THC/THC.h: No such file or directory

[cchan-lm]

It appears since PyTorch 1.11, TH/THC include files have been removed. warp_rnnt wasn't the only one affected. Other repos relying on these include files were as well (see https://github.com/open-mmlab/mmdetection3d/issues/1332).

Python version: 3.9.11 OS: Ubuntu 18.04.6 LTS

pip install torch==1.11.0+cu113 --extra-index-url https://download.pytorch.org/whl/cu113
pip install warp_rnnt

Output:

Looking in indexes: https://pypi.org/simple, https://pypi.ngc.nvidia.com
Collecting warp_rnnt
  Downloading warp_rnnt-0.5.0.tar.gz (10 kB)
  Preparing metadata (setup.py) ... done
Collecting pybind11
  Downloading pybind11-2.10.0-py3-none-any.whl (213 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 213.3/213.3 kB 32.3 MB/s eta 0:00:00
Collecting numpy
  Downloading numpy-1.23.3-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (17.1 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 17.1/17.1 MB 10.9 MB/s eta 0:00:00
Requirement already satisfied: torch>=1.0.0 in ./.pyenv/versions/3.9.11/envs/rnnt/lib/python3.9/site-packages (from warp                                  _rnnt) (1.12.1+cu113)
Requirement already satisfied: typing-extensions in ./.pyenv/versions/3.9.11/envs/rnnt/lib/python3.9/site-packages (from                                   torch>=1.0.0->warp_rnnt) (4.3.0)
Using legacy 'setup.py install' for warp_rnnt, since package 'wheel' is not installed.
Installing collected packages: pybind11, numpy, warp_rnnt
  Running setup.py install for warp_rnnt ... error
  error: subprocess-exited-with-error

  × Running setup.py install for warp_rnnt did not run successfully.
  │ exit code: 1
  ╰─> [23 lines of output]
      running install
      running build
      running build_py
      creating build
      creating build/lib.linux-x86_64-3.9
      creating build/lib.linux-x86_64-3.9/warp_rnnt
      copying warp_rnnt/test.py -> build/lib.linux-x86_64-3.9/warp_rnnt
      copying warp_rnnt/__init__.py -> build/lib.linux-x86_64-3.9/warp_rnnt
      running build_ext
      /home/linuxbrew/.linuxbrew/Cellar/pyenv/2.3.4/libexec/pyenv-exec: /home/craig/bin/ninja: /usr/local/bin/python: ba                                  d interpreter: No such file or directory
      /home/linuxbrew/.linuxbrew/Cellar/pyenv/2.3.4/libexec/pyenv-exec: line 48: /home/craig/bin/ninja: Success
      /home/craig/.pyenv/versions/3.9.11/envs/rnnt/lib/python3.9/site-packages/torch/utils/cpp_extension.py:411: UserWar                                  ning: Attempted to use ninja as the BuildExtension backend but we could not find ninja.. Falling back to using the slow                                   distutils backend.
        warnings.warn(msg.format('we could not find ninja.'))
      /home/craig/.pyenv/versions/3.9.11/envs/rnnt/lib/python3.9/site-packages/torch/utils/cpp_extension.py:813: UserWar                                  ning: The detected CUDA version (11.1) has a minor version mismatch with the version that was used to compile PyTorch (1                                  1.3). Most likely this shouldn't be a problem.
        warnings.warn(CUDA_MISMATCH_WARN.format(cuda_str_version, torch.version.cuda))
      building 'warp_rnnt._C' extension
      creating build/temp.linux-x86_64-3.9
      gcc -pthread -Wno-unused-result -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -fPIC -I/home/craig/.pyenv/versions/3                                  .9.11/envs/rnnt/lib/python3.9/site-packages/torch/include -I/home/craig/.pyenv/versions/3.9.11/envs/rnnt/lib/python3.9/s                                  ite-packages/torch/include/torch/csrc/api/include -I/home/craig/.pyenv/versions/3.9.11/envs/rnnt/lib/python3.9/site-pack                                  ages/torch/include/TH -I/home/craig/.pyenv/versions/3.9.11/envs/rnnt/lib/python3.9/site-packages/torch/include/THC -I/us                                  r/local/cuda/include -I/home/craig/.pyenv/versions/3.9.11/envs/rnnt/include -I/home/craig/.pyenv/versions/3.9.11/include                                  /python3.9 -c binding.cpp -o build/temp.linux-x86_64-3.9/binding.o -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_T                                  YPE="_gcc" -DPYBIND11_STDLIB="_libstdcpp" -DPYBIND11_BUILD_ABI="_cxxabi1011" -DTORCH_EXTENSION_NAME=_C -D_GLIBCXX_USE_CX                                  X11_ABI=0 -std=c++14
      binding.cpp:4:10: fatal error: THC/THC.h: No such file or directory
       #include <THC/THC.h>
                ^~~~~~~~~~~
      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: legacy-install-failure

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

note: This is an issue with the package mentioned above, not pip.

The above error also occurs for PyTorch 1.12, which we're required to use for our software. Will warp_rnnt be updated, or do you recommend another way forward?

[pooya-mohammadi]

The same error here. Do you have any plans to update the repository?

[1ytic]

Hey, thank you for your feedbacks. Could you try a new version pip install warp-rnnt==0.6.0? Should be solved now.

[pooya-mohammadi]

@1ytic Thanks for the update. I get the following error:

139 | #error -- unsupported GNU version! gcc versions later than 10 are not supported! The nvcc flag '-allow-unsupported-compiler' can be used to override this version check; however, using an unsupported host compiler may cause compilation failure or incorrect run time execution. Use at your own risk.

[1ytic]

Please share your environment details: OS, Python version, PyTorch version, CUDA/cuDNN version, GCC version. Could you try another version of gcc?

[pooya-mohammadi]

@1ytic

OS: Ubuntu 22.04
python version: 3.9.12
cuda version: NVIDIA-SMI 510.85.02    Driver Version: 510.85.02    CUDA Version: 11.6
cudatoolkit: 11.3.1 -> installed by conda
cudnn: cudnn-linux-x86_64-8.3.3.40_cuda11.5
cudnn: py3.9_cuda11.3_cudnn8.2.0_0    installed by conda along with pytorch
gcc: gcc (Ubuntu 11.3.0-1ubuntu1~22.04) 11.3.0
pytorch: 1.11.0 -> installed by conda

Unfortunately, I cannot install another version of gcc

[cchan-lm]

Installation works on my end! Thank you for fixing this :) Leaving as open pending resolution for @pooya-mohammadi, but feel free to close afterwards!

[pooya-mohammadi]

@cchan-lm would you mind reporting your environments details?