How to build theseus with Python 3.7 and Pytorch 1.12.1

[TruongKhang]

❓ Questions and Help

Hello, I'm having trouble when trying to install theseus with Python 3.7 and Pytorch 1.12.1.
I created a conda virtual environment with Python 3.7.

First, I tried to install from source by pip install -e .. But it yielded some errors and the installed Pytorch version was 1.13 by default.
Then, I tried to build by using ./build_scripts/build_wheel.sh, but it produced a Python 3.10 by default.

Do you have any suggestions?

Thanks for your help!

[luisenp]

Hi @TruongKhang. When you ran pip install -e ., did you install the desired Pytorch version ahead of time? Our recommended steps are to first install torch and suitesparse separately, then run the Theseus installer.

[TruongKhang]

Thanks for your response @luisenp
I installed Pytorch 1.12.1 in a conda virtual environment.
Here is the result when I ran pip install -e ., I have no idea why the code uninstalled Pytorch 1.12.1 and then installed torch 1.13.1 instead.

  Preparing metadata (setup.py) ... done
Requirement already satisfied: numpy>=1.19.2 in /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages (from theseus-ai==0.1.3) (1.21.5)
Requirement already satisfied: scipy>=1.5.3 in /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages (from theseus-ai==0.1.3) (1.7.3)
Collecting scikit-sparse>=0.4.5
  Using cached scikit-sparse-0.4.8.tar.gz (209 kB)
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
  Preparing metadata (pyproject.toml) ... done
Requirement already satisfied: pytest>=6.2.1 in /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages (from theseus-ai==0.1.3) (7.1.2)
Collecting pybind11>=2.7.1
  Using cached pybind11-2.10.3-py3-none-any.whl (222 kB)
Collecting functorch
  Using cached functorch-1.13.1-py2.py3-none-any.whl (2.1 kB)
Requirement already satisfied: attrs>=19.2.0 in /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages (from pytest>=6.2.1->theseus-ai==0.1.3) (21.4.0)
Requirement already satisfied: iniconfig in /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages (from pytest>=6.2.1->theseus-ai==0.1.3) (1.1.1)
Requirement already satisfied: packaging in /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages (from pytest>=6.2.1->theseus-ai==0.1.3) (21.3)
Requirement already satisfied: pluggy<2.0,>=0.12 in /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages (from pytest>=6.2.1->theseus-ai==0.1.3) (1.0.0)
Requirement already satisfied: py>=1.8.2 in /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages (from pytest>=6.2.1->theseus-ai==0.1.3) (1.11.0)
Requirement already satisfied: tomli>=1.0.0 in /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages (from pytest>=6.2.1->theseus-ai==0.1.3) (2.0.1)
Requirement already satisfied: importlib-metadata>=0.12 in /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages (from pytest>=6.2.1->theseus-ai==0.1.3) (3.10.0)
Collecting torch<1.13.2,>=1.13.1
  Using cached torch-1.13.1-cp37-cp37m-manylinux1_x86_64.whl (887.5 MB)
Requirement already satisfied: zipp>=0.5 in /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages (from importlib-metadata>=0.12->pytest>=6.2.1->theseus-ai==0.1.3) (3.8.0)
Requirement already satisfied: typing-extensions>=3.6.4 in /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages (from importlib-metadata>=0.12->pytest>=6.2.1->theseus-ai==0.1.3) (4.3.0)
Collecting nvidia-cublas-cu11==11.10.3.66
  Using cached nvidia_cublas_cu11-11.10.3.66-py3-none-manylinux1_x86_64.whl (317.1 MB)
Collecting nvidia-cuda-runtime-cu11==11.7.99
  Using cached nvidia_cuda_runtime_cu11-11.7.99-py3-none-manylinux1_x86_64.whl (849 kB)
Collecting nvidia-cuda-nvrtc-cu11==11.7.99
  Using cached nvidia_cuda_nvrtc_cu11-11.7.99-2-py3-none-manylinux1_x86_64.whl (21.0 MB)
Collecting nvidia-cudnn-cu11==8.5.0.96
  Using cached nvidia_cudnn_cu11-8.5.0.96-2-py3-none-manylinux1_x86_64.whl (557.1 MB)
Requirement already satisfied: setuptools in /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages (from nvidia-cublas-cu11==11.10.3.66->torch<1.13.2,>=1.13.1->functorch->theseus-ai==0.1.3) (63.4.1)
Requirement already satisfied: wheel in /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages (from nvidia-cublas-cu11==11.10.3.66->torch<1.13.2,>=1.13.1->functorch->theseus-ai==0.1.3) (0.37.1)
Requirement already satisfied: pyparsing!=3.0.5,>=2.0.2 in /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages (from packaging->pytest>=6.2.1->theseus-ai==0.1.3) (3.0.9)
Building wheels for collected packages: scikit-sparse
  Building wheel for scikit-sparse (pyproject.toml) ... done
  Created wheel for scikit-sparse: filename=scikit_sparse-0.4.8-cp37-cp37m-linux_x86_64.whl size=573884 sha256=6f8af4f73cce203df9eee9260301602b1707cd8b30d94a9e002f1f33c29bdedd
  Stored in directory: /home/khangtg/.cache/pip/wheels/e7/ae/fc/aa2b7d5325fd50f6aceb23b5c82e6f37b72653af336ddce5ea
Successfully built scikit-sparse
Installing collected packages: pybind11, nvidia-cuda-runtime-cu11, nvidia-cuda-nvrtc-cu11, nvidia-cublas-cu11, scikit-sparse, nvidia-cudnn-cu11, torch, functorch, theseus-ai
  Attempting uninstall: torch
    Found existing installation: torch 1.12.1
    Uninstalling torch-1.12.1:
      Successfully uninstalled torch-1.12.1
  Running setup.py develop for theseus-ai
    error: subprocess-exited-with-error
    
    × python setup.py develop did not run successfully.
    │ exit code: 1
    ╰─> [23 lines of output]
        Traceback (most recent call last):
          File "/home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages/torch/__init__.py", line 172, in _load_global_deps
            ctypes.CDLL(lib_path, mode=ctypes.RTLD_GLOBAL)
          File "/home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/ctypes/__init__.py", line 364, in __init__
            self._handle = _dlopen(self._name, mode)
        OSError: /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages/torch/lib/../../nvidia/cublas/lib/libcublas.so.11: symbol cublasLtGetStatusString version libcublasLt.so.11 not defined in file libcublasLt.so.11 with link time reference
        
        During handling of the above exception, another exception occurred:
        
        Traceback (most recent call last):
          File "<string>", line 36, in <module>
          File "<pip-setuptools-caller>", line 34, in <module>
          File "/home/khangtg/Documents/lab/CVPR23/theseus/setup.py", line 13, in <module>
            import torch
          File "/home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages/torch/__init__.py", line 217, in <module>
            _load_global_deps()
          File "/home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages/torch/__init__.py", line 178, in _load_global_deps
            _preload_cuda_deps()
          File "/home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages/torch/__init__.py", line 158, in _preload_cuda_deps
            ctypes.CDLL(cublas_path)
          File "/home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/ctypes/__init__.py", line 364, in __init__
            self._handle = _dlopen(self._name, mode)
        OSError: /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages/nvidia/cublas/lib/libcublas.so.11: symbol cublasLtGetStatusString version libcublasLt.so.11 not defined in file libcublasLt.so.11 with link time reference
        [end of output]
    
    note: This error originates from a subprocess, and is likely not a problem with pip.
error: subprocess-exited-with-error

× python setup.py develop did not run successfully.
│ exit code: 1
╰─> [23 lines of output]
    Traceback (most recent call last):
      File "/home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages/torch/__init__.py", line 172, in _load_global_deps
        ctypes.CDLL(lib_path, mode=ctypes.RTLD_GLOBAL)
      File "/home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/ctypes/__init__.py", line 364, in __init__
        self._handle = _dlopen(self._name, mode)
    OSError: /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages/torch/lib/../../nvidia/cublas/lib/libcublas.so.11: symbol cublasLtGetStatusString version libcublasLt.so.11 not defined in file libcublasLt.so.11 with link time reference
    
    During handling of the above exception, another exception occurred:
    
    Traceback (most recent call last):
      File "<string>", line 36, in <module>
      File "<pip-setuptools-caller>", line 34, in <module>
      File "/home/khangtg/Documents/lab/CVPR23/theseus/setup.py", line 13, in <module>
        import torch
      File "/home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages/torch/__init__.py", line 217, in <module>
        _load_global_deps()
      File "/home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages/torch/__init__.py", line 178, in _load_global_deps
        _preload_cuda_deps()
      File "/home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages/torch/__init__.py", line 158, in _preload_cuda_deps
        ctypes.CDLL(cublas_path)
      File "/home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/ctypes/__init__.py", line 364, in __init__
        self._handle = _dlopen(self._name, mode)
    OSError: /home/khangtg/anaconda3/envs/torch12cu113/lib/python3.7/site-packages/nvidia/cublas/lib/libcublas.so.11: symbol cublasLtGetStatusString version libcublasLt.so.11 not defined in file libcublasLt.so.11 with link time reference
    [end of output]

note: This error originates from a subprocess, and is likely not a problem with pip.```

[TruongKhang]

I somehow can manage to install Theseus successfully with Python 3.8 and Pytorch 1.12.1.
I built a wheel file by ./build_scripts/build_wheel.sh (I changed Python version in this file from 3.10 to 3.8). After that:

• install Pytorch 1.12.1 by using pip
• install functorch 0.2.1
• install the built wheel file above

[luisenp]

Hi @TruongKhang, good to hear you got it working. It's interesting that in your error trace above functorch was getting installed as version 1.13, even though our default requirements specify 0.2.1. Not sure why that was the case.

[mhmukadam]

Could be something to do with functorch now upstreamed into PyTorch proper.

[TruongKhang]

@luisenp , @mhmukadam, Let's take a look at line 38 of setup.py: reqs.append(line.split("==")[0])
I think maybe the functorch version 0.2.1 (functorch==0.2.1) was ignored due to this line.

[luisenp]

Oh, that's a great catch, thanks for mentioning it! I'll fix this soon.

[luisenp]

Fixed in #448. Thanks again!