ModuleNotFoundError

[ghost]

I got the following error. Please let me know how to resolve it.

Traceback (most recent call last):
File "C:\Users\Users\desktop\Music-Source-Separation-Training-main\train.py", line 515, in
train_model(None)
File "C:\Users\Users\desktop\Music-Source-Separation-Training-main\train.py", line 332, in train_model
model, config = get_model_from_config(args.model_type, args.config_path)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\Users\desktop\Music-Source-Separation-Training-main\utils.py", line 24, in get_model_from_config
from models.demucs4ht import get_model
File "C:\Users\Users\desktop\Music-Source-Separation-Training-main\models\demucs4ht.py", line 10, in
from demucs.demucs import Demucs
ModuleNotFoundError: No module named 'demucs'

[ZFTurbo]

pip install demucs

All required modules are here:
https://github.com/ZFTurbo/Music-Source-Separation-Training/blob/main/requirements.txt

[ghost]

Hi, I downloaded and redid it.
However, I get the same error with the following command.
Is there any solution?

python train.py --model_type htdemucs --config_path configs/config_musdb18_htdemucs.yaml --start_check_point results/f7e0c4bc-ba3fe64a.th --results_path results/ --data_path 'datasets/'

[ZFTurbo]

Your python doesn't see the demucs module for some reason.

[ghost]

pip install demucs
Now it works! Thank you.

[ghost]

Thanks for answering so many questions.
Thanks to you I understand how to use most of them.
I would like to make my own advanced models like the MDX23C.

For example, like separating the metronome sound from the drum sound.
For this, I need to create a learning model from scratch that does not use a pre-training model.
How can I create a new one without using the provided checkpoints?

--start_check_point results/model.ckpt \

[ZFTurbo]

Just don't use:

--start_check_point results/model.ckpt 

In this case training will start from scratch.

[ghost]

num_epochs: 1000
num_steps: 1000

We now have a new model that has been trained from the beginning with the first set of numbers.
However, the separation only generated noise that was not present in any of the data sets.
The noise is like a hum.
Does this mean that I have not learned enough yet?

[ZFTurbo]

What was the SDR on validation?

[ghost]

I was doing a lot of things to solve the problem. Thanks!
The problem was simply a lack of learning rate.
Lerning Rate is getting lower and lower, do you have a target number?
Would it be better to train until it is as close to 0 as possible?
Or should I target the SDR outputted by TEST, if there is something specific?

[ZFTurbo]

You can restart from higher LR (using latest weights) after it became too low.

[ghost]

I see, I understand.
Thanks.
So you don't have to worry so much about LR.

[Kenus23]

Python 3.13.0 (tags/v3.13.0:60403a5, Oct 7 2024, 09:38:07) [MSC v.1941 64 bit (AMD64)] on win32
Type "help", "copyright", "credits" or "license()" for more information.

= RESTART: C:\Users\user\Desktop\Music-Source-Separation-Training-1.0.9\train.py
Traceback (most recent call last):
File "C:\Users\user\Desktop\Music-Source-Separation-Training-1.0.9\train.py", line 9, in
from tqdm.auto import tqdm
ModuleNotFoundError: No module named 'tqdm'