option to be able to run it in the dwarf. I see we can do the capture through the new tonehunt website. Do you recommend it?

[Funkeq]

Maybe your problem somehow connected to mine:

No succes in retraining NAM caputres for AIDA-X. Help needed

Thank you all for your help. I will look if I can get it to work within the next days. @spunktsch do you mean that the input.wav file provided with the AIDA-X colab is not a good choice for training AIDA-X models?

[pilal]

I did two trainings with the new tone3000 website. The results were good. The possibility to use your own custom pair of input/output files make it easy to use files prepared for Aida.