window11, is it possible to install tiny-cuda-nn?

[kotran88]

I did everything including format all my environment but it is not working
window11 with gpu nvidia 3090
installed visual studio 2022 with C build tool checked
cuda version is 11.8
nvcc --version

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2022 NVIDIA Corporation
Built on Wed_Sep_21_10:41:10_Pacific_Daylight_Time_2022
Cuda compilation tools, release 11.8, V11.8.89
Build cuda_11.8.r11.8/compiler.31833905_0

>>> print(torch.__version__)
2.1.2+cu118

(nerfstudio) C:\Users\jung\tiny-cuda-nn>cmake --version
cmake version 3.29.3

CMake suite maintained and supported by Kitware (kitware.com/cmake).

(nerfstudio) C:\Users\jung\tiny-cuda-nn>ninja --version
1.9.0

every version is meet as required in tinycudann specification...

but it cause error in installing it.

first,
pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch

and it cause error...

 ninja: build stopped: subcommand failed.
      Traceback (most recent call last):
        File "C:\Users\jung\AppData\Local\anaconda3\envs\nerfstudio\lib\site-packages\torch\utils\cpp_extension.py", line 2100, in _run_ninja_build
          subprocess.run(
        File "C:\Users\jung\AppData\Local\anaconda3\envs\nerfstudio\lib\subprocess.py", line 516, in run
          raise CalledProcessError(retcode, process.args,
      subprocess.CalledProcessError: Command '['ninja', '-v']' returned non-zero exit status 1.

      The above exception was the direct cause of the following exception:

      Traceback (most recent call last):
        File "<string>", line 2, in <module>
        File "<pip-setuptools-caller>", line 34, in <module>

   File "C:\Users\jung\AppData\Local\anaconda3\envs\nerfstudio\lib\site-packages\torch\utils\cpp_extension.py", line 2116, in _run_ninja_build
          raise RuntimeError(message) from e
      RuntimeError: Error compiling objects for extension
      [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
  ERROR: Failed building wheel for tinycudann
  Running setup.py clean for tinycudann
Failed to build tinycudann
ERROR: Could not build wheels for tinycudann, which is required to install pyproject.toml-based projects

also, I tried
https://github.com/NVlabs/tiny-cuda-nn?tab=readme-ov-file#compilation-windows--linux

on Developer PowerShell for vs2022,

[Edward-HE]

I've been lately delving into the installation of tiny-cuda-nn, particularly on the Windows platform (which is acknowledged to have a much more arduous installation environment than the Linux platform). I discovered the misunderstandings existing in the previous issues and summarized a viable solution. You can directly try out the tips as follows:

• Install PyTorch and the corresponding version of full CUDA (apparently, it seems you've done it correctly).
• Install the MSVC 2022 compilation chain (just ensure to install the following components):
  • MSVC Build tools (latest)
  • Windows 11 SDK
  • C++ CMake tools for Windows
• [Key Point] Compile and Install with x64 Native Tools Command Prompt for VS 2022 !!! (Use the compile install instead of the pip command)

git clone --recursive https://github.com/nvlabs/tiny-cuda-nn
cd tiny-cuda-nn/bindings/torch
# Switch to the Python environment you wish to install.
python setup.py install

• Wait for approximately 3 minutes.
• Finish!

By using the tips above, you can bypass almost all potential problems and fully leverage the super-fast compilation of ninja.

[tim-win]

I'd like to meet someone who pip install git+https://github.com/NVlabs/tiny-cuda-nn/#subdirectory=bindings/torch works for, I've run into issues every time I've tried to perform the install that way.

Manually cloning the repo and building the bindings/torch manually is required for linux far as I can tell, giving up on the standard install recommendation was the first step toward getting this to run correctly.

Is NVLabs open to contracting a build engineer? Install issues have been plaguing this (otherwise great) repo since, based on google searches, 2022 if not earlier. Hate to see a bunch of cool tech utilize this library only for it to be a massive build-chain barrier for entry.