如何在 Windows 中安装 pytorch？

Question

我正在尝试在 Windows 上安装 pytorch，并且有一个可用但显示错误。

conda install -c peterjc123 pytorch=0.1.12

Answer 1

Warning: Unless you have a very specific reason not to, just follow the official installation instructions from https://pytorch.org . They are far more likely to be accurate and up-to-date.

---

Here is how to install the PyTorch package from the official channel, on Windows using Anaconda, as of the time of writing this comment (31/03/2020):

PyTorch without CUDA:

conda install pytorch torchvision cpuonly -c pytorch

PyTorch with CUDA 10.1:

conda install pytorch torchvision cudatoolkit=10.1 -c pytorch

Answer 2

go to the official website: http://pytorch.org/

1. Select Windows as your operating system
2. Select your Package Manager such as pip or conda
3. Select you python version
4. Select CUDA or choose none You will get the command that will install pytorch on your system based on your selection.

For example, if you choose Windows, pip, python 3.6 and none in the listed steps, you will get the following commands:

pip3 install http://download.pytorch.org/whl/cpu/torch-0.4.0-cp36-cp36m-win_amd64.whl 
pip3 install torchvision

Answer 3

Actual answer :

---

Best way is to check on the official website for up-to-date options. Here are the ones working as of 2020-03:

# for windows 10, CUDA 10.1
conda install pytorch torchvision cudatoolkit=10.1 -c pytorch

#for windows 10, CUDA 9
conda install pytorch torchvision cudatoolkit=9.2 -c pytorch -c defaults -c numba/label/dev 

Previous answer(out-of-date)

---

It seems that the author (peterjc123) released 2 days ago conda packages to install PyTorch 0.3.0 on windows. Here is a copy:

# 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

source: https://github.com/pytorch/pytorch/issues/494#issuecomment-350527200

Answer 4

If you are trying to install on windows 10 and you are not having the anaconda installation than the best options are below:

Python 2.7

pip install https://download.pytorch.org/whl/cpu/torch-1.0.1-cp27-cp27mu-linux_x86_64.whl
pip install torchvision

If the above command does not work, then you have python 2.7 UCS2, use this command

pip install https://download.pytorch.org/whl/cpu/torch-1.0.1-cp27-cp27m-linux_x86_64.whl

Python 3.5

pip3 install https://download.pytorch.org/whl/cpu/torch-1.0.1-cp35-cp35m-win_amd64.whl
pip3 install torchvision

Python 3.6

pip3 install https://download.pytorch.org/whl/cpu/torch-1.0.1-cp36-cp36m-win_amd64.whl
pip3 install torchvision

Python 3.7

pip3 install https://download.pytorch.org/whl/cpu/torch-1.0.1-cp37-cp37m-win_amd64.whl
pip3 install torchvision

Answer 5

Update June 2019: pytorch has a dedicated conda channel now and can be installed easily with anaconda. 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 6

The trick is to go to the PyTorch website and select the things you need:

Answer 7

You may want to consider using Docker for Windows . This would enable you to install pytorch as you would on Linux. Although, I believe that DfW has limited CUDA support, so you may want to explore a different option if you plan on using CUDA.

Answer 8

Try running:

conda install -c pytorch pytorch

The command will update/install: conda, cudatoolkit, pytorch.

Answer 9

pip install torch==1.5.0+cpu torchvision==0.6.0+cpu -f https://download.pytorch.org/whl/torch_stable.html

Answer 10

conda install -c peterjc123 pytorch   #for windows

This one simple line will get your work done for windows.

OR you may use this one for linux or osx.

conda install -c soumith pytorch 

Answer 11

I was getting some kind of Rollback error on Git bash and Windows Cmd prompt so had to run Anaconda prompt as admin for:

conda install pytorch-cpu -c pytorch

and then I got another when I tried the following command on Anaconda prompt:

pip3 install torchvision

so I switched back to Windows prompt to enter it and it worked.

To test the installation, I ran this from Git Bash:

$ python reinforcement_q_learning.py

with source code that looks like (the snippet near the top anyways):

"""

import gym
import math
import random
import numpy as np
import matplotlib
import matplotlib.pyplot as plt
from collections import namedtuple
from itertools import count
from PIL import Image

import torch
import torch.nn as nn
import torch.optim as optim
import torch.nn.functional as F
import torchvision.transforms as T


env = gym.make('CartPole-v0').unwrapped

# set up matplotlib
is_ipython = 'inline' in matplotlib.get_backend()
if is_ipython:
    from IPython import display

plt.ion()

# if gpu is to be used
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")


######################################################################
# Replay Memory
# -------------
#
# We'll be using experience replay memory for training our DQN. It stores
# the transitions that the agent observes, allowing us to reuse this data
# later. By sampling from it randomly, the transitions that build up a
# batch are decorrelated. It has been shown that this greatly stabilizes
# and improves the DQN training procedure.
#
# For this, we're going to need two classses:
#
# -  ``Transition`` - a named tuple representing a single transition in
#    our environment. It maps essentially maps (state, action) pairs
#    to their (next_state, reward) result, with the state being the
#    screen difference image as described later on.
# -  ``ReplayMemory`` - a cyclic buffer of bounded size that holds the
#    transitions observed recently. It also implements a ``.sample()``
#    method for selecting a random batch of transitions for training.
#

Transition = namedtuple('Transition',
                        ('state', 'action', 'next_state', 'reward'))


class ReplayMemory(object):

    def __init__(self, capacity):
        self.capacity = capacity
        self.memory = []
        self.position = 0

    def push(self, *args):
        """Saves a transition."""
        if len(self.memory) < self.capacity:
            self.memory.append(None)
        self.memory[self.position] = Transition(*args)
        self.position = (self.position + 1) % self.capacity

    def sample(self, batch_size):
        return random.sample(self.memory, batch_size)

    def __len__(self):
        return len(self.memory)

############continues to line 507...

Answer 12

A few steps as described here may help to install Pytorch in Windows:

1. First, we need to install Shapely. For this download Shapely as Shapely-1.6.3-cp36-cp36m-win_amd64.whl from here .

Then go to the directory where you have downloaded the whl file and then press SHIFT and right click and select open command prompt here and then execute this:

pip install Shapely-1.6.3-cp36-cp36m-win_amd64.whl

2. Secondly, we need to install Fiona. For this go to Gohlke and download Fiona‑1.7.13‑cp36‑cp36m‑win_amd64.whl

Then execute this command:

pip install Fiona‑1.7.13‑cp36‑cp36m‑win_amd64.whl

3. Third and final step is to download PyTorch . Again just as before execute this in command prompt:

pip install torch-0.4.0-cp36-cp36m-win_amd64.whl

Answer 13

On windows 10, Python 3.6 with pip install following works for me,

    pip3 install https://download.pytorch.org/whl/cu90/torch-1.1.0-cp36-cp36m-win_amd64.whl

also you need to install torchvision

pip3 install https://download.pytorch.org/whl/cu90/torchvision-0.3.0-cp36-cp36m-win_amd64.whl

Answer 14

Try this

1. cd C:\\Program files\\Anaconda3\\condabin
2. conda install pytorch-cpu torchvision-cpu -c pytorch

https://pytorch.org/get-started/locally/#anaconda

Answer 15

for python 3.7 which is the latest till date

for pytorch on cpu

pip install https://download.pytorch.org/whl/cpu/torch-1.0.1-cp37-cp37m-win_amd64.whl

pip install torchvision

Answer 16

I was using the official website( https://pytorch.org/get-started/locally/ ) where the following command is mentioned for Windows 10 & Conda environment:

conda install pytorch torchvision cudatoolkit=10.2 -c pytorch

I ran this command in Anaconda command prompt but I was getting stuck as there were following errors

ERROR conda.core.link:_execute(502): An error occurred while uninstalling package 'defaults::pycurl-7.43.0.1-py36h74b6da3_0'. WindowsError(5, 'Access is denied')

To rectify this I opened the Anaconda command prompt as administrator and then ran the same command again. It solved the access issue and allowed the package to get installed.

So you just have to use following two steps:

Step 1: Open Anaconda prompt as administrator

Step 2: run following command

conda install pytorch torchvision cudatoolkit=10.2 -c pytorch

Answer 17

Below is the working configuration:

Python 3.8

Pytorch version 1.5.1

Windows 10

pytorch installation command:

conda install pytorch torchvision cudatoolkit=9.2 -c pytorch -c defaults -c numba/label/dev

Answer 18

If @x0s answer gives dependency issues then try updating conda before that.

conda update conda  
conda install -c peterjc123 pytorch_legacy cuda80

Answer 19

This Line of code did the trick for me:

conda install -c peterjc123 pytorch

Check these links out in case you have any problem installing:

Superdatascience Tutorial Explains Clearly how to do it.

Or just go to the anaconda pytorch page: https://anaconda.org/peterjc123/pytorch

It worked for me.Hope my answer was useful.