Installation and getting started with pytoch

Pytorch Installation

The goal is

1. Know how to install pytorch

1. Pytorch Introduction to

Pytorch Is a facebook Published in-depth learning framework , Because of its ease of use , Friendliness , Favored by the majority of users .

2. Pytorch Version of

3. Pytorch Installation

Installation address introduction ：https://pytorch.org/get-started/locally/

belt GPU Installation steps ：

conda install pytorch torchvision cudatoolkit=9.0 -c pytorch

No GPU Installation steps

conda install pytorch-cpu torchvision-cpu -c pytorch

Open after installation ipython

Input ：

In [1]:import torch
In [2]: torch.__version__
Out[2]: '1.0.1'

Be careful ： When installing the module, you install pytorch , But in the code, we use torch

There was an article before the detailed introduction , You can go there by yourself if necessary

The first step of deep learning （anaconda、pytorch install ）

Pytorch To get started with

The goal is

1. Know tensor sum Pytorch The tensor in
2. know pytorch How to create tensors in
3. know pytorch in tensor Common ways to do it
4. know pytorch in tensor Data type of
5. know pytorch How to realize tensor stay cpu and cuda Medium transformation

1. tensor Tensor

Tensor is a general term , There are many types ：

1. 0 Order tensor ： Scalar 、 constant ,0-D Tensor
2. 1 Order tensor ： vector ,1-D Tensor
3. 2 Order tensor ： matrix ,2-D Tensor
4. 3 Order tensor
5. …
6. N Order tensor

2. Pytorch Create tensors in

1. Use python Create a list or sequence in tensor

   torch.tensor([[1., -1.], [1., -1.]])
   tensor([[ 1.0000, -1.0000],
           [ 1.0000, -1.0000]])
   

2. Use numpy Array creation in tensor

   torch.tensor(np.array([[1, 2, 3], [4, 5, 6]]))
   tensor([[ 1,  2,  3],
           [ 4,  5,  6]])
   

3. Use torch Of api establish tensor

   1. torch.empty(3,4) establish 3 That's ok 4 The column is empty tensor, Will be filled with useless data

   2. torch.ones([3,4]) establish 3 That's ok 4 Column All for 1 Of tensor

   3. torch.zeros([3,4]) establish 3 That's ok 4 Column All for 0 Of tensor

   4. torch.rand([3,4]) establish 3 That's ok 4 Column Random value Of tensor, The interval of random values is [0, 1)

      >>> torch.rand(2, 3)
      tensor([[ 0.8237,  0.5781,  0.6879],
      [ 0.3816,  0.7249,  0.0998]])
      

   5. torch.randint(low=0,high=10,size=[3,4]) establish 3 That's ok 4 Column Random integers Of tensor, The interval of random values is [low, high)

      >>> torch.randint(3, 10, (2, 2))
      tensor([[4, 5],
      	[6, 7]])
      

   6. torch.randn([3,4]) establish 3 That's ok 4 Column random number Of tensor, The distributed mean of random values is 0, The variance of 1

3. Pytorch in tensor The common method of

1. obtain tensor Data in ( When tensor Only one element is available in )：tensor.item()

   In [10]: a = torch.tensor(np.arange(1))
   
   In [11]: a
   Out[11]: tensor([0])
   
   In [12]: a.item()
   Out[12]: 0
   

2. Turn into numpy Array

   In [55]: z.numpy()
   Out[55]:
   array([[-2.5871205],
          [ 7.3690367],
          [-2.4918075]], dtype=float32)
   

3. Get shape ：tensor.size()

   In [72]: x
   Out[72]:
   tensor([[    1,     2],
           [    3,     4],
           [    5,    10]], dtype=torch.int32)
   
   In [73]: x.size()
   Out[73]: torch.Size([3, 2])
   

4. Shape change ：tensor.view((3,4)). similar numpy Medium reshape, It's a shallow copy , Just a change in shape

   In [76]: x.view(2,3)
   Out[76]:
   tensor([[    1,     2,     3],
           [    4,     5,    10]], dtype=torch.int32)
   

5. Get order ：tensor.dim()

   In [77]: x.dim()
   Out[77]: 2
   

6. Get the maximum ：tensor.max()

   In [78]: x.max()
   Out[78]: tensor(10, dtype=torch.int32)
   

7. Transposition ：tensor.t()

   In [79]: x.t()
   Out[79]:
   tensor([[    1,     3,     5],
           [    2,     4, 	  10]], dtype=torch.int32)
   

8. tensor[1,3] obtain tensor The value of the first row and the third column in

9. tensor[1,3]=100 Yes tensor The position of the first row and the third column in the is assigned 100

10. tensor The section of

In [101]: x
Out[101]:
tensor([[1.6437, 1.9439, 1.5393],
     [1.3491, 1.9575, 1.0552],
     [1.5106, 1.0123, 1.0961],
     [1.4382, 1.5939, 1.5012],
     [1.5267, 1.4858, 1.4007]])

In [102]: x[:,1]
Out[102]: tensor([1.9439, 1.9575, 1.0123, 1.5939, 1.4858])

​

4. tensor Data type of

tensor There are many data types in , The common types are as follows ：

In the picture above Tensor types This kind of type Of tensor Is an example

1. obtain tensor Data type of :tensor.dtype

   In [80]: x.dtype
   Out[80]: torch.int32
   

2. Specify the type when creating data

   In [88]: torch.ones([2,3],dtype=torch.float32)
   Out[88]:
   tensor([[9.1167e+18, 0.0000e+00, 7.8796e+15],
           [8.3097e-43, 0.0000e+00, -0.0000e+00]])
   

3. Type modification

   In [17]: a
   Out[17]: tensor([1, 2], dtype=torch.int32)
   
   In [18]: a.type(torch.float)
   Out[18]: tensor([1., 2.])
   
   In [19]: a.double()
   Out[19]: tensor([1., 2.], dtype=torch.float64)
   

5. tensor Other operations of

1. tensor and tensor Add up

   In [94]: x = x.new_ones(5, 3, dtype=torch.float)
   
   In [95]: y = torch.rand(5, 3)
   
   In [96]: x+y
   Out[96]:
   tensor([[1.6437, 1.9439, 1.5393],
           [1.3491, 1.9575, 1.0552],
           [1.5106, 1.0123, 1.0961],
           [1.4382, 1.5939, 1.5012],
           [1.5267, 1.4858, 1.4007]])
   In [98]: torch.add(x,y)
   Out[98]:
   tensor([[1.6437, 1.9439, 1.5393],
           [1.3491, 1.9575, 1.0552],
           [1.5106, 1.0123, 1.0961],
           [1.4382, 1.5939, 1.5012],
           [1.5267, 1.4858, 1.4007]])
   In [99]: x.add(y)
   Out[99]:
   tensor([[1.6437, 1.9439, 1.5393],
           [1.3491, 1.9575, 1.0552],
           [1.5106, 1.0123, 1.0961],
           [1.4382, 1.5939, 1.5012],
           [1.5267, 1.4858, 1.4007]])
   In [100]: x.add_(y)  # The underlined method will be correct for x Make local modifications 
   Out[100]:
   tensor([[1.6437, 1.9439, 1.5393],
           [1.3491, 1.9575, 1.0552],
           [1.5106, 1.0123, 1.0961],
           [1.4382, 1.5939, 1.5012],
           [1.5267, 1.4858, 1.4007]])
   
   In [101]: x #x Change 
   Out[101]:
   tensor([[1.6437, 1.9439, 1.5393],
           [1.3491, 1.9575, 1.0552],
           [1.5106, 1.0123, 1.0961],
           [1.4382, 1.5939, 1.5012],
           [1.5267, 1.4858, 1.4007]])
   

   Be careful ： Underlined methods （ such as :add_) Would be right tensor Make local modifications

2. tensor And digital operation

   In [97]: x +10
   Out[97]:
   tensor([[11., 11., 11.],
           [11., 11., 11.],
           [11., 11., 11.],
           [11., 11., 11.],
           [11., 11., 11.]])
   

3. CUDA Medium tensor

   CUDA（Compute Unified Device Architecture）, yes NVIDIA The new computing platform . CUDA It's a kind of NVIDIA General parallel computing architecture , The architecture enables GPU Able to solve complex computing problems .

   torch.cuda This module adds a pair of CUDA tensor Support for , In the cpu and gpu Use the same method on tensor

   adopt .to Method can put a tensor Transfer to another device ( For instance from CPU go to GPU)

   #device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
   if torch.cuda.is_available():
       device = torch.device("cuda")          # cuda device object 
       y = torch.ones_like(x, device=device)  #  Create one in cuda Upper tensor
       x = x.to(device)                       #  How to use x To cuda  Of tensor
       z = x + y
       print(z)
       print(z.to("cpu", torch.double))       # .to Method can also set the type 
       
   >>tensor([1.9806], device='cuda:0')
   >>tensor([1.9806], dtype=torch.float64)
   

Through the previous study , You can find torch The various operations of are almost the same as numpy equally

Video learning website ：pytorch Getting started video