Tacotron2環境構築でApexがインストールできない：ModuleNotFoundError: No module named 'torch'

実現したいこと

Tacotron2を用いて自分の声で文章を読ませるプログラムを作成したいと思っています。そのために、Tacotron2を導入しようと考えています。環境は、AnacondaのJupyterLabを使っています

発生している問題・分からないこと

Tacotron2を導入するために以下のコマンドを実行します。

Python
1! git clone https://github.com/NVIDIA/tacotron2
2! git submodule init; git submodule update
3! pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu117
4! git clone https://github.com/nvidia/apex
5
6import os
7os.chdir('apex')

ここまでは上手く動作します。そして次のコマンドで問題は発生します。

Python
1$ pip install -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./

エラー内容は以下の通りとなります。

ModuleNotFoundError: No module named 'torch'

torchが見つからないそうです。
しかし、

!pip install torch

を実行すると、

Requirement already satisfied:

と普通にtorchはインストールされています。

しかし、何故か上記のような「torchが見つからない」というエラーが出てしまいます。

エラーメッセージ

error
1```エラー内容
2Using pip 22.3.1 from C:\Users\takespring\anaconda3\lib\site-packages\pip (python 3.10)
3Processing c:\users\takespring\source\anaconda_python_study\tacotron2_study\apex
4  Installing build dependencies: started
5  Installing build dependencies: finished with status 'done'
6  Getting requirements to build wheel: started
7  Getting requirements to build wheel: finished with status 'error'
8WARNING: Implying --no-binary=:all: due to the presence of --build-option / --global-option / --install-option. Consider using --config-settings for more flexibility.
9DEPRECATION: --no-binary currently disables reading from the cache of locally built wheels. In the future --no-binary will not influence the wheel cache. pip 23.1 will enforce this behaviour change. A possible replacement is to use the --no-cache-dir option. You can use the flag --use-feature=no-binary-enable-wheel-cache to test the upcoming behaviour. Discussion can be found at https://github.com/pypa/pip/issues/11453
10  Running command pip subprocess to install build dependencies
11  DEPRECATION: --no-binary currently disables reading from the cache of locally built wheels. In the future --no-binary will not influence the wheel cache. pip 23.1 will enforce this behaviour change. A possible replacement is to use the --no-cache-dir option. You can use the flag --use-feature=no-binary-enable-wheel-cache to test the upcoming behaviour. Discussion can be found at https://github.com/pypa/pip/issues/11453
12  Collecting setuptools
13    Using cached setuptools-75.1.0.tar.gz (1.3 MB)
14    Getting requirements to build wheel: started
15    Getting requirements to build wheel: finished with status 'done'
16    Preparing metadata (pyproject.toml): started
17    Preparing metadata (pyproject.toml): finished with status 'done'
18  Collecting wheel
19    Using cached wheel-0.44.0.tar.gz (100 kB)
20    Installing build dependencies: started
21    Installing build dependencies: finished with status 'done'
22    Getting requirements to build wheel: started
23    Getting requirements to build wheel: finished with status 'done'
24    Preparing metadata (pyproject.toml): started
25    Preparing metadata (pyproject.toml): finished with status 'done'
26  Building wheels for collected packages: wheel, setuptools
27    Building wheel for wheel (pyproject.toml): started
28    Building wheel for wheel (pyproject.toml): finished with status 'done'
29    Created wheel for wheel: filename=wheel-0.44.0-py3-none-any.whl size=67059 sha256=79b3427c1f582be3c4102c06d28c196d1b9815870f25d2f431c864fd160b2072
30    Stored in directory: c:\users\takespring\appdata\local\pip\cache\wheels\39\7d\c2\27f9739db1d0af92c01bc1219b8ee6601d4229aa88b3a4ec45
31    Building wheel for setuptools (pyproject.toml): started
32    Building wheel for setuptools (pyproject.toml): finished with status 'done'
33    Created wheel for setuptools: filename=setuptools-75.1.0-py3-none-any.whl size=1248534 sha256=a0b46c01c145e0713677ca5495b22515f133a86c208a0dc23757cced1f7c9b43
34    Stored in directory: c:\users\takespring\appdata\local\pip\cache\wheels\d0\c1\3b\a0bc89d43586b8e15a4c99d36be536ba6705da7668c04772c0
35  Successfully built wheel setuptools
36  Installing collected packages: wheel, setuptools
37  Successfully installed setuptools-75.1.0 wheel-0.44.0
38  Running command Getting requirements to build wheel
39  Traceback (most recent call last):
40    File "C:\Users\takespring\anaconda3\lib\site-packages\pip\_vendor\pep517\in_process\_in_process.py", line 351, in <module>
41      main()
42    File "C:\Users\takespring\anaconda3\lib\site-packages\pip\_vendor\pep517\in_process\_in_process.py", line 333, in main
43      json_out['return_val'] = hook(**hook_input['kwargs'])
44    File "C:\Users\takespring\anaconda3\lib\site-packages\pip\_vendor\pep517\in_process\_in_process.py", line 118, in get_requires_for_build_wheel
45      return hook(config_settings)
46    File "C:\Users\takespring\AppData\Local\Temp\pip-build-env-v4gn59cg\overlay\Lib\site-packages\setuptools\build_meta.py", line 332, in get_requires_for_build_wheel
47      return self._get_build_requires(config_settings, requirements=[])
48    File "C:\Users\takespring\AppData\Local\Temp\pip-build-env-v4gn59cg\overlay\Lib\site-packages\setuptools\build_meta.py", line 302, in _get_build_requires
49      self.run_setup()
50    File "C:\Users\takespring\AppData\Local\Temp\pip-build-env-v4gn59cg\overlay\Lib\site-packages\setuptools\build_meta.py", line 318, in run_setup
51      exec(code, locals())
52    File "<string>", line 10, in <module>
53  ModuleNotFoundError: No module named 'torch'
54  error: subprocess-exited-with-error
55  
56  Getting requirements to build wheel did not run successfully.
57  exit code: 1
58  
59  See above for output.
60  
61  note: This error originates from a subprocess, and is likely not a problem with pip.
62  full command: 'C:\Users\takespring\anaconda3\python.exe' 'C:\Users\takespring\anaconda3\lib\site-packages\pip\_vendor\pep517\in_process\_in_process.py' get_requires_for_build_wheel 'C:\Users\TAKESP~1\AppData\Local\Temp\tmpuwbupcak'
63  cwd: C:\Users\takespring\source\anaconda_python_study\tacotron2_study\apex
64error: subprocess-exited-with-error
65
66Getting requirements to build wheel did not run successfully.
67exit code: 1
68
69See above for output.
70
71note: This error originates from a subprocess, and is likely not a problem with pip.


### 該当のソースコード

特になし


### 試したこと・調べたこと
- [x] teratailやGoogle等で検索した
- [x] ソースコードを自分なりに変更した
- [ ] 知人に聞いた
- [ ] その他

##### 上記の詳細・結果
[https://github.com/NVIDIA/apex/issues/1823
](url)
このサイトで同じ問題を見つけましたが、解決はしなかったです。

### 補足
特になし