In this article, we will see some notable online cloud platforms that offer support for Jupyter Notebook which can be accessed as easily as a single click. There is no hassle of doing any installation or setup, hence it is very useful for beginners. But before going through the list let us first take a look at what is Jupyter notebook and why we need cloud platforms.
What is Jupyter Notebook
Jupyter notebook is a browser-based IDE for coding, visualizing, documenting and collaborating. It is very popular among data scientists and machine learning engineer community. Perhaps the main reason behind this popularity is the ease of documenting one’s work and sharing it with others. This is why it is called a notebook. It supports Python, R and Julia programming languages.
Jupyter notebook can be easily installed on your laptop or local workstation. The easiest way to set up is to install Anaconda which is a popular data science distribution for Python and R that offers Jupyter notebook IDE out of the box.
Why we need cloud platforms
Sometimes your laptop or computer configurations might not be good enough to train machine learning models efficiently with time. This is especially true when training big deep neural networks that might take too much time on your laptop CPU. Deep learning models train fast on GPUs (NVIDIA), but they are very expensive to purchase and set up on your laptop. Even if you purchase a GPU, you will find that very soon it will become outdated due to the fast-evolving computational hardware industry.
This is where GPU compute power offered by various cloud platforms can be very useful and cost-effective. The leading cloud platforms in this area are Google Cloud Platform and AWS EC2. But even these platforms require you to set up everything on their cloud as per your needs. It might not be easy for everyone to make their hands dirty in setting up a cloud platform, especially for beginners. A beginner should focus on machine learning and deep learning, not on setting up cloud platforms.
Do not worry, there are many cloud platforms who understand this gap and offer ready to use Jupyter Notebook on their cloud, which can be accessed by simply a single click. There is absolutely no set up needed. So let us take a look at these platforms.
Floyd Hub is a fully managed platform for training your deep learning models. It requires no setup and Jupyter notebook can be launched at a single click. It comes with all the popular python packages installed which include Scikit Learn, Pytorch, Tensorflow, and Keras.
They charge a fix monthly bill for using their platform plus the compute power that you have utilized. Their beginner plan has zero monthly billing and their compute plan starts with $1.20/hr for GPU and $0.48/hr for CPU.
Paperspace gradient offers one-click access to Jupyter Notebook on their cloud with zero setups. All the python libraries like Scikit Learn, Keras, Tensorflow, Pytorch, Caffe are preinstalled.
It too has a fixed monthly plan in addition to computing power billing. Their monthly plan starts for free for beginners. And the good thing is that they also offer free GPU for certain NVIDIA instances. Apart from this, they offer TPU (Tensor Processing Unit) compute power which is more powerful than GPU and can be used for training of some hardcore deep learning models.
Kaggle is a data science community platform that is very popular for hosting data science and machine learning competitions. It also offers Kaggle kernels which are Jupyter notebooks that come with preinstall python and R.
The integration of Jupyter notebook with its platforms makes it easy for Kaggle contest participants and other practitioners to work. And the good thing is that all this is free. It also offers free GPU compute power to spice up the training process of your models.
Colab is a google research product that hosts Jupyter notebook service without any need for setup or installation. It only supports Python currently and contains all the machine learning packages pre-installed.
Colab is free to use including their GPU compute power. They also offer TPU computing power which comes at a cost, however. The pricing of TPU starts at $1.35/hr.
It is worth noting that Colab’s free resources have certain limitations. It does not guarantee the availability of unlimited computational resources for their free plan. This dynamic fluctuation in serving computational resources helps them to meet demands.
Amazon Sagemaker is an AWS platform to deploy and train machine learning models where you can also run Jupyter notebook. AWS offers both CPU and GPU computation and has many billing plans depending on the type of resource you opt for.
AWS, however, does offer some free compute resources under their AWS Free Trial offers.
Disclaimer – The information provided here is to the best of our knowledge and research. We recommend you to please do your own research with respect to the service and pricing these platforms provide and if it suits your needs.