While working in the field of Deep Learning we always require a library or framework that reshapes and reimagines things. So the library which matches our list of requirements is PyTorch. It helps our researchers to accelerate their research in the field of Deep Learning by making them computationally faster, efficient and less expensive (dream of every Data Scientist!).
If I talk about myself, I personally found PyTorch of much use and not only me but all those who are into the field of computer vision all delve due to its flexibility and efficiency.
So in this article, I will give you a bit of an overview about how PyTorch works, so that you can take over from here and work on it from today itself.
Getting Started with PyTorch
As we move ahead in this article you will also agree from the above statement. But before we look into the subtle distinction of PyTorch, let’s look at some of the key features of this framework which make it unique and easy to use.
TorchScript
PyTorch TorchScript helps to create serial and optimizable models. If we train these models in Python. However, they can run independently from Python as well. When we will be in the deployment stage of the Data Science Model then it will help us a lot.
So, you can train a model in PyTorch using Python and then export the model through TorchScript to a production environment where Python is not available.
After going through this blog you will be able to easily understand the deployment of Models and also able to work on those Models.
Pytorch Distributed Training
Along with all these, PyTorch supports distributed training which enables researchers and practitioners to work together and also parallelize their computations. By distributing the training data into different sets makes it possible to use multiple GPUs to process larger batches of input data.
also, This reduces the computation time and increases our models efficiency.
Python Support
Interaction of PyTorch is somewhat similar to that of Python and also coding in Python is quite similar to that of Python. So if you have previously worked in Python, you are going to love working on PyTorch.
Tensors
They are the same as numpy multidimensional arrays. in addition to this Tensors are also used as a GPU to accelerate computing which decreases our execution time and make our model more efficient.
Tensors as a multi dimensional Matrices –
This will form a X by Y dimensional Matrix using Tensor that has been filled with random values.
A randomly selected Tensor can be created using PyTorch and the values are selected from Uniform Distribution between -1 and 1.
So here we will create a Tensor of 5X3:
Tensors have a size attribute that is used to check their size
Mathematical operations
Have you ever performed mathematical operations on NumPy arrays? If not, let me quickly give the overview of that.
Now we will quickly initialize two arrays that are ‘a’ and ‘b’ and then perform basic mathematical operations on them like addition, subtraction, multiplication, and division
Output : 2 1
Here are the two NumPy arrays ‘a’ and ‘b’ we have initialized. Now let’s try to perform mathematical operations on these arrays:
Output : 3
-1
2
2.0
Now let’s perform the same task in PyTorch and then compare both PyTorch and NumPy .
So, first, let’s initialize two tensors:
Output : tensor(2) tensor(1)
Now perform the above performed operations with PyTorch also:
Output : tensor(3)
tensor(-1)
tensor(2)
tensor(2)
Do you get the similarities between two?The codes that are written to perform mathematical operations in NumPy or PyTorch are almost same.
Now our next step is to see how to initialize a matrix along with how to perform matrix operations in PyTorch.
Matrix Operations
We will first initialize two matrices in NumPy:
Now we will perform basic mathematical operations on both the matrices
Output :
Concatenating Tensors
Let’s say we have two tensors as shown below:
Output :
Reshaping Tensors
Let’s say we have the following tensor:
Output :
We can use the .reshape() function and pass the required shape as a parameter. Let’s reshape the above defined tensor of shape (2,4) to a tensor of shape (1,8):
Output :
End Notes: On Pytorch
In this article, we get the basic ideas of PyTorch including it’s similarity with Numpy.
This article is the start of our journey with the library. Now we can easily learn how to build convolutional neural networks in PyTorch, model Check pointing techniques, and how to deploy trained deep learning models using the library.
Article by: Amit Shukla
If you are Interested In Machine Learning You Can Check Machine Learning Internship Program
Also Check Other Technical And Non Technical Internship Programs
