Examples of Google Colab
Google Colaboratory or Google Colab is a Jupyter Notebook-based live-runtime environment that allows you to execute all your code purely on the cloud. It is intended to provide research and development of Machine Learning and Artificial Intelligence. Colab is a priceless resource available to any person who wants to explore the world of machine learning and data science but does not want to worry about hardware capacity. It has integrated support with popular machine learning libraries that include TensorFlow, PyTorch, NumPy and Pandas which are pre-installed, and makes them easy to set up and configure. Colab can be used by both students and data scientists and AI researchers and enables them to create and share their projects easily. Its usage with Google drive is so fluid in sharing, version controlling and real-time editing of the notebooks.
To connect to Google Colab
The only requirements to learn how to use Google Colab are having a Google account and a web browser. Google Colab website is available to visit at colab.research.google.com where you can sign in using your Google account. After you successfully log in, the subsequent creation of a fresh notebook can be achieved either by menu selection upon clicking on File > New Notebook or New Python 3 notebook. It is also possible to open Google Drive, GitHub projects or upload a notebook directly. When you create notebooks in Google Colab, they get automatically saved on your Google Drive and are stored in the sub folder Colab Notebooks. By clicking on the title, the default name of the notebook, which is usually, Untitled.ipynb can be renamed.
Interface Overview
A Google Colab document will have both text and code cells that give a user the live coding experience.
Code Cells: In code cells, you write and run Python code. One of the ways to run a cell is by clicking the play button to the left of the cell, or by pressing Shift + Enter. The execution results are shown below the cell in the form of output which aids in debugging and editing on-the-fly. The variables declared in one cell may be used in any other cell.
Text Cells Expressions can be added to a notebook using these cells, where documentation, explanations, or any text in Markdown format can be added. This feature is useful as it assists in organizing your notebook with titles, descriptions, and visualization. To add a new cell with text you can go to the top of the notebook and click on + Text or in the menu bar go to Insert > Text cell.
The toolbar: The toolbar has shortcuts to some common activities such as running cells and the addition of new cells.
Sidebar Tabs: A lot of helpful tabs can be found on the left-hand sidebar like:
The Table of Contents: Table of Contents automatically parses the titles of your text cells with the goal of creating an interactive table of contents that allows easy navigation, especially in notebooks that will be complicated.
Find and Replace: This is a tool that can be used to search and replace text in your notebook in case you need to make a quick correction or edit.
Variables: This tab gives us a summary of all the variables available in the running environment, their names and types in addition to their values that help in tracing the data and tracking the errors in programming.
This functions as a file manager that can be used to upload, download, as well as handling files related to the project by loading it through Google Drive or even the local systems.
Packages Installation in Saturn Python
Although Colab comes with most popular Python libraries used in deep learning and machine learning, you may find yourself in the situation when the library you need is not installed or a different version is required. The packages can be installed with the help of pip package manager by prepending an exclamation mark (!) to the command.
Files and Datasets
Google Colab has a variety of tools to manage files, and datasets efficiently, which is essential when working on machine learning models that use large datasets most of the time.
Downloading on Kaggle: In order to download datasets on Kaggle, it is required to generate a Kaggle API token on your account. Then, install Kaggle using the shell commands in Colab, create the API_keys directory, copy the kaggle.json file, set their permissions and copy the competition dataset.
Generic Website Downloading By using curl or commands in Colab, it is possible to obtain files from any site model using the prefix ! in front of the command. One can use browser extensions, such as Firefox client or CurlWget in Chrome to produce these commands. Keep in mind that any downloaded datasets in your Colab workspace will be destroyed when the runtime loses connection, hence it is a good habit to copy them to your cloud drive.
TPUs And GPUs
Deep learning is a computationally demanding technique, so Google Colab makes GPUs (Graphics Processing Units) and TPUs (Tensor Processing Units) available free of charge to counteract this.
Runtime Type: Switching to the GPU or TPU This can be done by opening the toolbar and changing the Runtime Type: enable GPU or TPU by going to Runtimesubsectionletting change runtime type then selecting either GPU or TPU in hardware accelerator dropdown.
Runtime Constraints: Colab provides free access to GPUs/TPUs, but there is a considerable chance that it will eventually run out, and its sessions can be set to time out because of inactivity after a few hours, about 12 hours at a time. With longer session capabilities and GPUs capable of working at a higher rate, customers can subscribe to Colab Pro at 10 dollars per month, which is currently accepted in the US and Canada. A good practice is to log off your notebook runtime when you have finished work so that other people can use resources.
Mathematics/State of the art-facilities
TensorBoard Integration: TensorBoard augments the machine learning datasets and enables visualization of the loss and accuracy over epochs as a TensorFlow toolkit. In order to monitor your model training and performance, you may use TensorBoard, which can be imported to the notebook by the extension and its callbacks must be defined in Colab.
Running shell commands: You can execute shell commands directly in a python code cell by placing an exclamation mark in front of the shell command. This will enable you to do things such as listing files (!ls) or cloning Git repositories (!git clone).
Avoiding Disconnection: In order to prevent disconnection during long training runs, one can use a piece of JavaScript in one of the cells with the %%javascript magic to have it continuously refresh the reconnect button.
Keyboard shortcuts can make the work flow much faster. Code That Ctrl + M + A adds a cell above, Code That Shift + Enter executes the current cell and Ctrl + M + B results in a new cell at the bottom.
Awesome learning of Python with Google colab using Courses
Some of the many free Python learning options are provided with the use of Google Colab.
OHSC Free Course on Google Colab: This course will help beginners to learn the fundamentals of Google Colab, and Python, and then how to manage data and perform machine learning on it. It will assist in equipping the learners with the working experience of a commonly used code environment and improve their coding and data science skills without the consideration of having a local environment.
Google Python Class: Google has a free course to people who know some other programming language and indeed want to learn Python. The course contains typed notes, video explanations, and practice code tasks that teach the fundamentals of Python, such as strings and lists, and proceed to more advanced topics where students write complete programs that work with files and make network requests. It is internal training material at Google, and it is freely available to individuals under the Creative Commons Attribution 2.5 license.
Uncodemy Python (Programming Language) Course Module: Uncodemy is an educational institution that offers courses on Python, data science, machine learning, etc. The different modules they have in Python include functions, Numpy, Pandas, Matplotlib, and Object-Oriented Programming (OOP). The instructors are well-trained instructors who possess knowledge about the industry.
