Pytorch install with anaconda error

Question

I get this error:

C:\Users>conda install pytorch torchvision -c soumith
Fetching package metadata .............

PackageNotFoundError: Package missing in current win-64 channels:
  - pytorch

I got conda install pytorch torchvision -c soumith from Pytorch official website and I have OSX/conda/3.6/none for settings on Pytorch site(should be correct). I am new to conda, any tips how to solve this?

Answer 1

Update: From PyTorch 0.4, there is an official Anaconda channel with packages for Windows as well. You can install the latest pre-built (binary) version of PyTorch (GPU version by default) on windows using:

conda install -c pytorch pytorch

For CPU only version: conda install -c pytorch pytorch-cpu

For specific version of CUDA say CUDA9.1: conda install -c pytorch pytorch cuda91

Official instructions for windows are now available here

Old answer for previous versions (<0.4) of PyTorch on Windows:

It looks like you are on windows (win 64) and you are trying to install pytorch by choosing OSX because you don't have an option listed for win64 on pytorch site . (Correct me if that is not the case, I will revise my answer). -c soumith will use soumith's channel which only has packages for Linux and OSX not for windows. That is why you got that error when you ran conda from a windows machine.

You can install pytorch on windows through conda using this command:

conda install -c peterjc123 pytorch

This will fetch the pytorch package using peterjc123's channel which has packages for Win64.

You can install the torchvision package using pip like this: pip install torchvision

Answer 2

Use the following commands to install pytorch on windows

for Windows 10 and Windows Server 2016, CUDA 8

conda install -c peterjc123 pytorch cuda80

for Windows 10 and Windows Server 2016, CUDA 9

conda install -c peterjc123 pytorch cuda90

for Windows 7/8/8.1 and Windows Server 2008/2012, CUDA 8

conda install -c peterjc123 pytorch_legacy cuda80

Answer 3

On June,2019, The command generated at pytorch will require dependencies before it can be executed successfully. For example I chose stable pytorch 1.1 build with python 3.6 and Cuda 10.0. The command generated by pytorch page was as follows:

conda install pytorch torchvision cudatoolkit=10.0 -c pytorch

But it will not work if you have created a new conda environment like me. The step by step process for setting up pytorch is as follows:

• First install the cudatoolkit as follows:

conda install -c anaconda cudatoolkit=10.0

• Then install the mkl_fft as follows:

conda install -c anaconda mkl_fft

• Assuming you will face no more dependency issues. Use the following command to setup pytorch:

conda install -c pytorch pytorch

This worked for me. But I had setup my new conda environment with scikit-learn and jupyter notebook before starting the pytorch setup. So if any dependency problem arise, it would be a good idea to install both scikit-learn and jupyter notebook as well.

Answer 4

I have face similar problem because I have installed pytorch cpu only version. I tried everything and update pytorch to gpu version but nothing helped. Simple solution is to create new environment and then install pytorch gpu version. It resolve my problem