pytorch install 시작해보기(설치기)

What is PyTorch?

It’s a Python based scientific computing package targeted at two sets of audiences:

• A replacement for NumPy to use the power of GPUs
• a deep learning research platform that provides maximum flexibility and speed

PyTorch는 파이썬기반의 과학 계산용 패키지 입니다.
- GPU들의 계산을 사용하기 위해서 NumPy 교체 합니다.
- 최대 유연성과 속도를 제공하는 Deep Learning 연구 플랫폼입니다.

0. 설치 환경

Windows 10
CUDA 지원 안함

1. 아나콘다 설치 

요즘 아나콘다가 대세네요
https://www.anaconda.com/download/
3.6 버전 설치
64bit 용 설치 바랍니다. PyTorch용이 64bit만 존재합니다.(2019.1.13일 기준)

2. PyTorch 설치

https://pytorch.org/ 접속
운영체제와 python 버전 선택

anaconda shell에서 실행
conda install pytorch-cpu -c pytorch
pip3 install torchvision
=> pip3는 동작안합니다.
pip install torchvision 을 실행하도록 합니다.

(base) E:\>pip install torchvision
Collecting torchvision
  Using cached https://files.pythonhosted.org/packages/ca/0d/f00b2885711e08bd71242ebe7b96561e6f6d01fdb4b9dcf4d37e2e13c5e1/torchvision-0.2.1-py2.py3-none-any.whl
Requirement already satisfied: pillow>=4.1.1 in e:\programdata\anaconda3\lib\site-packages (from torchvision) (5.1.0)
Requirement already satisfied: numpy in e:\programdata\anaconda3\lib\site-packages (from torchvision) (1.14.3)
Requirement already satisfied: torch in e:\programdata\anaconda3\lib\site-packages (from torchvision) (0.4.0)
Requirement already satisfied: six in e:\programdata\anaconda3\lib\site-packages (from torchvision) (1.11.0)
distributed 1.21.8 requires msgpack, which is not installed.
Installing collected packages: torchvision
Successfully installed torchvision-0.2.1

빨간색으로 msgpack이 없다고 에러가 난다면 아래와 같이 msgpack를 설치해줍니다.

(base) E:\>pip install msgpack
Collecting msgpack
  Downloading https://files.pythonhosted.org/packages/04/81/c6363198f24ec1c56e5c48ce685cb532e175125adade0cdb181c8c5fea6e/msgpack-0.5.6-cp36-cp36m-win_amd64.whl (85kB)
    100% |████████████████████████████████| 92kB 303kB/s
Installing collected packages: msgpack
Successfully installed msgpack-0.5.6

다시 설치해보면 에러 문구는 나타나지 않습니다.

(base) E:\>pip install torchvision
Requirement already satisfied: torchvision in e:\programdata\anaconda3\lib\site-packages (0.2.1)
Requirement already satisfied: pillow>=4.1.1 in e:\programdata\anaconda3\lib\site-packages (from torchvision) (5.1.0)
Requirement already satisfied: numpy in e:\programdata\anaconda3\lib\site-packages (from torchvision) (1.14.3)
Requirement already satisfied: torch in e:\programdata\anaconda3\lib\site-packages (from torchvision) (0.4.0)
Requirement already satisfied: six in e:\programdata\anaconda3\lib\site-packages (from torchvision) (1.11.0)

3. 예제 실행해보기

아직 뭐가 뭔지는 모르겠지만 제대로 설치되었는지 샘플을 실행해보면 됩니다.
아래에서 제일아래 tensor_tutorial.py 예제를 다운로드 받을 수 있습니다.
https://pytorch.org/tutorials/beginner/blitz/tensor_tutorial.html#getting-started
실행은 아래와 같이 합니다.
(base) E:\pytorch>python tensor_tutorial.py

결과

tensor(1.00000e-43 *
       [[ 0.0000,  0.0000,  0.0000],
        [ 0.0000,  0.0000,  0.0000],
        [ 0.0000,  0.0000,  0.0000],
        [ 0.0000,  1.6816,  0.0000],
        [ 0.0000,  0.0000,  0.0000]])
tensor([[ 0.9836,  0.3660,  0.1008],
        [ 0.9361,  0.0545,  0.4045],
        [ 0.7703,  0.6550,  0.0720],
        [ 0.5599,  0.6790,  0.6733],
        [ 0.8420,  0.5594,  0.4168]])
tensor([[ 0,  0,  0],
        [ 0,  0,  0],
        [ 0,  0,  0],
        [ 0,  0,  0],
        [ 0,  0,  0]])
tensor([ 5.5000,  3.0000])
tensor([[ 1.,  1.,  1.],
        [ 1.,  1.,  1.],
        [ 1.,  1.,  1.],
        [ 1.,  1.,  1.],
        [ 1.,  1.,  1.]], dtype=torch.float64)
tensor([[ 0.1702, -0.5912,  0.8690],
        [-0.0291,  0.5152,  1.1440],
        [-0.2709, -0.3549,  0.0495],
        [ 1.6595,  0.5233, -0.2273],
        [ 0.9720,  0.5231,  0.8539]])
torch.Size([5, 3])
tensor([[ 1.0499,  0.2793,  1.6822],
        [ 0.4391,  0.9968,  1.4158],
        [ 0.4017,  0.1225,  0.2590],
        [ 2.1562,  1.3099, -0.0821],
        [ 1.7752,  1.0990,  0.9617]])
tensor([[ 1.0499,  0.2793,  1.6822],
        [ 0.4391,  0.9968,  1.4158],
        [ 0.4017,  0.1225,  0.2590],
        [ 2.1562,  1.3099, -0.0821],
        [ 1.7752,  1.0990,  0.9617]])
tensor([[ 1.0499,  0.2793,  1.6822],
        [ 0.4391,  0.9968,  1.4158],
        [ 0.4017,  0.1225,  0.2590],
        [ 2.1562,  1.3099, -0.0821],
        [ 1.7752,  1.0990,  0.9617]])
tensor([[ 1.0499,  0.2793,  1.6822],
        [ 0.4391,  0.9968,  1.4158],
        [ 0.4017,  0.1225,  0.2590],
        [ 2.1562,  1.3099, -0.0821],
        [ 1.7752,  1.0990,  0.9617]])
tensor([-0.5912,  0.5152, -0.3549,  0.5233,  0.5231])
torch.Size([4, 4]) torch.Size([16]) torch.Size([2, 8])
tensor([ 0.3910])
0.39095234870910645

결과가 뭘 뜻하는지는 getting-started 페이지를 천천히 읽어봐야 할것 같습니다만, 에러가 없다는것에 만족하며 여기에서 끝내야 할 것 같습니다.