Pytorch

Pytorch is an open source machine learning library based on the torch library used for applications such as computer vision and natural language processing.

Pytorch tensors

Pytorch defines a class called tensor(torch.tensor) to store and operations on homogeneous multidimensional rectangular array of number.

Tensors

Pytorch tensor is the fundamental unit of the pytorch framework whose operation are similar to python numpy array. Convert python list to pytorch tensor.

 import torch
 import numpy as np
 t =torch.tensor([[15., -33.,44.], [-81., -54., 55]])
 t
                

tensor([[ 15., -33.,  44.],
[-81., -54.,  55.]])
                

t1=torch.tensor(np.array([[45, 27, 63], [144, 549, 72]]))
t1
                

tensor([[ 45,  27,  63],
        [144, 549,  72]])
                

Creating different types of pytorch tensors

Pytorch zeros tensor

t2 =torch.zeros([3,4])
t2
                

tensor([[0., 0., 0., 0.],
        [0., 0., 0., 0.],
        [0., 0., 0., 0.]])
                

Pytorch ones tensor

t4 =torch.ones([3,3])
t4
                

tensor([[1., 1., 1.],
        [1., 1., 1.],
        [1., 1., 1.]])
                

Pytorch random tensor

t3 =torch.randn([3,3])
t3
                

tensor([[-1.2749,  1.9119,  0.1540],
        [ 0.2169,  1.6054,  0.0332],
        [-0.2685, -0.8797,  2.6951]])
                

Simple join the tensor

x = torch.tensor([[45, 27, 63], [144, 549, 45]])
y = torch.tensor([[4, 5, 9], [5.4, 6.3, 9.1]])

t1 =torch.cat([x,y],dim=1)
t1
                

tensor([[ 45.0000,  27.0000,  63.0000,   4.0000,   5.0000,   9.0000],
        [144.0000, 549.0000,  45.0000,   5.4000,   6.3000,   9.1000]])
                

Mathematical operations

Simple Add two matrix

import torch
import numpy as np
x =torch.tensor([[32,45,86],[33,65,73]])
y =torch.tensor([[22,33,21],[45,33,35]])
x1=torch.add(x,y)
x1
                

tensor([[ 54,  78, 107],
        [ 78,  98, 108]])
                

Subtract two matrix

x2 =torch.tensor([[20,20,20],[20,20,20]])
x3 =torch.tensor([[45,66,45],[33,55,76]])
x4 =torch.sub(x,y)
x4
                

tensor([[ 10,  12,  65],
        [-12,  32,  38]])
                

Multipy two matrix

x5 =torch.mul(x,y)
x5
                

tensor([[ 704, 1485, 1806],
        [1485, 2145, 2555]])
                

Divide two matrix

x6 =torch.div(x,y)
x6
                

tensor([[1.4545, 1.3636, 4.0952],
        [0.7333, 1.9697, 2.0857]])
                

FashionMNIST dataset

In the we have 70,000 images.In the training data we have 60,000 images and the test data we have 10,000 images.

How we can load the FashionMNIST dataset from torchvision

We load the FashionMNIST Dataset with the following parameters:

1. root is the path where the train/test data is stored,
2. train specifies training or test dataset
3. download=True downloads the data from the internet if it’s not available at root.
4. transform and target_transform specify the feature and label transformations

import torch
from torchvision import datasets
from torchvision.transforms import ToTensor
training_data = datasets.FashionMNIST(
    root="data",
    train=True,
    download=True,
    transform=ToTensor(),
)
                

test_data = datasets.FashionMNIST(
    root="data",
    train=False,
    download=True,
    transform=ToTensor(),
)
                

Visualizing the Dataset

We use matplotlib to visualize some samples in our training data.

import matplotlib.pyplot as plt
labels_map = {
    0: "T-Shirt",
    1: "Trouser",
    2: "Pullover",
    3: "Dress",
    4: "Coat",
    5: "Sandal",
    6: "Shirt",
    7: "Sneaker",
    8: "Bag",
    9: "Ankle Boot",
}
figure = plt.figure(figsize=(8, 8))
cols, rows = 3, 3
for i in range(1, cols * rows + 1):
    sample_idx = torch.randint(len(training_data), size=(1,)).item()
    img, label = training_data[sample_idx]
    figure.add_subplot(rows, cols, i)
    plt.title(labels_map[label])
    plt.axis("off")
    plt.imshow(img.squeeze(), cmap="gray")
plt.show()
                

Conclusion

Firstly i learn what is pytorch than i performed the basic operation i.e convert python list to pytorch tensor then pytorch zero tensor then ones tensor than random tensor then perfomed simple mathematical operation. then i performed how we can load the dataset from torchvision. then i performed visulization on the dataset with the help of matplotlib to visulize some images in our training dataset.