installation issue

[Tsingularity]

Hi, thanks for the great library!

I met the following issue when trying to install functorch:

1. in the doc, it says functorch is already included in pytorch 1.13, but after i install pytorch-nightly, it still tells me no module name functorch.
2. so i install with pip, but after installation, it will give me the following userwarning when running my code. looks like torchvision is broken? i saw another issue had similar error message but on torchaudio.

   anaconda3/envs/nv/lib/python3.9/site-packages/torchvision/io/image.py:13: 
   UserWarning: Failed to load image Python extension: /local/lumingt/anaconda3/envs/nv/lib/python3.9/site-packages/torchvision/image.so:
   undefined symbol: _ZN2at4_ops19empty_memory_format4callEN3c108ArrayRefIlEENS2_8optionalINS2_10ScalarTypeEEENS5_INS2_6LayoutEEENS5_INS2_6DeviceEEENS5_IbEENS5_INS2_12MemoryFormatEEE
   warn(f"Failed to load image Python extension: {e}")

   could you please take a look at this and see what should i do now? thanks!

[zou3519]

What version of torch do you have installed? If you could run the following script and show the output then that would be helpful to diagnose

wget https://raw.githubusercontent.com/pytorch/pytorch/master/torch/utils/collect_env.py
# For security purposes, please check the contents of collect_env.py before running it.
python collect_env.py

[Tsingularity]

thanks for help!

here're the outputs:

Collecting environment information...                                                                                                                                               [3/1048]PyTorch version: 1.13.0+cu117
Is debug build: False
CUDA used to build PyTorch: 11.7
ROCM used to build PyTorch: N/A

OS: Ubuntu 18.04.6 LTS (x86_64)
GCC version: (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0
Clang version: Could not collect
CMake version: version 3.24.3
Libc version: glibc-2.27

Python version: 3.9.13 | packaged by conda-forge | (main, May 27 2022, 16:56:21)  [GCC 10.3.0] (64-bit runtime)
Python platform: Linux-4.15.0-189-generic-x86_64-with-glibc2.27
Is CUDA available: True
CUDA runtime version: 11.3.109
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration: GPU 0: NVIDIA RTX A6000
Nvidia driver version: 515.76
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

Versions of relevant libraries:
[pip3] functorch==1.13.0
[pip3] mypy-extensions==0.4.3
[pip3] numpy==1.23.4
[pip3] pytorch-ranger==0.1.1
[pip3] torch==1.13.0
[pip3] torch-optimizer==0.3.0
[pip3] torchvision==0.13.1
[conda] blas                      2.116                       mkl    conda-forge
[conda] blas-devel                3.9.0            16_linux64_mkl    conda-forge
[conda] cudatoolkit               11.6.0              hecad31d_10    conda-forge
[conda] cudatoolkit-dev           11.3.1           py39h3811e60_0    conda-forge
[conda] functorch                 1.13.0                   pypi_0    pypi
[conda] libblas                   3.9.0            16_linux64_mkl    conda-forge
[conda] libcblas                  3.9.0            16_linux64_mkl    conda-forge
[conda] liblapack                 3.9.0            16_linux64_mkl    conda-forge
[conda] liblapacke                3.9.0            16_linux64_mkl    conda-forge
[conda] mkl                       2022.1.0           h84fe81f_915    conda-forge
[conda] mkl-devel                 2022.1.0           ha770c72_916    conda-forge
[conda] mkl-include               2022.1.0           h84fe81f_915    conda-forge
[conda] numpy                     1.23.4           py39h3d75532_1    conda-forge
[conda] pytorch-mutex             1.0                        cuda    pytorch
[conda] pytorch-ranger            0.1.1                    pypi_0    pypi
[conda] torch                     1.13.0                   pypi_0    pypi
[conda] torch-optimizer           0.3.0                    pypi_0    pypi
[conda] torchvision               0.13.1               py39_cu116    pytorch

[Tsingularity]

follow-up above:

after installing functorch with pip, looks like some of my custom cuda kernel also got broken, with a similar undefined symbol error message.

[zou3519]

It looks like you have torch 1.13.0 and functorch 1.13.0. The version of torchvision that is compatible with those is 0.14.0, but it looks like you have 0.13.0 installed. I don't know what custom cuda kernels you're referring to, but if you are having undefined symbol errors, then that means that they don't work with torch 1.13.0.