01Py-basics.py

# -*- coding: utf-8 -*-
"""

Automatically generated by Colaboratory.



# **PyTorch (Basics)**

**PyTorch** is an optimized tensor library for deep learning using GPUs and CPUs.

**Torch** is an open-source machine learning library, a scientific computing framework, and a script language based on the Lua programming language.
"""

import torch

# define a number
t = torch.tensor(5.0)
t

# type variable
t.dtype

# vector
v = torch.tensor([1,2,3,4])
v

# matrix
m = torch.tensor([[1,2,3,4],[1,2,3,4]])
m

# 3 dimentional array
t1 = torch.tensor([[[1,2,3,4],[1,2,3,4]],[[1,2,3,4],[1,2,3,4]]])
t1

# shapes
print('number: ', t.shape)
print('vector: ', v.shape)
print('matrix: ', m.shape)
print('tensor: ', t1.shape)

# create tensor
x = torch.tensor(3.)
b = torch.tensor(4., requires_grad=True)
w = torch.tensor(5., requires_grad=True)

# arithmetic operation 
y = x * w   +   b
y

"""$y = 3x + 4$"""

# compute derivative
y.backward()

# gradiants
print('dy/dx =', x.grad)
print('dy/dw =', w.grad)
print('dy/db =', b.grad)

"""Using Numpy """

import numpy as np

# array
v = np.array([1,2,3,4])
v

# convert numpy array to torch tersor
y = torch.from_numpy(v)
y

# convert torch tersor to numpy array
x = y.numpy()
x