How to install pytorch from source for Asus GeForce 710 GT with CUDA CC 3.5 and supported CUDA Toolkit 11.0

l123456 · August 16, 2020, 4:19pm

How to install pytorch from source for Asus GeForce 710 GT with CUDA CC 3.5 and supported CUDA Toolkit 11.0?

l123456 · August 16, 2020, 4:23pm

The solution is that I have to build my own version from source. From Pytorch 1.3.1 on, 3.5 is not supported in the binaries anymore, they do this to reduce the size of the binary. See the comment at gpu - Which PyTorch version is CUDA 3.0 compatible? - Stack Overflow and Issues · pytorch/pytorch · GitHub. That means I can well use pytorch 1.3.1 and higher with CUDA cc 3.5, I simply have to compile it myself.

And this link shows how to install pytorch from source:

github.com

pytorch/pytorch/blob/master/README.md

![PyTorch Logo](https://github.com/pytorch/pytorch/blob/master/docs/source/_static/img/pytorch-logo-dark.png)

--------------------------------------------------------------------------------

PyTorch is a Python package that provides two high-level features:
- Tensor computation (like NumPy) with strong GPU acceleration
- Deep neural networks built on a tape-based autograd system

You can reuse your favorite Python packages such as NumPy, SciPy, and Cython to extend PyTorch when needed.

Our trunk health (Continuous Integration signals) can be found at [hud.pytorch.org](https://hud.pytorch.org/ci/pytorch/pytorch/master).

<!-- toc -->

- [More About PyTorch](#more-about-pytorch)
  - [A GPU-Ready Tensor Library](#a-gpu-ready-tensor-library)
  - [Dynamic Neural Networks: Tape-Based Autograd](#dynamic-neural-networks-tape-based-autograd)
  - [Python First](#python-first)
  - [Imperative Experiences](#imperative-experiences)
  - [Fast and Lean](#fast-and-lean)

This file has been truncated. show original

I need for example MSVC 2019 to compile for CUDA CC 3.5

l123456 · August 16, 2020, 4:32pm

Main help to install from source came from:

l123456 · August 16, 2020, 4:36pm

Here are the detailed steps:

install cmake from the internet
install git from the internet
MS Visual Studio 2019 16.7.1
→ Indivudual components:

most recent MSVC v142 - VS 2019 C+±x64/x86-Buildtools (v14.27) (the most recent x64 version at that time)
most recent Windows 10 SDK (10.0.19041.0) (the most recent x64 version at that time)
(myenv) C:\Users\Admin\Downloads\Pytorch>git clone GitHub - pytorch/pytorch: Tensors and Dynamic neural networks in Python with strong GPU acceleration
(myenv) C:\Users\Admin\Downloads\Pytorch\pytorch>git submodule update --init --recursive
(myenv) C:\Users\Admin\Downloads\Pytorch>git clone GitHub - peterjc123/pytorch-scripts: A few Windows specific scripts for PyTorch
(Copy all files in C:\Users\Admin\Downloads\Pytorch\pytorch-scripts that are not: “\internal”, “.github”, README.md, .gitignore to C:\Users\Admin\Downloads\Pytorch\pytorch)
conda install numpy ninja pyyaml mkl mkl-include setuptools cmake cffi typing_extensions future six requests
set CMAKE_GENERATOR=Visual Studio 16 2019
# using --cmake: I can't build pytorch 1.3.1 from sources. · Issue #30910 · pytorch/pytorch · GitHub, unclear if needed
(myenv) C:\Users\Admin\Downloads\Pytorch\pytorch>python setup.py install --cmake
CMake Error: Error: generator : Ninja Does not match the generator used previously: Visual Studio 16 2019 Either remove the CMakeCache.txt file and CMakeFiles directory or choose a different binary directory.

l123456 · August 16, 2020, 4:39pm

if this has not worked: set USE_NINJA=OFF
(myenv) C:\Users\Admin\Downloads\Pytorch\pytorch>python setup.py install --cmake
if this has not worked:
(myenv) C:\Users\Admin\Downloads\Pytorch\pytorch>set DMSVC_Z7_OVERRIDE=OFF
(myenv) C:\Users\Admin\Downloads\Pytorch\pytorch>python setup.py install --cmake
if this has not worked (according to output):
(myenv) C:\Users\Admin\Downloads\Pytorch\pytorch>set “CUDAHOSTCXX=C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.27.29110\bin\Hostx64\x64\cl.exe”
(myenv) C:\Users\Admin\Downloads\Pytorch\pytorch>python setup.py install --cmake