Unexpected key(s) in state_dict: "istft.ola_window". about music-separation-gui

[Freddd13]

In fact, if you have a torch environment, you don't need the full package lol but just the code as I mentioned in README.md. This package is large because it contains a independent python environment and cuda.
Anyway, glad to help you run successfully. Good luck.

[Freddd13]

Hi, lucas. Are you using GUI only or the full package？ I think it is caused by an improper torch version. If you install torch by yourself, you should install 1.7.1 version.

[Freddd13]

Besides, I found that you are not using the latest version. Please refer to the release to get 003 version. It should work fine.

[User0404]

16:36:25 Traceback (most recent call last):
16:36:25 File "D:_easyHelper_\Python\Lib\site-packages\qgui\base_tools.py", line 169, in new_func
func(obj)
16:36:25 File "D:_easyHelper_\resources\MainUI.py", line 232, in run_single
progress = self
16:36:25 File "D:_easyHelper_\resources\SepProcessor.py", line 149, in separate_file
audio = load_audio(audio_path=audio_path, mono=False, sample_rate=sample_rate) # audio: (input_channels, audio_samples)
16:36:25 File "D:_easyHelper_\Python\Lib\site-packages\bytesep\utils.py", line 62, in load_audio
audio_path, sr=sample_rate, mono=mono, offset=offset, duration=duration
16:36:25 File "D:_easyHelper_\Python\Lib\site-packages\librosa\core\audio.py", line 163, in load
y, sr_native = _audioread_load(path, offset, duration, dtype)
16:36:25 File "D:_easyHelper\Python\Lib\site-packages\librosa\core\audio.py", line 187, in audioread_load
with audioread.audio_open(path) as input_file:
16:36:25 File "D:_easyHelper\Python\Lib\site-packages\audioread_init.py", line 116, in audio_open
raise NoBackendError()
16:36:25 audioread.exceptions.NoBackendError
16:36:25 None

请问这是什么问题呢？
cpu R5
显卡 rtx 1650

[Freddd13]

你好 @User0404 。你这个是因为没有安装ffmpeg, 十分抱歉，因为我的失误，以为上次修改后已经不需要ffmpeg依赖，实际上有个包还是需要的。你可以尝试将解压后的 `resources/tools`目录加入[环境变量](https://jingyan.baidu.com/article/00a07f3876cd0582d128dc55.html)，因为暂时没装虚拟机没法测试不保证这个能成功，不过这可能是最简单的方法。 如果还是不行，请参考[这里](https://blog.csdn.net/qq_43803367/article/details/110308401)安装 ffmpeg并将其加入环境变量 Sent with a Spark 在 2022年4月10日 +0800 16:40，User0404 ***@***.***>，写道：

…

16:36:25 Traceback (most recent call last): 16:36:25 File "D:_easyHelper_\Python\Lib\site-packages\qgui\base_tools.py", line 169, in new_func func(obj) 16:36:25 File "D:_easyHelper_\resources\MainUI.py", line 232, in run_single progress = self 16:36:25 File "D:_easyHelper_\resources\SepProcessor.py", line 149, in separate_file audio = load_audio(audio_path=audio_path, mono=False, sample_rate=sample_rate) # audio: (input_channels, audio_samples) 16:36:25 File "D:_easyHelper_\Python\Lib\site-packages\bytesep\utils.py", line 62, in load_audio audio_path, sr=sample_rate, mono=mono, offset=offset, duration=duration 16:36:25 File "D:_easyHelper_\Python\Lib\site-packages\librosa\core\audio.py", line 163, in load y, sr_native = _audioread_load(path, offset, duration, dtype) 16:36:25 File "D:_easyHelper\Python\Lib\site-packages\librosa\core\audio.py", line 187, in audioread_load with audioread.audio_open(path) as input_file: 16:36:25 File "D:_easyHelper\Python\Lib\site-packages\audioread_init.py", line 116, in audio_open raise NoBackendError() 16:36:25 audioread.exceptions.NoBackendError 16:36:25 None 请问这是什么问题呢？ cpu R5 显卡 rtx 1650 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***>

[playdasegunda]

Good afternoon, sorry for the late reply,

Unfortunately I'm still facing the same problem reported,

I downloaded the package with the latest version of this GUI but I still get the error "istft.ola_window"

I have no idea what it could be,

I can only say that probably the problem is because I have installed some pip dependency, but I don't know what it is, do you have an idea what it might be?

[Freddd13]

@lucasbr15 How do you launch the program? Please just open run.exe to launch instead of using python MainUI.py.
Remeber if you run it for the first time, you need to run install.cmd first.

To install, run install.cmd.

Wait the terminal to download and install packages. If you encounter a win32 dll import error, try to run the fix_win32_error.bat.

This may take some time, please be patient. If it is stuck for too long, please try to press "Enter" on your keyboard!!! (This problem may only occurs during the first opening.)

After the first installation, you can quickly start the program by running run.exe in the future. In fact, the install.cmd will provoke the run.exe. The former is to deal with torch packages, so there's no need to run it after the installation.

[playdasegunda]

Hello;

When running run.exe I get the following error

[Freddd13]

The directory you use should not contain a space.

[playdasegunda]

Good afternoon, now it is working 100% I moved the Programs Files folder to Downloads, and installed torch torch 1.11.0+cu113 and it seems to work, probably the error I was getting was because I was using torch 1.9.0, I am running it from the exe file as suggested and it is also working normally now, thanks a lot for the help!