ERROR: Failed building wheel for colossalai

[LSTM-Kirigaya]

When I attempt to install the colossalai for the first time, error happened during the process of building wheels.

pip install colossalai 

Many errors look like:

      D:/python/lib/site-packages/torch/include\c10/util/Optional.h(184): warning C4624: “c10::constexpr_storage_t<T>”: 已将析构函数隐式定义为“已删除”
              with
              [
                  T=at::Tensor
              ]
      D:/python/lib/site-packages/torch/include\c10/util/Optional.h(383): note: 查看对正在编译的 类 模板 实例化“c10::constexpr_storage_t<T>”的引用
              with
              [
                  T=at::Tensor
              ]
      C:/Program Files (x86)/Microsoft Visual Studio/2019/Community/VC/Tools/MSVC/14.28.29333/include\type_traits(630): note: 查看对正在编译的 类 模板 实例化“c10::trivially_copyable_optimization_optional_base<T>”的引用
              with
              [
                  T=at::Tensor
              ]
      D:/python/lib/site-packages/torch/include\c10/util/Optional.h(526): note: 查看对正在编译的 类 模板 实例化“std::is_copy_constructible<c10::trivially_copyable_optimization_optional_base<T>>”的引用
              with
              [
                  T=at::Tensor
              ]
      D:/python/lib/site-packages/torch/include\c10/util/Optional.h(526): note: 查看对正在编译的 别名 模板 实例化“c10::OptionalBase<at::Tensor>”的引用

In the end of errors:

...
...
      1 error detected in the compilation of "colossalai/kernel/cuda_native/csrc/multi_tensor_l2norm_kernel.cu".
      error: command 'C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v11.5\\bin\\nvcc.exe' failed with exit code 4294967295
      Error in atexit._run_exitfuncs:
      Traceback (most recent call last):
        File "D:\python\lib\site-packages\colorama\ansitowin32.py", line 59, in closed
          return stream.closed
      ValueError: underlying buffer has been detached
      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
error: legacy-install-failure

× Encountered error while trying to install package.
╰─> colossalai

note: This is an issue with the package mentioned above, not pip.
hint: See above for output from the failure.

I can use cuda in torch for training.

Environment:

• NVCC: V11.5.50
• torch: 1.10.1+cu113
• python: 3.7.6

[FrankLeeeee]

Hi, thanks for your issue. We have not tested Colossal-AI on windows, but we can investigate this issue.

[LSTM-Kirigaya]

Thanks. I have installed colossalai successfully in WSL2 in my Windows 😄. Awesome!