Install PyTorch on Windows with Visual Studio

Quick walk-through on installing PyTorch (circa 1.3) on Visual Studio (2017) and Windows 10.

Get set up for Deep Learning without having to struggle through finding all the right tools and settings -- about 20 minutes including download/install time. A good first start for machine learning engineers who want to build intelligent systems.

[maxlau]

Hi Geoff, kindly ping! I'm a software engineer trying to hands on more Machine Learning stuff and preparing for the potential ML related tasks. Found your tutorial very very helpful! Appreciate it so much for the help! Hope could learn more interesting knowledge from you in the future~

[iamlettuce]

I came across this randomly, thank you it was very useful.

[gtrent]

Not even close. Can't even do step one on VS 2022 CE. No Such option as Anaconda. Only have 6 option. Github Copilot, Python 3 (Out of support), Python native development tools, Life share and Azure Cloud Services core tools. That it.

[kareemshaik]

Thanks for the video. Much much useful. Here Pytorch is prebuild package that is being installed using pip. But what I want is to build pytorch from source code and then use this custom build pytorch package to create virtual env in visual studio project how to do this?

[brahxter]

Best video Ive seen so far on how to do this shit. lol. Great Job man!

[naomijames]

When I used the Anaconda command prompt, I typed (base) PS C:\Users\naomi> activate PytorchVS as I named my file PytorchVS.

I expected the next line to start with Pytorch, however it created a second line as this (base) PS C:\Users\naomi>

I then copy and pasted the Pytorch code after this and pressed enter, it now looks like this

(base) PS C:\Users\naomi> activate PytorchVS
(base) PS C:\Users\naomi> conda install pytorch torchvision cpuonly -c pytorch
Collecting package metadata (current_repodata.json): done
Solving environment: done

## Package Plan ##

etc.

I am worried that even though this is done, I did not create my environment successfully because the second line was supposed to start with PytorchVS but it still starts with base. How can I undo this?

[LLlikeme]

Hi! Very helpful but if I want to install the CUDA toolkit... I can with the option of the pytorch page? or I need to press the CUDA option on pytorch and also download NVIDIA CUDA from the page?

[gabrielebasaglia]

Hi Geoff, I tried your tutorial.
Everything worked fine before I wrote the "import torch" instruction. What am I doing wrong?
>>> from __future__ import print_function
>>> import torch
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "C:\Users\Gabriele\AppData\Local\conda\conda\envs\PytorchVS\lib\site-packages\torch\__init__.py", line 81, in <module>
ctypes.CDLL(dll)
File "C:\Users\Gabriele\AppData\Local\conda\conda\envs\PytorchVS\lib\ctypes\__init__.py", line 348, in __init__
self._handle = _dlopen(self._name, mode)
OSError: [WinError 126] Impossibile finding the specified module
>>>
Thanks,
GB