ERROR: Failed building wheel for torchmcubes

[yefl2064]

     from /home/yefl/miniconda3/lib/python3.11/site-packages/torch/include/torch/csrc/api/include/torch/all.h:7,
                   from /home/yefl/miniconda3/lib/python3.11/site-packages/torch/include/torch/extension.h:5,
                   from cxx/grid_interp_cpu.cpp:1:
  /home/yefl/miniconda3/lib/python3.11/site-packages/torch/include/c10/util/string_view.h:10:23: fatal error: string_view: 没有那个文件或目录
   #include <string_view>
                         ^
  compilation terminated.
  error: command '/usr/bin/gcc' failed with exit code 1
  [end of output]

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

[pookiefoof]

Hi @yefl2064, what's your gcc version?

[Dont-look-at-me]

how can i check my gcc version?
i faced the same problem.

[DiamondGlassDrill]

Mine is gcc (Rev11, Built by MSYS2 project) 12.2.0 and i am now running in the same issue as I killed my TripoSR environment which worked before and now I cannot install trochmcubes

[DiamondGlassDrill]

6h later I still have the same issue it is really frustrating I have used 12.1 and 12.2 cuda version but nothing helps

[DiamondGlassDrill]

A fresh day, a fresh mind
I solved it this way:
First Part:

1. I now removed all CUDA 11.6,11.8,12.1,12.2 from the PC, was anyways time to do a clean up.
2. Updated to latest NVIDIA GPU driver
3. Installed CUDA 12.4.1 (not 12.5 too new for pytorch) deinstalled all torch related parts from pytorch CUDA_12.1 from my old TripoSR environment in my env.
4. Installed latest pytorch CUDA_12.4 parts as usual from (https://pytorch.org/get-started/locally)

Tried again:
pip install git+https://github.com/tatsy/torchmcubes.git

Second Part (new issues):
5. Ran into an issue as of above CUDA version 11.8, NVIDIA is no longer shipping NvToolsExt in lib format but in the new NVTX3 (header only) and the pytorch team couldn't fix it since 6months

Solution:
6. Downloaded 11.8 CUDA installer and unchecked all parts besides the NVTX part as explained at the bottom of (https://discuss.pytorch.org/t/failed-to-find-nvtoolsext/179635/8)
7. Copy in case of windows the file "nvToolsExt64_1.lib" afterwards from:
C:\Program Files\NVIDIA Corporation\NvToolsExt\lib\x64
to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4\lib\x64

Tried again:
pip install git+https://github.com/tatsy/torchmcubes.git

Worked!!!

That was easy ;)
I can confirm TripoSR runs on CUDA 12.4!

PS: Never update a running TripoSR env :) (without a backup)

[YuQiao0303]

I solved it on Linux following #48 (comment)

pip install git+https://github.com/tatsy/torchmcubes.git@3aef8afa5f21b113afc4f4ea148baee850cbd472

[AshMartian]

I solved it on Linux following #48 (comment)

pip install git+https://github.com/tatsy/torchmcubes.git@3aef8afa5f21b113afc4f4ea148baee850cbd472

This worked for me on Windows too

[IuryChagas25]

I solved it on Linux following #48 (comment)

pip install git+https://github.com/tatsy/torchmcubes.git@3aef8afa5f21b113afc4f4ea148baee850cbd472

Worked for me!