incompatible package versions about meta-tts HOT 9 CLOSED

能請您提供安裝時的error message嗎？我最近更新套件版本時並沒有明確出現問題，過去經驗中較容易出現的安裝問題主要是pytorch/torchaudio/torchtext彼此版本的不匹配，這部分可以參考我在另一個repo的回覆。若要直接使用pretrained checkpoint，目前確認可直接執行的程式版本是這個commit。
Hope these helps，也祝我們彼此都畢業順利！

from meta-tts.

非常感谢您的及时解答。在您的建议下我更换了commit，但在执行到Processing中的python3 preprocess.py config/LibriTTS/preprocess.yaml时依然出现了包不兼容的错误：
File "pyworld/pyworld.pyx", line 1, in init pyworld.pyworld
ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 96 from C header, got 80 from PyObject
这个错误看起来是numpy与puyworld不兼容导致的，在我将numpy=1.19.2升级到1.20时依然存在这个问题，于是我将pyworld=0.3.2改到0.3.0时解决了这个问题。但是与此同时又产生了新的问题：
OSError: /home/yuexin/anaconda3/envs/metatts-1/lib/python3.8/site-packages/nvidia/cublas/lib/libcublas.so.11: symbol cublasLtGetStatusString version libcublasLt.so.11 not defined in file libcublasLt.so.11 with link time reference
这个错误看起来是torch版本跟cublas不兼容导致的，于是我将torch=1.13.1改到1.9.1版本，从而解决了上述问题。
但在更改完torch版本后出现了新的红色警告：
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
torchvision 0.14.1 requires torch==1.13.1, but you have torch 1.9.1 which is incompatible.
torchaudio 0.13.1 requires torch==1.13.1, but you have torch 1.9.1 which is incompatible.
pytorch-lightning 1.9.3 requires torch>=1.10.0, but you have torch 1.9.1 which is incompatible.
我不确定这是否会导致新的问题，所以可能的话希望您能提供成功运行时的各种安装包的版本。
再次感谢！

from meta-tts.

我觉得问题可能出在这个地方：在运行pip install -r requirement.txt 时，系统给我安装的各个包版本跟您执行时的版本并不相同，所以会导致接下来的错误

from meta-tts.

但在更改完torch版本后出现了新的红色警告：
ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
torchvision 0.14.1 requires torch==1.13.1, but you have torch 1.9.1 which is incompatible.
torchaudio 0.13.1 requires torch==1.13.1, but you have torch 1.9.1 which is incompatible.
pytorch-lightning 1.9.3 requires torch>=1.10.0, but you have torch 1.9.1 which is incompatible.

這個就是我說的pytorch/torchaudio/torchtext彼此版本的不匹配問題，可以參考我在另一個repo的回覆

至於我自己的package版本，因為我是偶爾想到才升級，基本上也沒有出錯，所以就沒有特別限定。以下是我目前環境的版本：
learn2learn 0.1.7
comet-ml 3.22.0
g2p-en 2.1.0
inflect 5.3.0
librosa 0.9.2
matplotlib 3.6.3
matplotlib-inline 0.1.3
numba 0.56.4
numpy 1.23.5
pypinyin 0.44.0
pyworld 0.3.0
PyYAML 6.0
scikit-learn 1.2.1
scipy 1.9.3
soundfile 0.11.0
tensorboard 2.11.2
tensorboard-data-server 0.6.1
tensorboard-plugin-wit 1.8.0
tgt 1.4.4
hypergrad 0.1 /home/r06942045/myProjects/Meta-TTS/hypertorch
pytorch-lightning 1.6.4
torch 1.13.1
torchaudio 0.9.1
torchmeta 1.8.0
torchmetrics 0.9.0
torchvision 0.14.1
tqdm 4.64.1
torchaudio 0.9.1
seaborn 0.11.2

from meta-tts.

感谢您的回复。但我发现您的包也有冲突的地方，比如torch 1.13.1与torchaudio 0.9.1是冲突的，请问是之前用的torch1.9.1然后再升级到1.13.1的吗，现在您是不是只是test而没有再train模型了

from meta-tts.

Using cache found in /home/yuexin/.cache/torch/hub/descriptinc_melgan-neurips_master
Traceback (most recent call last):
File "main.py", line 196, in
main(args, configs)
File "main.py", line 105, in main
model = system(
File "/home/yuexin/Downloads/Meta-TTS-main/lightning/systems/baseline.py", line 20, in init
super().init(*args, **kwargs)
File "/home/yuexin/Downloads/Meta-TTS-main/lightning/systems/base_adaptor.py", line 26, in init
super().init(*args, **kwargs)
File "/home/yuexin/Downloads/Meta-TTS-main/lightning/systems/system.py", line 47, in init
self.vocoder = LightningMelGAN()
File "/home/yuexin/Downloads/Meta-TTS-main/lightning/utils.py", line 11, in init
vocoder = torch.hub.load(
File "/home/yuexin/anaconda3/envs/metatts/lib/python3.8/site-packages/torch/hub.py", line 382, in load
model = _load_local(repo_or_dir, model, *args, **kwargs)
File "/home/yuexin/anaconda3/envs/metatts/lib/python3.8/site-packages/torch/hub.py", line 411, in _load_local
model = entry(*args, **kwargs)
File "/home/yuexin/.cache/torch/hub/descriptinc_melgan-neurips_master/hubconf.py", line 16, in load_melgan
return MelVocoder(path=None, github=True, model_name=model_name)
File "/home/yuexin/.cache/torch/hub/descriptinc_melgan-neurips_master/mel2wav/interface.py", line 38, in init
self.fft = Audio2Mel().to(device)
File "/home/yuexin/.cache/torch/hub/descriptinc_melgan-neurips_master/mel2wav/modules.py", line 42, in init
mel_basis = librosa_mel_fn(
TypeError: mel() takes 0 positional arguments but 5 were given

此外在运行Meta-TTS-main分支的main来train时候报了这个错误，看起来是MelVocoder构造函数中使用了librosa_mel_fn，该函数传入5个位置参数，但是没有被接收， 有没有可能是因为库的版本问题导致函数的参数发生了变化

from meta-tts.

感谢您的回复。但我发现您的包也有冲突的地方，比如torch 1.13.1与torchaudio 0.9.1是冲突的，请问是之前用的torch1.9.1然后再升级到1.13.1的吗，现在您是不是只是test而没有再train模型了

我還是有在train誒，雖然不是meta的部分，但像是baseline還是train的起來啊？不過我確實是從1.9.1升上去的，您可以再試試

from meta-tts.

"/home/yuexin/.cache/torch/hub/descriptinc_melgan-neurips_master/mel2wav/modules.py", line 42, in init mel_basis = librosa_mel_fn( TypeError: mel() takes 0 positional arguments but 5 were given

此外在运行Meta-TTS-main分支的main来train时候报了这个错误，看起来是MelVocoder构造函数中使用了librosa_mel_fn，该函数传入5个位置参数，但是没有被接收， 有没有可能是因为库的版本问题导致函数的参数发生了变化

這感覺是您的librosa有問題誒？就算是最新的librosa.filters.mel也是有用兩個positional arguments的（sr, n_fft），而且後面的keywork arguments理論上也要可以照順序assign，您的error message比較像是您的系統以為mel()完全沒有argument，這反倒很奇怪？

from meta-tts.

感谢您的及时回复。如您所言，确实是librosa版本有问题，在将版本改为 0.9.2后解决了这个问题。

from meta-tts.