Training models increasingly requires processing power. GPUs have become increasingly necessary to serve different fields. But GPUs are not cheap either to buy or to subscribe to as a service. Me as a student, you or anyone who doesn't have the means to have one. We need to look for alternatives to use and be able to learn too, I don't know about your machine, but my laptop is simple and doesn't support so much processing, so having an alternative helps a lot.
In the options below I will leave the notebook I created as an example of execution, the notebook has no specific objective, just for use in the context of the post. Also for you to test.
Dear reader, if you can like and share I would greatly appreciate it, so I can see where my posts are going and bring you news.
Google Colab
I've been using Google Colab for a few years now, being simple you can work with Notebooks, being able to use CPU, GPU and TPU, all of which is free, of course there is a limit so use it in moderation, with it we can use it together with our Google account, shared files directly from Google Drive, this is very useful if you have very heavy image files and you can import them simply. Additionally, you can save notebooks to your account as well. Its interface is simple, containing a terminal, code snippets, keyboards and different themes. There are some notebooks available for you to start some examples.
Benefits:
- Short learning curve
- Has integration with Google account
- Wide range of features
- Darkmode
Disadvantages:
- Little variety of subscription and credit purchase plans
- Terminal only available for PRO account.
- You cannot create environments from scratch, Python and packages are already defined in the environment.
- There is no integration with Gitlab to save notebooks, only for Github
Kaggle
Kaggle is my favorite, I don't know about you but I like studying, within Kaggle I can combine many resources to learn. Starting with the community where you can search for new content, start discussions and other debates. You have a huge library of notebooks, datasets and models available for you to use. To use it is simple, just create a notebook and add by name which dataset or model you want to use, this is very dynamic. In addition to all this, you have short courses to learn the topics of machine learning and deep learning. You can use CPU, GPU and TPU, of course also with a limit. The most incredible thing about this is that it has competitions for teams and users to improve their skills.
Benefits:
- Short learning curve
- Environment with a large number of resources to use with notebooks
- Datasets, Models available with one click
- Courses
- Darkmode
Disadvantages:
- Responsive for small screens doesn't adapt very well, I have a laptop, but when I need to split the screen to read something next to the notebook, it becomes uncomfortable
- It takes a while to start the machine
- There is no ability to purchase more credits to use with GPUs
- Some keyboard commands do not work on the kaggle notebook
- There is no integration with Gitlab to save notebooks, only for Github
Kaggle
Kaggle Datasets
Kaggle Models
Amazon SageMaker Studio Lab
SageMaker Studio Lab, I found it complicated, to start you need to be approved after registering, I don't know if this has changed since I created the account. But you want to learn a new software and then run some notebooks and you can't because of that, then you have to check your cell phone which is another delay, you have to check if it will work, I hope it works for you, reader, well, out that. You have two CPU and GPU options to use on your notebooks, as well as a repository full of examples for you to learn about Generative Deep Learning, Computer Vision and NLP.
Benefits
- Short learning curve
- Simple and straight-to-the-point interface
- Example notebooks for training
Disadvantages:
- A lot of bureaucracy to create and verify the account
- Little resource on the dashboard and user profile
- No darkmode
Amazon SageMaker Studio Lab
Notebooks
PaperSpace
PaperSpace, my dear reader, I'll also apologize, I couldn't run a notebook, I even had to take a print of this one, because it's free, but when you instantiate the machine, you open a new box to add the card, the machine is free even though I marked it in the image. It is always unavailable, I ended up not being able to create any notebook, I hope you can. But don't be discouraged to check it out, as well as having CPUs and GPUs. You can create notebooks with different types of machines, whether blank notebooks or notebooks with libraries already defined, and you can also work with deploys, data, models, all in one place.
Benefits
- Short learning curve
- Wide range of features
- Wide variety of machines
- Darkmode
- Development-ready environments
- Good variety of plans
Disadvantages
- Free machine difficult to install, due to lack of availability
- There is no credit plan, for consumption only.
- Very low data storage option, example of the 50Gb growth option, I imagine situations with very heavy image and video sets, this could be a disadvantage, there may be an alternative. I'll research more, and if so, I'll update the post.
About the author:
A little more about me...
Graduated in Bachelor of Information Systems, in college I had contact with different technologies. Along the way, I took the Artificial Intelligence course, where I had my first contact with machine learning and Python. From this it became my passion to learn about this area. Today I work with machine learning and deep learning developing communication software. Along the way, I created a blog where I create some posts about subjects that I am studying and share them to help other users.
I'm currently learning TensorFlow and Computer Vision
Curiosity: I love coffee
My Latest Posts
Top comments (0)